Boards should treat AI agent autonomy as a tiered, criticality-scored decision, not an all-or-nothing rollout, because network and workflow failures scale faster than governance teams can react. The consensus challenged here is the industry assumption, promoted by many automation vendors, that reaching TM Forum Autonomous Network Level 4 is primarily a technology milestone rather than a human-AI teaming and guardrail design problem. New research on criticality-based guard rail validation for AI agent decisions shows that a single ungated agent action, such as deactivating a zero-traffic cell that happens to be the sole emergency-services coverage, can create outsized operational risk. With the agentic AI market on pace to grow more than 40% annually through 2031 and carriers racing toward higher autonomy levels, the stakes for getting human-AI team design and agent guardrails right are rising every quarter.
Only 23% of carriers expect Level 4 autonomy by 2026, yet 85% target it by 2030
Criticality-based guard rails, not blanket human review, are emerging as the scalable safety pattern
EU AI Act high-risk obligations for embedded systems phase in through August 2028, reshaping telecom compliance timelines
1. Why Autonomous Telecom Networks Need Guard-Railed AI Agents Now
The complication is that most autonomy roadmaps assume uniform risk across all AI agent decisions, when in practice the consequences of an agent action range from trivial to catastrophic. Research on criticality-based guard rail validation for AI agent decisions in autonomous telecom networks demonstrates that treating "read a performance counter" the same as "shut down an emergency-services cell" is either too slow to be useful or too unsafe to deploy at scale. The paper's Guard Rail Validation (GRV) framework scores every AI agent decision across multiple risk dimensions and applies a proportionate level of validation, ranging from simple logging to multi-agent consensus and mandatory human sign-off, directly addressing a gap in current 3GPP and O-RAN autonomy discussions. Separately, enterprise security guidance on securing autonomous agentic AI systems confirms that access control, environment segregation, and monitoring are now treated as first-class requirements for any agent with write access to production systems, not optional add-ons.
The resolution is a layered governance model that pairs criticality-scored technical guardrails with explicit human-AI teaming design, so that autonomy expands only where evidence supports it. This means codifying which decisions autonomous agents can make unsupervised, which require asynchronous human review, and which demand real-time human-in-the-loop approval before execution, a structure closely aligned with the GRV framework's escalation tiers. Analysis of design patterns for safe agentic AI, including guardrails, policies, and human approval flows, shows that organizations adopting this tiered approach report fewer rollback events and faster incident resolution than those relying on either full autonomy or blanket manual review. For telecom operators and any enterprise deploying AI agents into mission-critical workflows, the practical decision is not "how autonomous should we be," but "which decisions earn autonomy, and under what guardrail tier."
2. The Evidence: Guardrail Frameworks, Market Scale, and Governance Gaps
On the governance side, the gap between agent capability and agent oversight is now well documented by independent research bodies. The Cloud Security Alliance's research note on the AI agent governance gap finds that most enterprises deploying autonomous agents lack formal criteria for escalation, audit trails, and cross-agent conflict resolution, echoing the exact failure mode that the GRV framework was designed to prevent in telecom networks. Network capital expenditure context reinforces why this matters financially: network capex represents 15% to 25% of revenue for major operators, totaling tens of billions of dollars annually worldwide, meaning that ungoverned agent errors touching network infrastructure carry proportionally large financial and reputational exposure.
The number one financial risk is ungoverned agent action inside high-capex network infrastructure, where a single unvalidated autonomous decision, such as an incorrect capacity reallocation or an erroneous cell shutdown, can cascade into service outages, SLA penalties, and regulatory exposure across a network base that already consumes 15% to 25% of operator revenue annually. Because criticality is unevenly distributed across the millions of decisions an autonomous network makes daily, operators that apply uniform guardrails either bottleneck routine operations or, worse, apply insufficient scrutiny to the rare high-impact decision, which is precisely the scenario the GRV framework was built to prevent. The number one financial opportunity is the inverse: proportionate, criticality-scored autonomy lets operators capture the bulk of the projected 55% operations and maintenance savings on low-risk, high-volume decisions while reserving human bandwidth for the small subset of genuinely high-stakes actions, effectively decoupling cost savings from risk exposure rather than trading one for the other.
3. MD-Konsult Research View
The prevailing consensus, reflected in TM Forum's own maturity roadmap and echoed by most systems integrators, is that reaching Autonomous Network Level 4 by a target date is fundamentally an architecture and AI capability milestone, achieved by building the right reference architecture, closed-loop automation, and intent-driven interfaces. MD-Konsult's contrarian position is that Level 4 autonomy without a criticality-scored guardrail and human-AI teaming layer is not a maturity achievement at all, but an unmanaged liability that most operators are currently underpricing.
Two data points support this position. First, the GRV research explicitly motivates its framework with a scenario where an energy-saving agent nearly disables a cell that is the sole emergency-services coverage for its area, a failure mode that a pure architecture-and-capability view of autonomy would not have caught, because the network was technically "autonomous" and functioning as designed until the decision context changed. This is documented directly in the criticality-based guard rail validation paper. Second, independent governance researchers at the Cloud Security Alliance find that the majority of enterprises running autonomous agents today, across industries including telecom, still lack the escalation criteria, audit trails, and conflict-resolution protocols needed to catch this class of failure before it happens, as detailed in the CSA research note on the AI agent governance gap.
Operators and enterprises that build criticality-scored guardrails and explicit human-AI teaming protocols now, ahead of the 2027 to 2028 EU AI Act high-risk enforcement windows, will be positioned to scale autonomy faster and with fewer costly rollbacks than peers retrofitting governance after an incident. Being early also creates a compounding advantage in regulatory relationships and customer trust, since demonstrable, criticality-tiered oversight is likely to become a competitive differentiator as autonomous network levels rise industry-wide.
4. Practitioner Perspective
"We do not resist autonomy because we distrust the models; we resist blanket autonomy because our incident reviews keep surfacing the same pattern, a low-probability edge case that a generic guardrail never anticipated. Criticality scoring changes the conversation from 'should this agent be autonomous' to 'which specific decisions has this agent earned the right to make alone,' and that distinction is what gets safety and operations teams to agree." — Head of Network Operations, Tier-1 Communications Service Provider
This practitioner view is grounded in survey findings from enterprise security researchers, who report in their analysis of agentic AI guardrails for safe scaling that access control, input validation, and staged autonomy expansion consistently outperform one-time authorization models in production environments, reinforcing why operations leaders favor decision-level rather than system-level autonomy grants.
5. Strategic Implications by Stakeholder
Stakeholder
What to Do Now
Risk to Manage
CTO / CIO
Map every AI agent decision type in network and workflow systems to a criticality tier using a framework modeled on guard rail validation research, and require multi-agent consensus for the highest tiers.
Deploying autonomy uniformly across domains without decision-level risk segmentation, creating single points of unmonitored failure.
COO / Operations
Redesign human-AI team workflows so operations staff review the small subset of high-criticality decisions in real time while low-risk decisions execute autonomously, following patterns described in safe agentic AI design guidance.
Alert fatigue and slow response times if all agent decisions, regardless of severity, are routed through the same human review queue.
CFO / Board
Fund guardrail and audit infrastructure as a capital efficiency investment, not a compliance cost, given that ungoverned agent errors in network infrastructure representing 15% to 25% of revenue can erase projected savings from autonomy-driven operational efficiency.
Treating guardrail investment as discretionary, then absorbing larger remediation and regulatory costs after an unguarded high-criticality failure.
6. What the Critics Get Wrong
Some technology leaders argue that criticality-based guardrails and human-AI teaming layers slow down the autonomy roadmap and undercut the operational savings that justified the investment in the first place, since manual escalation paths inherently add latency to decisions that autonomous agents could otherwise make instantly. This concern is not unfounded in naive implementations, where every guardrail routes to the same overloaded human review queue, and it echoes broader industry hesitation captured in analyses of AI adoption practicalities for telecom operators.
The direct rebuttal is that the GRV framework and comparable guardrail architectures are explicitly designed to avoid this bottleneck by scoring decisions individually rather than applying a single review standard network-wide, meaning the vast majority of low-criticality decisions still execute with zero added latency, as demonstrated in the criticality-based guard rail validation research. Independent governance analysis further shows that enterprises without any tiering system experience more rollback events and slower incident resolution than those using tiered guardrails, because undifferentiated autonomy eventually forces reactive, ad hoc human intervention at the worst possible moment, a dynamic documented in the CSA research note on agent governance gaps. In practice, tiered guardrails accelerate net autonomy rollout by making the high-risk minority of decisions safe enough to automate sooner, not later.
7. Frequently Asked Questions
What is the difference between AI agent automation and true network autonomy?
Automation follows human-defined rules for every action, while autonomy involves AI agents making independent decisions based on real-time context, a distinction TM Forum formalizes through its Autonomous Networks Levels framework, where Level 4 marks the shift from rule-based automation to genuinely independent decision-making across network domains.
How does criticality-based guard rail validation actually work?
The GRV approach scores each AI agent decision across multiple risk dimensions in real time, then routes the decision to the appropriate validation tier, ranging from simple logging for low-risk actions to multi-agent consensus or mandatory human sign-off for high-risk actions, as detailed in the original criticality-based guard rail validation paper.
What regulatory deadlines should telecom operators track for AI agent deployment?
Under the EU AI Act, prohibited AI practices have applied since February 2025 and general-purpose AI model obligations since August 2025, while high-risk standalone systems under Annex III face a conformity assessment deadline of December 2027, and embedded high-risk systems under Annex I face an August 2028 deadline, according to the EU AI Act compliance deadlines timeline.
How big is the market opportunity behind agentic AI in telecom?
The global agentic AI market is projected to grow from roughly $9.89 billion in 2026 to $57.42 billion by 2031 at a 42.14% compound annual growth rate, with IT and telecom named explicitly as a target vertical, according to Mordor Intelligence's agentic AI market analysis.
Why can't operators just require human approval for every AI agent decision?
Requiring universal human approval defeats the purpose of autonomy and does not scale against the volume of decisions modern networks generate, which is why criticality-tiered guardrails, rather than blanket review, are emerging as the practical safety pattern documented in safe agentic AI design pattern research.
What operational savings are at stake if governance is done right?
TM Forum research indicates mature autonomous network operations can deliver up to a 55% reduction in operations and maintenance costs alongside a 71% rise in customer satisfaction, benefits that are only fully realizable when guardrails prevent costly rollback events, per the TM Forum Autonomous Networks Mission overview.
Autonomous telecom networks are a
fundamental business model transformation that rewires how operators create,
deliver, and capture value. A network that self-configures, self-heals,
self-optimizes, and translates business intent directly into network
behavior is not a better version of what exists today. It is a different
kind of business: one where connectivity becomes programmable, where SLAs
can be guaranteed to millisecond precision, and where services that were
previously unscalable become commercially viable at any volume. The
operators who understand this distinction are already building towards Level
4 autonomous networks with a product roadmap as their guide. Those treating
it as a cost-reduction technology programme will arrive years late, with the
wrong architecture, and without the revenue model to justify what they
built. The
global autonomous networks market is forecast to grow from USD 8.53
billion in 2025 to USD 37.9 billion by 2033, and
90 percent of telcos already report AI is increasing revenue and reducing
costs, with autonomous networks ranked the top ROI use case in the industry.
Autonomous telecom networks are a prerequisite for the telco-to-AICO
transition: operators who achieve them will move intelligence, not just
bits, across their infrastructure, opening a new identity and revenue
model for the first time in twenty years.
What must be true to achieve them is specific and demanding: clean
unified data, cloud-native OSS, cross-domain AI agents, intent-based
orchestration, and a governance model that ties every automation
decision to a named business outcome.
The competitive clock cannot be paused: Telefonica now operates 12
Level 4 use cases in production, China Mobile has eliminated manual
fault resolution across its IP network, and Vodafone has reported over
500 million euros in combined savings. Every quarter of delay widens the
AI model experience gap and the talent gap simultaneously.
1. The Real Nature of Autonomous Telecom Networks
Most organizations describe autonomous telecom networks as networks that
automate themselves and that framing is technically accurate and strategically
misleading. Automation is a feature, while autonomy is a business model and the
distinction matters enormously for how capital is committed, how
architecture is designed, and what outcomes are measured.
An autonomous telecom network operates across four dimensions
simultaneously.
It self-configures: the network designs and deploys its own
optimal configuration in response to changing demand, spectrum conditions,
and service objectives without human instruction.
It self-heals: faults are
detected, diagnosed, and resolved through closed-loop AI decision-making
before service degradation reaches a customer.
It self-optimizes: the
network continuously adjusts resource allocation, routing, and energy
consumption to maintain performance against stated goals.
And, critically,
it is intent-driven: a network decision-maker or enterprise customer states
a desired outcome, a performance target, an SLA requirement, a quality of
service threshold, and the network figures out how to fulfil it,
continuously, across all domains from the radio access layer to the core and
service layer simultaneously.
Ericsson's canonical definition
of autonomous networks describes systems built on three pillars: agentic
closed-loop automation, intent-driven interactivity and interoperability,
and intelligent data management. Remove any one of these pillars and you
have a more efficient version of what operators run today, not an autonomous
telecom network.
The
TM Forum Autonomous Network Framework
provides the maturity model the industry uses to measure progress: six
levels from Level 0, which is fully manual, to Level 5, which is fully
self-evolving. Level 3 is conditional autonomy, meaning the network manages
itself within predefined rules but requires human approval for anything
outside them. Level 4 is highly autonomous, meaning the network acts on
intent without human intervention across complex, multi-domain scenarios
including those not explicitly anticipated during configuration.
The
financial difference between Level 3 and Level 4 is not marginal. Level 3
delivers perhaps half of the available OPEX savings, and none of the
revenue-side value.
As analysts observing MWC 2026 noted: automation reduces effort, but autonomy aligns outcomes. That is where
business model transformation begins.
The broader strategic frame being adopted by the most forward-looking
operators is the concept of AICO: AI infrastructure company.
Sebastian Barros of Circles, whose views were cited in the NVIDIA 2026
telecom AI survey, describes the transition directly: telcos will shift from moving bits
across networks toward moving intelligence across local and regulated
infrastructure. The customers will not just be humans, rather they will be AI
models, humanoid robots, autonomous machines, and hyperscalers (HCP). Every one of
them requires low-latency, guaranteed-SLA, programmable connectivity to
function. An autonomous telecom network is the infrastructure layer that
makes that commercially viable at scale. A traditionally managed network is
not.
2. What Must Be True to Achieve Autonomous Telecom Networks
The most important strategic question for any organization considering ATN
investment is not which vendor to choose or which use case to start with. It
is: what conditions must exist inside our organization before autonomous
networks can deliver at Level 4? This question is almost never asked with
sufficient rigor, and its absence explains the majority of program failures in the field today.
TierOne's CTO, presenting at MWC 2026, was direct: autonomy is not a software switch, it's an architectural
evolution. Real autonomous networks require five specific conditions to be
true simultaneously.
First, clean and actively reconciled inventory data:
the network cannot make autonomous decisions about resources it cannot
accurately see.
Second, cross-domain end-to-end service visibility: AI
agents operating in siloed domains cannot coordinate the multi-layer
responses that Level 4 autonomy requires.
Third, real-time service impact
intelligence: the system must be able to evaluate the customer-facing
consequence of any network event within milliseconds, not hours.
Fourth,
closed-loop AI-augmented orchestration: the feedback between observation,
decision, and action must be continuous and machine-driven, not
batch-processed and human-approved.
Fifth, measurable operational KPIs
linked to business outcomes, not technical metrics: without this link,
autonomy remains a technology ambition disconnected from the profit and loss
account.
Without all five, what organizations build is sophisticated
automation that stalls at Level 3. Detecon's autonomous networks transformation framework
identifies six organizational pillars that must be built in parallel, not
sequentially. Business value and use case prioritization must come first:
every initiative must begin with a defined business problem and a
quantifiable KPI, not a technology specification. Process digitalization
must remove manual workflows entirely, not layer automation on top of them.
Analytics, AI, and data management must be treated as infrastructure in
their own right: DataOps pipelines, ML model governance, and real-time data
availability are as critical as the AI models themselves. Cloud-native
architecture with standardised APIs must replace monolithic OSS systems:
microservices-based architectures are the only substrate on which
cross-domain autonomous decision-making can operate at the required speed
and scale. People, skills, and organization must be transformed: autonomous
networks require cross-functional teams combining network engineering, data
science, software development, and AI operations expertise in a single
accountability structure. And governance must enforce accountability at
every level, from individual AI agent decisions to board-level value
reporting. The operators who are succeeding at Level 4 have all six pillars
in motion simultaneously. Those who treat any one as optional will stall.
The data infrastructure requirement deserves particular focus because it is
the most common underinvestment point.
Google Cloud's autonomous networks architecture presented with Deutsche
Telekom and Vodafone at MWC 2026
identified the critical bottleneck explicitly: one of the biggest hurdles to
achieving Level 4 to 5 autonomy is manual delays caused by disconnected
legacy systems. The solution implemented was a unified graph data layer that
breaks down silos between operational and analytical data, enabling
real-time updates and deep historical pattern detection without slow ETL
processes.
Symphonica's white paper on intent-driven autonomous networks
is equally direct: legacy OSS architectures are not just suboptimal for
autonomous networks, they actively block them. Monolithic OSS lack the
modularity required for autonomous operations, and fragmented data prevents
AI models from operating with the reliability Level 4 demands. OSS
modernization is not a prerequisite that can be deferred until after the ATN
program is underway. It is part of the ATN program itself, and it must
be funded and staffed accordingly from day one.
3. The Evidence: Financial Performance and Market Momentum
The financial case for autonomous telecom networks has moved past modelling.
The
Capgemini Research Institute
surveyed operators globally and found that before most had reached Level 3,
they had already documented a 20 percent improvement in operational
efficiency and an 18 percent reduction in network OPEX. The study benchmarks
average investment of approximately 87 million USD over five years against
projected OPEX savings of 150 to 300 million USD, delivering an ROI of 1.7x
at the conservative end and 3.4x under optimistic deployment conditions,
with payback periods of 1.5 to 2.9 years.
Vodafone's autonomous network program has delivered over 500 million
euros in combined CAPEX and OPEX savings, a figure that elevates ATN from a technology cost-optimization discussion
to a balance-sheet strategic decision. Bharti Airtel has reported a 30 to 50
percent reduction in mean time to repair and a 20 to 25 percent reduction in
operational costs through autonomous AI-driven network and service
operations, according to
Forbes analysis of the 2026 telecom AI shift.
The operator cohort that has committed most fully to ATN is already seeing
the separation effects.
Telefonica closed 2025 with 12 fully operational Level 4 use cases
across Spain, Brazil, and Germany, including autonomous capacity creation,
digital twin-driven transport network optimization, and intelligent
multi-domain correlation for 4G and 5G simultaneously. The operator targets
an average autonomy level of 3.75 across its entire network by 2028 and full
Level 4 by 2030.
China Mobile achieved an 80 percent reduction in major network faults
after elevating its network operations centre from Level 3.2 to Level 4
using AI agents for fault management and customer complaint resolution.
Telstra achieved 30 percent faster time to market and 20 percent OPEX
reduction in 5G network slicing
through autonomous domain deployments. These are production outcomes
verified through the TM Forum assessment framework, not projected results
from a pilot program. The proof-of-concept phase for autonomous telecom
networks is over.
Metric
Value
Source
Global autonomous networks market (2025 to 2033)
USD 8.53B in 2025; forecast USD 37.9B by 2033 at 20.8% CAGR
Consensus position: The dominant framing in the market,
most explicitly stated in the
Bain and Company/TM Forum Autonomous Networks Reality Check published in
August 2025, defines ATN value primarily through the lens of OPEX reduction. The
report benchmarks success against a 30 percent cost savings target by 2028
and recommends operators begin with operational use cases like fault
management and service assurance before engaging revenue-generating domains.
This is sensible sequencing advice. It is not a sufficient strategic frame.
MD-Konsult position: The operators who will
define competitive advantage through 2030 are not building autonomous
networks to spend less. They are building them to sell something
fundamentally new: programmable, intent-based, guaranteed-SLA connectivity
as a service, at any scale, to any type of connected entity. The OPEX
savings are the financial case that gets the program funded. The revenue
transformation is the strategic case that determines whether the program was worth building at all.
Two data points validate this contrarian position. First,
TM Forum Inform's April 2026 analysis of NaaS strategy
states explicitly that NaaS represents a transformation of business models
rather than solely a technological advancement, and that it is the mechanism
through which 5G monetization becomes commercially viable at scale.
Operators who design their ATN architecture around fault management as the
primary use case will arrive at Level 4 without the intent-driven service
orchestration layer that NaaS requires. The architecture that delivers 30
percent OPEX savings and the architecture that enables programmable NaaS are
not identical. Choosing one without planning for the other is a multi-year
capital allocation error. Second,
SDxCentral's analysis of the NVIDIA 2026 telecom survey
cites the survey director directly: 80 percent of operators expect AI-native
networks before 6G, the industry is in a full investment and adoption phase,
and the question of quantifying ROI is a thing of the past. The operators
leading this shift are not asking whether ATN is worth building. They are
asking what their network must be capable of to compete in a market where
intelligence, not bandwidth, is the scarce and valuable resource.
The strategic advantage of being early on the full-value architecture path
compounds through time in a way that is not quickly reversible. AI agents
that manage autonomous networks improve through operational experience. That
experience takes the form of labelled training data, validated decision
patterns, and calibrated confidence thresholds across thousands of real
network events. It is not transferable between operators. An operator that
begins building these AI systems in 2026 will be operating more capable,
more reliable, and more commercially differentiated autonomous networks in
2029 than an operator who begins in 2028, because the early mover's agents
have been learning from live traffic every day that the late mover spent
evaluating vendor proposals.
5. Practitioner Perspective
"The hardest conversation we have with operator leadership teams is not
about technology. It is about identity. Autonomous networks are not a better
version of what you operate today. They are a different kind of business.
The question is not whether your network can reach Level 4 autonomy. It is
whether your organization can define a commercial product that only Level 4
autonomy makes possible, and then work backwards to build the architecture,
the data foundation, the operating model, and the governance structure that
product requires. Operators who start from the product question get there.
Operators who start from the technology question arrive at Level 4 with no
commercial home for what they built."
-- Director of Network Strategy Transformation, Global Systems
Integrator
This framing is reinforced by findings from the
Capgemini Research Institute, which found that only 17 percent of telcos have adopted a comprehensive
ATN transformation strategy with well-defined goals and measurable
timelines, and that only 15 percent have appointed a dedicated
cross-functional leader for autonomous network transformation. These two
factors, strategic comprehensiveness and dedicated ownership, are the
highest-correlation organizational predictors of faster deployment timelines
and higher realized ROI across the global operator sample.
The Fast Mode's 2026 analysis of autonomous network trends
describes five trends that will define the year: agentic AI transforming
operations into business-aligned systems; modular observability replacing
dashboard-based monitoring; autonomous assurance replacing reactive incident
management; cross-domain AI coordination replacing siloed automation; and
mature governance replacing experimental oversight controls. Each of these
trends is a prerequisite condition for Level 4 in production, not a
nice-to-have feature. Together they describe the operating environment that
only organizations who have done the prerequisite work can sustain.
6. The Regulatory and Standards Landscape
The regulatory environment for autonomous telecom networks is converging on
a set of requirements that make governance architecture, not just technical
architecture, a determinant of program success. The most immediate
constraint is the
EU AI Act, fully applicable from August 2, 2026. Autonomous network systems that make decisions with material impact on
service quality, infrastructure availability, or customer experience are
likely to qualify as high-risk AI systems under Article 6 and Annex III of
the Act. The practical obligations for Level 4 deployments include
documented conformity assessments before deployment, human oversight
mechanisms capable of intervening in or overriding AI decisions,
transparency documentation covering data governance and model performance,
and post-market monitoring for continuous compliance.
US-headquartered operators with European network operations face the same
compliance deadline
regardless of corporate domicile. The Act's requirements are not in tension
with ATN architecture. They define the governance architecture that
trustworthy Level 4 systems require regardless of regulatory mandate: an AI
agent whose decisions cannot be explained or overridden is not safe to
operate in a production network.
Standards bodies are converging simultaneously on a common technical
foundation that reduces deployment risk. The
ETSI ISG ZSM group
defines the cross-domain automation architecture and AI enabler
specifications. The
AI-RAN Alliance, now 132 members following MWC 2026, defines AI-native RAN architecture, bringing intelligence into the radio
access layer as an inherent property rather than an operational overlay. The
joint
GSMA, ETSI, TM Forum, ITU, and IEEE GenAINet self-healing autonomous
networks initiative launched in 2026
is producing AI model benchmarks and evaluation criteria that will likely
become the regulatory technical baseline in multiple jurisdictions. For
decision makers, the practical implication is clear: building to open
standards is not a vendor management preference, it is a program risk
management requirement. Proprietary architectures that cannot be assessed
against ANLET, ZSM, or AI-RAN benchmarks will face regulatory and
procurement scrutiny that slows deployment and increases compliance cost
over the program's life.
7. Strategic Implications by Stakeholder
Stakeholder
What to Do Now
Risk to Manage
Strategy and Technology Leadership
Begin not with a technology assessment but with a product question:
what service can you offer at Level 4 that you cannot profitably offer
today? Define that product first. Then use the
TM Forum ANLET framework
to identify the two or three domains where Level 4 capability is
achievable within 18 months and where it directly enables the defined
product. Appoint a dedicated ATN program leader with a mandate that
spans network, IT, and commercial teams simultaneously. Evaluate the
Ericsson and Nokia multivendor rApp cooperation
and the
Huawei AN L4 Phase 2 solution with the A2A-T agent protocol
against open-standards compliance criteria before committing to either
architecture.
Starting from fault management as the anchor use case and building an
architecture that reaches Level 4 in the NOC without the intent-driven
orchestration layer that NaaS and programmable enterprise connectivity
require. That is the most common and most expensive architectural
mistake in the field today and it is not correctable without
restarting major workstreams.
Operations and Delivery Leadership
Treat legacy OSS decommissioning as a first-class ATN workstream, not
a dependency to be resolved later.
AI-driven, cloud-native, no-code OSS is not an incremental
upgrade: it is the data substrate on which autonomous AI agents must
operate. Without it, Level 4 AI models will produce unreliable
decisions on incomplete data. Redesign NOC workflows around the
closed-loop model China Mobile deployed. Reframe workforce transformation around AI-augmented engineering
roles: organizations that lead with downsizing as the ATN narrative
create cultural resistance that the
Bain and TM Forum survey
identifies as one of the top three deployment blockers across the
global operator sample.
Building automation on top of legacy infrastructure rather than
replacing it. That approach produces Level 3 at best, at Level 4 cost,
because AI models cannot achieve production reliability without clean,
unified, real-time data across all domains simultaneously.
Finance and Board Leadership
Require a five-year ATN investment case, not a two-year pilot
justification. Use the
Capgemini 87 million USD investment against 150 to 300 million USD
in savings model
as the floor case and the Vodafone 500 million euro outcome as the
ceiling reference. Establish a cross-functional ATN governance
committee with quarterly reporting against business outcome KPIs
defined in the
TM Forum AN Framework. Require every ATN use case approval to include a named commercial
product or revenue line, not just an efficiency metric, before capital
is committed. Ensure EU AI Act compliance budgets are included for any
Level 4 deployment in
EU-regulated markets ahead of the August 2, 2026 deadline.
Under-scoping the investment because the case is framed only on OPEX,
and arriving at a network that is more efficient but not more
commercially differentiated. The operators who win the enterprise
connectivity market through 2030 will be those who used the OPEX
savings to fund the NaaS architecture. Those who stopped at the
savings will have funded their competitors' advantage.
8. What the Critics Get Wrong
The skeptical position deserves serious engagement. The
Bain and TM Forum reality check
documents real and persistent barriers: multi-vendor interoperability gaps,
the complexity of legacy OSS migration, structural AI talent shortages, and
the absence of clean cross-domain data in most operator environments.
Critics correctly argue that 84 percent of operators cannot reach Level 4
next year, that the organizational change required is genuinely difficult,
and that Level 3 conditional autonomy may deliver 80 percent of available
OPEX savings with materially lower integration risk. These are honest
constraints. Any capital allocation framework that ignores them will produce
overruns and missed timelines.
Where the skeptical position fails is in its assumption that the competitive
clock pauses while operators deliberate.
Industry analysts tracking agentic network operations in 2026
are unequivocal: early-mover operators are accumulating AI model training
data, cross-domain decision experience, and engineering talent that creates
advantages compounding with each quarter of operation.
TM Forum's April 2026 progress report
confirms that the pace of Level 4 production deployment is accelerating, not
slowing, as the first cohort of operators demonstrates that the barriers
Bain identified in 2025 are solvable with committed roadmaps. The data
infrastructure problem is a sequencing challenge. Operators who begin OSS
modernization and cross-domain data unification now will have the foundation
required for production Level 4 by 2027 to 2028. Operators who defer that
work will not be one year behind. They will be three or more, because the
prerequisite program itself takes 18 to 24 months once the decision to
commit has been made.
9. Frequently Asked Questions
Why are autonomous telecom networks described as a business model change
rather than a technology change?
Because the product they make possible did not exist before, and cannot be
manufactured with the previous architecture. A network that can provision a
guaranteed-SLA enterprise slice in under two minutes, autonomously assure
that SLA in real time, and expose its capabilities through open APIs is not
a faster version of traditional connectivity. It is a different commercial
offer entirely.
TM Forum Inform's April 2026 analysis of NaaS strategy
states this plainly: NaaS represents a transformation of business models,
not solely a technological advancement. The technology is what makes the new
model possible. The business model is why the technology is worth building.
What are the five conditions that must be true before Level 4 autonomy is
achievable?
Based on
TierOne's MWC 2026 analysis
and the
Detecon transformation framework, the five conditions are: clean and actively reconciled inventory data
across all domains; cross-domain end-to-end service visibility with no data
siloes between network layers; real-time service impact intelligence that
connects network events to customer outcomes in milliseconds; closed-loop
AI-augmented orchestration that acts without batch processing or human
approval gates; and measurable operational KPIs linked directly to business
outcomes, not technical metrics alone. If any one of these five is absent,
autonomous AI agents will produce decisions that are either unreliable,
unauditable, or commercially irrelevant, and the program will stall at
Level 3 regardless of how much has been invested.
How long does it realistically take to reach Level 4 in a specific domain
from Level 2?
Operators who commit cross-functional teams, address data infrastructure in
parallel, and focus on a single high-value scenario can reach Level 4 in
that domain within 18 to 24 months of program launch.
China Mobile's deployment with Huawei
is the clearest available benchmark: the operator moved from Level 3.2 to
Level 4 in its network operations center with all IP software faults now
resolved through closed-loop automation without human intervention. The
prerequisite work, specifically OSS consolidation and cross-domain data
unification, typically requires 6 to 12 months running in parallel.
Operators who sequence OSS modernization before the ATN program rather
than running both concurrently add 12 to 18 months to the total timeline
without any acceleration benefit.
What is the relationship between autonomous telecom networks and 6G?
ATN at Level 4 is both a prerequisite for and an accelerant of 6G readiness.
80 percent of telecom operators in the NVIDIA 2026 survey expect
AI-native networks to be deployed significantly before 6G, and the AI-native RAN architectures being built now under the
AI-RAN Alliance framework
will form the radio access foundation of 6G. Operators who achieve Level 4
autonomy before the 6G deployment cycle will have the AI infrastructure, the
data assets, and the operating model already in place to integrate 6G
capabilities as they become available. Those who have not will face two
simultaneous transformations rather than one sequential one, compressing
timelines and multiplying execution risk at the worst possible moment.
What does the EU AI Act require for Level 4 autonomous network deployments
specifically?
Level 4 autonomous network systems operating in EU jurisdictions must be
assessed against high-risk AI classification criteria under the
EU AI Act, fully applicable from August 2, 2026. Systems that make consequential decisions about service availability or
infrastructure configuration without human intervention are the archetype of
high-risk AI under Annex III. Compliance requires documented conformity
assessments before deployment, human oversight mechanisms capable of
overriding AI decisions, transparency documentation covering training data
governance and model performance records, and ongoing post-market
monitoring. These requirements define the governance architecture that
trustworthy Level 4 systems require regardless of regulatory mandate.
How should organizations structure the build versus buy decision for ATN
components?
The decision is not binary and should not be treated as a single
program-level choice. The data infrastructure layer, unified OSS,
real-time inventory, and cross-domain telemetry pipelines, is where building
internal capability matters most because it reflects each operator's
specific network topology and cannot be standardized across operators
without losing fidelity. The AI agent layer, covering fault management,
capacity optimization, and intent-based orchestration, is where proven
platforms add speed, with the
Ericsson and Nokia multivendor ecosystem
and
Huawei AN L4 Phase 2
both offering production-grade components assessed against open standards.
The governance and compliance layer must be built internally because no
vendor can accept accountability for how an operator's AI agents perform
against its specific regulatory obligations under the EU AI Act or
equivalent jurisdiction-specific requirements.
Breakthrough Business Models 2026: McKinsey's Six Archetypes, the Composable Enterprise, and Outcome-Based Pricing, Which Architecture Should Your Company Build?
Three converging business model shifts are creating the most significant structural reset in corporate strategy since the transition from perpetual software licenses to cloud subscriptions: McKinsey's six breakthrough archetypes, the composable enterprise, and outcome-based pricing. The conventional wisdom, championed by BCG and several traditional strategy consultancies, holds that these are technology-layer choices best delegated to the CTO or CIO. MD-Konsult Research argues the opposite: each of these is a board-level architecture decision with direct implications for revenue recognition, capital allocation, and competitive moat. Companies applying McKinsey's six breakthrough business models are achieving CAGRs above 15 percent, while Gartner projects that composable architecture adopters will outpace competitors by 80 percent in the speed of new capability deployment. Meanwhile, outcome-based pricing is erasing nearly $1 trillion in enterprise software market cap from vendors still clinging to per-seat models.
McKinsey's six breakthrough models deliver 15%+ CAGR and are explicitly designed to be transferable from Asia to any global market, yet fewer than 10% of Western enterprises have formally evaluated which archetype applies to their business.
Gartner has stated that 70% of enterprises are now mandated to adopt composable architecture by 2026, with the composable infrastructure market growing at a 51.3% CAGR through 2034 from a $14.16B base.
Outcome-based pricing will expand from 5% to 25% of all B2B contracts by 2028, representing a 5x increase that will force every B2B company to redesign its revenue model before buyers do it for them.
Interactive Brief
MD-Konsult.com | Interactive brief
Breakthrough business models: executive decision brief
McKinsey identifies six breakthrough business model archetypes shaping global growth: emotion-first products, network-driven commerce, microsegments and microproduction, the knowledge economy, conglomerates 3.0, and AI-native consumer platforms. AlixPartners adds that hybrid usage- and outcome-based models could account for up to 40 percent of software revenue by 2026, linking growth-model innovation to commercial-model redesign.
6Breakthrough archetypes identified by McKinsey across markets and sectors.
15%+CAGR threshold used by McKinsey for breakout growth models.
40%Share of software revenue AlixPartners says could move toward hybrid usage- and outcome-based models by 2026.
Archetypes
Operating design
Pricing logic
Executive implication
Archetypes
McKinsey presents six repeatable growth architectures rather than isolated company stories. The common pattern is that demand generation, trust formation, supply responsiveness, ecosystem design, and monetization are built into the operating model itself instead of being treated as separate functions.
Emotion-first products convert fandom, ritual, and identity into repeat demand.
Network-driven commerce compresses media, trust, and transaction into the same environment.
Microsegments and microproduction use rapid demand sensing and small-batch response to scale niche demand.
The relevance comes from the underlying architecture, not from geographic imitation. McKinsey’s framing shows that the transferable element is the system design: how trust is built, how supply adapts, how customer participation is structured, and how monetization is linked to value creation.
A new growth architecture usually requires changes to onboarding, pricing, fulfillment, analytics, and service delivery. Modular or composable operating design makes those changes easier to execute without rebuilding the whole enterprise stack for every strategic shift.
Growth-model innovation changes where and how value is created. AlixPartners' 2026 outlook indicates that commercial models are shifting toward usage and outcomes, which means pricing logic increasingly needs to reflect delivered value rather than simple access or seat count.
The business model question has returned to the boardroom with unusual urgency. After two decades in which strategic planning meant choosing digital channels, optimising unit economics within fixed architectures, and scaling SaaS subscription revenues, three simultaneous forces are demanding that executives rethink the underlying architecture of how their companies create and capture value. These forces are not independent experiments happening in adjacent industries. They are converging into a single structural reset, and the companies that address all three together will gain a compounding architectural advantage over those that treat any one of them as a departmental IT project.
In March 2026, McKinsey published its analysis of six breakthrough business model archetypes that have driven CAGRs above 15 percent for companies across Asia, with explicit guidance that each archetype is transferable to Western markets without requiring the government digital infrastructure or super-app ecosystems that characterise Asian markets. At almost exactly the same time, AlixPartners documented what it called "the second major pricing transformation in fifteen years" in enterprise software, as per-seat SaaS licensing is being systematically replaced by outcome-linked and consumption-based contracts. Alongside both of these, the composable infrastructure market reached $14.16 billion in 2025 and is projected to grow at a 51.3% CAGR to $588 billion by 2034, as enterprises dismantle monolithic application estates and replace them with modular, API-first architectures that can be reconfigured without rebuilding entire systems.
What makes 2026 a decision point rather than a planning horizon is that all three shifts have crossed the threshold from early-adopter experimentation into mainstream competitive pressure. Approximately $1 trillion in enterprise software market cap has been erased since mid-2025, with the IGV Software ETF down 22% year-to-date, as financial markets reprice the long-term value of per-seat licensing. Gartner has stated that 70% of organizations are mandated to acquire composable technology rather than monolithic suites by 2026, compared to 50% in 2023. And McKinsey's archetype research documents that early movers are already compounding network effects, data advantages, and trust architecture at a pace that late movers in 2027 and 2028 will struggle to replicate within the same cost envelope. The complication is not that the direction is unclear. The complication is that most Western enterprises have not yet translated these developments into a specific board-level architecture decision.
2. The Three Concepts Explained
2a. McKinsey's Six Breakthrough Business Model Archetypes
What they are and why they matter
McKinsey's six breakthrough archetypes are distinct structural approaches to building a business, each of which has been documented generating CAGRs above 15 percent across Asian companies of varying sizes and sectors. They are described as "strategic choices made by leaders, not outcomes unique to Asia's context" and are explicitly intended for global application. The four structural dynamics that created fertile ground for these models in Asia (scale and speed, system-level public-private collaboration, depth of digital services, and regulation as a catalyst) are enabling conditions but not prerequisites. Each archetype can be applied wherever the underlying customer logic holds.
The first archetype, Emotion-First Products, involves engineering emotional engagement as a core economic driver rather than treating it as a brand-building by-product. Companies applying this model industrialise the mechanics of emotional engagement including scarcity, product drops, fan rituals, collection loops, live co-creation, and rich IP narratives. Pop Mart's blind-box collectibles are the defining example: its Monsters line generated $1.9 billion in revenue in the first half of 2025 alone, representing a 180% trailing 12-month revenue CAGR, making it the largest company in its collectibles peer group. HYBE, the agency behind BTS, surpassed $1.6 billion in revenue in 2023 by building communities spanning offline concerts, digital content, and direct fan-to-artist interactions on its Weverse platform. The implication for global companies is that emotional engagement can be measured and monetised at scale through data on user-generated content velocity, community sentiment, and collection behaviour, and that IP designed around emotional loops generates predictable recurring spending that is structurally different from traditional brand loyalty.
The second archetype, Network-Driven Commerce, turns trust in creators and communities into the primary distribution channel, evolving well beyond influencer marketing into a complete retail system where purchasing occurs inside livestreams, short video, and chat environments. Douyin's e-commerce GMV grew approximately sevenfold in five years, from an estimated $75 billion in 2020 to $490 billion in 2024. TikTok Shop's Southeast Asian GMV grew from $4.4 billion in 2022 to approximately $16.3 billion in 2023. The lesson for Western markets is not to replicate Asian creators but to replicate their trust architecture: transparent product claims, live Q&A, local cultural cues, fulfillment clarity, and authentic category expertise. The foundations for creator-driven commerce exist in Western markets, but the potential remains largely untapped because most brands continue allocating budget to upper-funnel sponsorships rather than conversion-linked creator partnerships with SKU-level margin accountability.
The third archetype, Microsegments and Microproduction, is about matching supply to demand at an extreme pace by delivering small-batch or personalised products at unit costs previously associated with mass production. Shein is the defining example: the company generates revenue growing at approximately 29% annually (from $23 billion in 2022 to $38 billion in 2024) by testing new designs in batches of 100 to 200 units compared to Zara's 300 to 500, listing up to 10,000 new SKUs per day, and moving designs from concept to production in as few as five days. For companies outside fashion, the applicable principle is using demand-sensing infrastructure and small-batch experimentation to identify high-margin micro-niches before committing to full production scale, which requires modernising supply chain systems around real-time data rather than periodic forecasting cycles.
The fourth archetype, The Knowledge Economy Model, involves treating free, high-quality education not as corporate social responsibility or marketing spend but as a genuine customer acquisition channel that builds trust and reduces acquisition costs in categories where customers have far less information than providers. Zerodha became India's largest stockbroker by building Varsity, a comprehensive ad-free educational platform covering markets and investing, which allowed it to acquire high-value customers at 12 to 13% of revenue versus competitors' 20 to 23%. Groww, which overtook Zerodha as market leader with over 12 million active users, reports that approximately 80% of customers find the company organically. ITC Limited's e-Choupal initiative has served four million customers across 35,000 villages over more than 20 years using shared market prices and agronomy advice as the acquisition vehicle. This model is directly applicable to any Western company operating in a trust-sensitive sector including finance, healthcare, insurance, energy, and professional services.
The fifth archetype, Conglomerates 3.0, replaces the traditional logic of pooling capital and management expertise across unrelated businesses with a model in which multiple verticals are integrated through shared digital assets such as identity platforms, payments infrastructure, data lakes, and loyalty systems. Ping An tracks and reports that a quarter of its 242 million retail customers hold four or more contracts across its ecosystem spanning financial services, healthcare, auto services, and smart-city solutions, all integrated through a unified identity and data platform with approximately 95% year-to-year customer retention. Reliance in India doubled its annual revenues over eight years to over $100 billion by integrating telecommunications, content, commerce, payments, and offline retail through shared digital assets. The board-level implication is that diversified companies should evaluate their portfolio not by asking whether business units share physical assets or management talent, but by asking whether shared digital assets (customer data, identity, loyalty, payments) can create a compounding cross-selling advantage that a standalone competitor cannot replicate.
The sixth archetype, AI-Native Consumer Platforms, encompasses services delivered primarily or entirely by AI rather than human labour, at near-zero marginal cost per additional user. AI tutors are mainstream in China, India, and South Korea: Zuoyebang alone reports more than 170 million monthly active users receiving personalised practice sessions adjusted in real time to student behaviour. Yuanfudao, valued at over $15 billion, grew device sales by 120% year-over-year in third-tier Chinese cities. The economic logic is that services which previously required human labour proportional to the number of users can now scale to unlimited users at near-zero incremental cost, converting what was a variable cost into a fixed infrastructure investment. For Western companies, AI-native services in customer advisory, compliance guidance, technical support, and styling are the most immediate applications of this archetype.
2b. The Composable Enterprise and Composable Architecture
What it is and why it matters
A composable enterprise is a business designed around interchangeable, independently deployable building blocks called Packaged Business Capabilities (PBCs), rather than large monolithic applications with hard-coded dependencies between functions. Gartner defines a composable business as "an organization that delivers business outcomes and adapts to the pace of business change through the assembly and combination of packaged business capabilities." In practical terms, this means that functions such as payment processing, customer onboarding, order fulfilment, fraud detection, pricing, and inventory management each exist as discrete, API-accessible modules that can be combined, replaced, or scaled without redesigning the surrounding system. The composable enterprise is simultaneously a business model decision and a technical architecture decision, because the organizational structure (which team owns which capability, how delivery units are structured around modules rather than applications) must align with the technical design for the speed benefits to materialise.
Composable architecture rests on three foundational principles that distinguish it from both traditional monolithic systems and simple microservices. The first is modularity, which means that data and business functions are organised into small, discrete units that can be used independently or combined to create new services without modifying the entire surrounding system. The second is autonomy, meaning that each component operates independently as a defined business capability while still functioning as part of the broader platform, eliminating the cascading failure risk that monolithic systems create when a change to one module forces untested changes across dependent modules. The third is interoperability, meaning that data flows between components through open APIs and standardised contracts without requiring manual IT intervention, so that a change in the payment module does not require a rebuild of the inventory or customer onboarding modules. Together, these three principles allow organizations to reconfigure business processes in response to a new market opportunity, a regulatory change, or a pricing model transition without paying for a full platform rebuild each time. Coderio's April 2026 guide to composable enterprise architecture describes this as replacing "the idea of one system doing everything with the idea of many bounded components doing specific work well."
Composable infrastructure, which is the physical and cloud-level complement to composable application architecture, takes this logic one layer deeper. CIO Magazine's March 2026 analysis defines composable infrastructure as a catalogue of independently addressable compute, storage, network, and security building blocks that can be assembled on-demand through policy-defined software rather than requiring physical reconfiguration of hardware or migration between vendor-locked cloud stacks. Instead of purchasing a single closed technology stack from a single vendor, enterprises assemble best-in-class components that interoperate through open standards, where each piece is replaceable, independently optimised for its role, and auditable as a distinct unit. The composable infrastructure market was valued at $14.16 billion in 2025 and is projected to grow to $588 billion by 2034 at a 51.3% CAGR, driven by financial services, healthcare, retail, and government sectors deploying composable infrastructure to support DevOps, AI workloads, and hybrid cloud strategies, with North America accounting for 48.1% of current global adoption.
The strategic relevance of composable architecture for the board is not about IT modernisation for its own sake. It is about purchasing the right to change direction quickly without paying for a complete system rebuild each time a market condition, pricing model, or regulatory requirement changes. When a company wants to introduce outcome-based pricing, for example, the billing and metering logic can be updated within an isolated PBC without touching the CRM, inventory, or service delivery systems. When a company wants to apply the microproduction archetype and introduce demand-sensing at the SKU level, it can connect a new demand-forecasting module through the existing API layer without re-engineering the entire supply chain system. This is why Gartner's mandate that 70% of enterprises adopt composable technology by 2026 is correctly interpreted as a board-level capital allocation decision about strategic agility rather than a CIO decision about software procurement.
2c. Outcome-Based Pricing
What it is and why it matters
Outcome-based pricing is a revenue model in which customers pay only when a specific, pre-agreed business result is achieved, rather than paying for access to a product (per-seat licensing), for time spent by the vendor (hourly billing), or for volume consumed (usage-based pricing). Metronome's definition captures the essential distinction: "Instead of paying a flat fee for software access, customers pay for tangible results, like successfully resolved sales or support conversations." The outcome is binary, verifiable, and defined in advance by both parties, so that both the buyer's incentive and the vendor's incentive align around delivery of the same result. This is categorically different from usage-based pricing, which charges for inputs (API calls, compute cycles, storage volume), and different from subscription pricing, which charges for access regardless of whether value is delivered.
The structural logic that is driving adoption of outcome-based pricing is the same logic that is eroding per-seat SaaS: when AI agents can perform work equivalent to hundreds of human employees at near-zero marginal cost, the connection between headcount (seats) and value delivered is severed. Accenture's April 2026 analysis describes this as a shift from "software as a user interface to software as a system of orchestration," arguing that the platforms which endure will be those that become systems of execution where AI owns outcomes end-to-end, rather than those that remain interface-centric access tools. Gartner projects that 40% of enterprise applications will include task-specific AI agents by the end of 2026, up from less than 5% the prior year, which means the majority of enterprise software platforms will have separated the number of users from the amount of work actually performed within a two-year window. In this environment, pricing on seat count is no longer defensible to a CFO who can directly observe the ratio of work output to licensed headcount.
In practice, outcome-based pricing takes several forms depending on how cleanly the vendor controls the execution path. A pure outcome model charges only on verified delivery of the defined result, with no floor payment and no access fee. A hybrid model, which is the structure that CRN's February 2026 channel research identifies as the dominant early-mover structure, combines a base platform or access fee that provides vendor cash-flow predictability with a performance bonus component tied to a verifiable outcome metric. RiskSpan's announced 2026 pricing structure of fixed fee plus bonuses based on performance outcomes is the most cited current example in the IT channel. McKinsey has stated that a quarter of its own fee pool is now staked on measurable client outcomes, providing one of the most credible signals that the model is viable at scale in complex professional services engagements. The critical design pre-condition for any outcome contract is that the vendor must own the last mile of the execution path: the model works where the vendor controls or observes the complete execution path from input to verified outcome, and it fails contractually wherever the outcome is influenced by factors the vendor cannot observe or reasonably control.
3. The Evidence
The financial performance data behind McKinsey's six archetypes spans asset classes, geographies, and company sizes in a way that makes the "Asia-only" objection difficult to sustain. Pop Mart's Monsters IP generated $1.9 billion in revenue in the first half of 2025 at a trailing 12-month CAGR of 180 percent, making it the largest company in its collectibles peer group by a substantial margin. Douyin's e-commerce GMV grew approximately sevenfold between 2020 and 2024, from an estimated $75 billion to $490 billion, while TikTok Shop's Southeast Asian GMV nearly quadrupled in a single year. Shein's microproduction architecture drove revenue from $23 billion in 2022 to $38 billion in 2024 at a 29% annual CAGR while remaining profitable on batch sizes of 100 to 200 units and a design-to-production timeline of as few as five days. Zerodha and Groww demonstrate the knowledge economy model generating customer acquisition costs of 12 to 13% of revenue against competitors at 20 to 23%. These are not experimental pilot programmes; they are the dominant companies in their respective categories, and they arrived at those positions through deliberate model architecture choices that McKinsey's research explicitly describes as replicable.
The composable enterprise data adds a parallel dimension of structural urgency. Gartner's research documents that composable architecture adopters achieve 80% faster new feature and capability implementation than non-composable peers, which in a market where AI-native entrants can ship new capabilities weekly translates directly into competitive positioning. The composable applications market was $7.84 billion in 2025 and will reach $38 billion by 2035 at a 17.1% CAGR. On outcome-based pricing, Kyle Poyar's research shows 25% of B2B companies expect to use outcome-based pricing by 2028, up from approximately 5% today, representing a 5x increase within three years. The market cap destruction is the sharpest financial signal available: approximately $1 trillion in enterprise software value has been erased from companies still pricing on per-seat models, reflecting financial markets pre-emptively repricing the long-term sustainability of the SaaS era's core assumptions.
Metric
Value
Source
CAGR of companies applying McKinsey's six breakthrough archetypes
Above 15% CAGR across documented cases; Pop Mart emotion-first model achieved 180% trailing 12-month revenue CAGR in H1 2025
Enterprise software market cap erased from per-seat model repricing
Approximately $1 trillion erased since mid-2025; IGV Software ETF down 22% year-to-date 2026; ServiceNow down 50% from highs; Salesforce and Workday each off 40%
The consensus position, articulated most clearly by BCG's November 2025 portfolio strategy report and echoed across the major consultancies, holds that business model innovation is an iterative process: companies should optimise their existing model first, experiment at the edges, and adopt new architectures only when competitors have proved them viable and the transition risk can be precisely quantified. BCG's framing treats model redesign as inherently capital-intensive and suitable only for companies with the balance-sheet capacity to absorb transformation costs without disrupting quarterly earnings cadence.
MD-Konsult Research's position is that the sequencing consensus has it exactly backwards: the greatest risk in 2026 is not the capital cost of redesigning your business model architecture, but the 18-month window of compounding structural advantage that you permanently surrender by waiting for others to prove the transition first.
The strategic implication of being early is asymmetric in the direction of the early mover. Companies that begin the composable transition in 2026 accumulate a capability velocity that late movers in 2027 and 2028 cannot replicate within the same cost envelope, because each new service or market added to a composable architecture costs progressively less to integrate, accelerates faster to deployment, and generates more reusable data than the equivalent addition inside a monolithic structure. On outcome-based pricing, the first movers define what "success" means in their category before buyers, regulators, or AI-native competitors define it for them, retaining the definitional power that late movers permanently surrender.
5. Practitioner Perspective
"We made the decision to move from fixed-fee project work to outcome-linked contracts two years ago, and every operational instinct at the time said it would compress near-term revenue and create attribution disputes with clients. What actually happened was the opposite: customers who pay for results stay longer, expand their engagement faster, and refer new clients more consistently, because their incentive and ours are aligned around the same defined result rather than around maximising billed hours. The composable architecture question is structurally identical. You are not spending capital on flexibility for its own sake; you are purchasing the right to change direction quickly without paying for a full system rebuild each time the market, a regulation, or a pricing model shifts. Boards that frame either of these as cost decisions are asking the wrong question. The correct question is what does it cost your business to be structurally slower than a new entrant that has made these decisions from day one."
--- Chief Strategy Officer, Global Professional Services Firm
This perspective is grounded in the operational evidence from the technology channel. CRN's February 2026 investigation into outcome-based business models documents multiple solution providers transitioning away from hourly billing, with Snowflake CEO Sridhar Ramaswamy explicitly endorsing outcome-driven partner models as "what is going to drive great results" and committing to pass on Snowflake's in-house services learnings to help partners adopt the model. RiskSpan, a Snowflake and AWS partner, announced a 2026 pricing structure of fixed fee plus bonuses tied to performance outcomes, a hybrid model that directly addresses the cash-flow risk of pure outcome pricing while retaining the alignment signal that enterprise buyers increasingly demand before renewing large contracts. The convergence of practitioner behaviour, CEO endorsements from major platform vendors, and financial market signals creates an unusually consistent picture of a shift that is already underway at the partner and services layer, ahead of the broader enterprise software market.
6. Regulatory Landscape
The EU AI Act is the most consequential active regulation affecting all three model transitions simultaneously. Prohibited AI practices have been subject to penalties since February 2025, and full requirements for high-risk AI systems take effect August 2, 2026, with US companies that have EU market exposure facing identical compliance obligations as EU-domiciled firms. For composable architectures, the Act's data lineage and auditability requirements are directly relevant to architectural design decisions: modular systems must maintain centralised, immutable audit logs across all components to satisfy both GDPR data flow requirements and the AI Act's traceability obligations for systems that make or influence decisions about individuals. For outcome-based pricing contracts, the Act's explainability requirements create a new form of contractual complexity: where a defined outcome is determined or measured by an AI system, that system's decision logic must be explainable to the buyer and potentially to regulators, which reinforces the case for hybrid pricing models with transparent, manually verifiable measurement definitions rather than purely AI-scored outcomes.
In the United States, the regulatory environment is less prescriptive but not absent. The SEC's evolving disclosure rules around material digital transformation investments create pressure on CFOs to articulate the strategic rationale and expected financial return for large-scale composable architecture transitions. The FTC has increased scrutiny of AI-enabled pricing practices, particularly where dynamic or outcome-based pricing could be demonstrated to disadvantage specific buyer categories. GEP's April 2026 enterprise AI regulation guide identifies data governance, model risk management, and contract transparency as the three areas generating the most active regulatory correspondence between enterprises and US federal agencies. The 2026 regulatory roadmap from ComplyAdvantage identifies operational resilience requirements, third-party AI risk management, data localisation obligations, and model governance as the four themes cutting across all three model transitions, reinforcing the case for companies adopting all three shifts in an integrated way rather than independently.
7. Strategic Implications by Stakeholder
Stakeholder
What to Do Now
Risk to Manage
CTO / CIO
Audit the current application portfolio against the composable framework: identify which core systems can be decomposed into Packaged Business Capabilities (PBCs) accessible via API, and prioritise the two or three capabilities where speed-to-change is most commercially valuable. Run a composable pilot within 90 days on a customer-facing function where the competitive cost of structural slowness is highest. Evaluate which of McKinsey's six archetypes most naturally extends the existing digital infrastructure and what measurement systems would need to be added to support an outcome-based pricing transition.
Vendor lock-in on monolithic platforms that do not expose APIs and will require costly exit negotiations when the composable transition begins. Data flow documentation gaps that will create compliance exposure when the EU AI Act's August 2026 high-risk system requirements take effect. Workflow portability risk if AI model infrastructure is built around a single provider's proprietary ecosystem rather than open standards.
COO / Operations
Map existing service delivery processes against outcome-based pricing logic by identifying where your organization already owns the last mile of a specific, verifiable result and can define a binary, observable success metric. Pilot one outcome-linked contract with a reference customer before year-end 2026, establishing the measurement baseline, attribution methodology, and success definition before procurement teams begin demanding it as a standard contract term. Apply the microproduction archetype thinking to any supply chain function where batch-size reduction and real-time demand-sensing can reduce inventory carrying costs while improving margin per unit.
Attribution complexity in multi-factor service environments where outcomes are influenced by variables outside the vendor's control, which must be addressed through contract design rather than through technology alone. Cash flow timing mismatch if outcome payments are structured to arrive quarterly or annually rather than against defined milestones. Internal change management resistance from delivery teams whose performance metrics are currently built around activity (hours, tickets, outputs) rather than customer results.
CFO / Board
Reframe the business model architecture decision as a capital allocation question with an explicit late-mover scenario: commission a scenario analysis that includes a base case (current architecture maintained), an early-mover case (composable transition plus one McKinsey archetype adopted by 2027), and a late-mover case (an AI-native competitor enters the primary revenue category with outcome-based pricing within 24 months). Initiate a pricing strategy review that distinguishes between usage-based, output-based, and true outcome-based contract structures, and defines where each is appropriate across the revenue portfolio. Ensure the audit committee is briefed on the EU AI Act August 2026 compliance deadline and its implications for any AI-enabled product or service with EU market exposure.
Revenue recognition complexity under IFRS 15 and ASC 606 when outcome-linked payments are spread across multi-period contracts and when the timing of outcome verification is uncertain. Investor skepticism about near-term margin compression during a composable transition that is correctly a multi-year programme rather than a single-quarter cost event. Forecast accuracy deterioration if outcome-based contracts create revenue timing variability that exceeds what financial planning and analysis teams can model accurately against quarterly guidance commitments.
8. What the Critics Get Wrong
The strongest version of the opposing argument appears in Forbes and Parloa's January 2026 analysis and in Moneycontrol's April 2026 commentary on outcome pricing in AI. The core objection runs as follows: outcome-based pricing is structurally flawed because outcomes are multi-factorial and unattributable, composable architecture imposes coordination costs between module teams that erode the theoretical speed gains, and McKinsey's Asian archetypes rely on structural conditions (government digital infrastructure, super-app ecosystems, user density at a national scale) that simply do not exist in Western markets. The Parloa piece specifically argues that outcome-based pricing converts software vendors into underwriters, asking them to take on balance-sheet risk for factors they cannot observe or control, and that the cash-flow timing problem alone makes it unsuitable for most B2B businesses with quarterly investor reporting obligations. These objections deserve a genuine hearing, and they explain why large incumbents including Salesforce and Microsoft have not yet fully adopted outcome models for their core product lines.
The rebuttal is supported by the evidence available in 2026 rather than by theoretical arguments. On composability, 80% of enterprises have already adopted or are actively evaluating composable architecture, which indicates that the coordination cost objection has been tested at scale in production environments and found manageable, with the key finding being that coordination costs are highest during the initial decomposition of existing monolithic systems and then decline as each new service added to a composable estate is born modular rather than requiring decomposition. On outcome-based pricing, the hybrid model structure that combines a base subscription fee with an outcome bonus tier directly addresses both the attribution and cash-flow problems that the critics identify, and is already the dominant structure in the IT channel's early-mover cohort rather than a theoretical proposal. On McKinsey's archetypes and their geographic preconditions, the report's own data is the clearest rebuttal: Zerodha's knowledge-economy model required no government super-app and no national-scale user density, only a decision to treat education as an acquisition channel, and Pop Mart's emotion-first model required no cultural conditions unique to China, only a decision to engineer emotional engagement as an economic driver rather than a brand by-product.
9. Frequently Asked Questions
Which of McKinsey's six breakthrough business models is most immediately applicable to a Western enterprise with no presence in Asian markets?
The knowledge economy model is the most directly transferable because it requires no government digital infrastructure, no super-app ecosystem, and no national-scale user density. The model's entire logic rests on a single strategic decision: allocate part of the customer acquisition budget to producing free, high-quality educational content that builds trust in a category where customers have significantly less information than the vendor. McKinsey's evidence from Zerodha and Groww shows customer acquisition costs of 12 to 13% of revenue against competitors at 20 to 23%, a structural advantage available to any company operating in finance, healthcare, insurance, energy, or professional services. The emotion-first model is the second most immediately applicable: the mechanics of emotional engagement through scarcity, collection loops, drops, and fan rituals can be designed and deployed in any consumer-facing category using data infrastructure that already exists in Western markets.
What is the most risk-controlled way to structure an outcome-based pricing contract without exposing the company to uncontrollable external variables?
The evidence-backed approach is the hybrid model: a base platform or access fee that covers vendor cost of delivery and provides cash-flow predictability, combined with a performance component tied to a single, binary, frequently measurable outcome that the vendor directly controls end-to-end. CRN's channel research documents RiskSpan's fixed fee plus performance bonus structure as the current reference model, and Kyle Poyar's practical guidance is to start with one outcome, one customer, and one verifiable metric with a pre-agreed observation window and attribution methodology before scaling the contract structure across the portfolio. The pre-condition that cannot be shortcut is owning the last mile of the result: the model is contractually viable where the vendor controls or observes the complete execution path from input to verified outcome, and it breaks down wherever the outcome is influenced by customer behaviour, third-party systems, or market conditions outside the vendor's reasonable control.
How should a company begin the composable architecture transition without creating operational disruption to current revenue-generating systems?
The recommended approach is a "strangler fig" decomposition: begin by identifying the two or three business capabilities where speed-to-change is most commercially valuable, typically customer-facing functions such as commerce, service delivery, personalisation, and billing, and expose those capabilities as API-first Packaged Business Capabilities running in parallel with the existing monolithic core. Coderio's April 2026 composable enterprise guide describes this as replacing the monolith incrementally rather than in a single migration event, with the composable layer expanding as each module proves stable and the monolithic dependencies shrink. Gartner's "fusion team" model, which blends technology architects and business domain owners into multidisciplinary capability units, is the organizational complement to this technical approach and is documented as the standard delivery structure for successful composable rollouts at production scale.
Is the per-seat SaaS model dying uniformly across all enterprise software categories, or are there categories where it remains the appropriate pricing structure?
The per-seat model is not dying uniformly. It remains appropriate and defensible in categories where individual user engagement is a genuine and observable proxy for value delivered. Moneycontrol's April 2026 analysis correctly notes that ChatGPT, Claude, Cursor, and Microsoft Copilot all charge per user because the value those tools deliver scales directly with the number of users adopting them as productivity tools, making seat count a reasonable proxy. What is dying is the per-seat model in categories where AI has structurally severed the relationship between licensed headcount and work actually performed, covering contact centre automation, document and contract review, code generation, and customer service, where a single AI deployment can perform the work of dozens or hundreds of human users simultaneously. Gartner projects global software spending at $1.43 trillion in 2026 but warns that approximately 9% of that increase is a hidden AI tax from vendors raising prices on products without delivering proportional new value, which is the market signal that the per-seat model is being stress-tested most aggressively in precisely those categories.
How do the three model shifts interact, and is there a recommended sequencing for companies that want to adopt all three?
The three shifts are mutually reinforcing rather than independent, and there is a logical sequencing that maximises the return on each transition. Composable architecture should come first because it creates the technical and organizational substrate that makes both the McKinsey archetypes and outcome-based pricing operationally feasible without requiring full-system rebuilds for each implementation. Without composable architecture, introducing a new pricing model requires rebuilding billing, metering, and reporting infrastructure across the entire application estate rather than updating a discrete, API-accessible billing PBC. IBM's composable architecture guidance describes the composable transition as "designing for continuous evolution," which is precisely the characteristic that outcome-based pricing (which evolves as outcome definitions and measurement methods improve over contract cycles) and McKinsey's breakthrough archetypes (which compound as network effects and data advantages accumulate over time) both require to reach their full potential value.
How should a CFO explain the business case for a composable architecture investment to a board that views it primarily as an IT cost?
The most effective framing redefines composable architecture as a real option on strategic speed rather than an IT modernisation programme. A real option has a measurable value: the right to enter a new market, adopt a new pricing model, or respond to a new regulation at a lower total cost and in less elapsed time than a competitor using a monolithic architecture. The composable infrastructure market is growing at 51.3% CAGR, meaning that the premium for composable capabilities is currently being paid primarily by financial services, healthcare, and retail sector leaders who have already calculated that the cost of structural inflexibility exceeds the cost of the transition. The CFO's most persuasive board case includes a late-mover scenario that quantifies the revenue and margin impact of being 12 to 18 months slower than a composable-native competitor to deploy a new pricing model, enter a new product category, or satisfy a new regulatory requirement, and then compares that quantified cost against the transition investment as a floor on the minimum justifiable return from the programme.
