BCG: The Formula for Agentic AI Value
Key takeaways
AI value is becoming measurable, but it remains concentrated. Almost 50% of companies are now generating measurable value from AI, while the largest share of that value remains concentrated among a relatively small group of top performers.
Strategic clarity and applied AI have to work together. Companies with both high strategic clarity and high applied AI maturity realise around five times as much AI value as companies lacking both.
Agentic AI is becoming a much larger part of the value pool. Its share of AI value increased from 17% in 2025 to 22% in 2026, and BCG projects it will reach 39% by 2030.
The operating model matters more than simply deploying better tools. Future built companies expect nearly 70% of their AI value to come from reshaping and inventing business processes.
Governance enables agentic value rather than slowing it down. 42% of companies expect agents to act autonomously by 2030, but only 5% currently have all the relevant agent controls in place. Companies with all six controls deployed enterprise wide generate roughly three times as much agentic AI value as companies with only one.
Workforce reinvention is a core part of agentic readiness. Strategic workforce planning is practised by 55% of future built companies versus 17% of laggards.
BCG expects AI to reshape work more than simply eliminate it. 89% of respondents expect AI to generate new work, while only 11% expect it primarily to replace existing jobs.
Human and agent roles have to be designed deliberately. BCG argues that companies need to work backwards from strategy to determine what to delegate to agents, where humans should remain in the loop or in the lead, and what roles and skills that implies.
The technology advantage is moving beyond basic cloud and data infrastructure. The new differentiator is agent ready data, shared semantics, operating context, enforceable access controls and an enterprise platform capable of orchestrating agents reliably.
Scale matters more than pilots. Future built companies run five times as many workflows at scale as laggards, despite having a portfolio that is only 1.5 times larger.
Transformation is predominantly an organisational challenge. BCG’s 10 20 70 model allocates 10% of AI transformation effort to algorithms, 20% to technology and data, and 70% to people, organisation and processes.
Executive summary
BCG’s Applied AI Index 2026 examines what separates companies that are beginning to generate measurable value from AI from those still struggling to translate investment into business impact.
The study surveyed 1,330 CxOs and senior executives who are AI decision makers, covering more than 20 sectors and respondents across more than 60 countries. BCG assesses companies across 41 capabilities and classifies them into four maturity groups: stagnating, emerging, scaling and future built.
The headline finding is that AI value is becoming real. Almost half of companies now fall into the scaling or future built categories and are generating measurable value. But value remains highly concentrated among a relatively small group of top performers.
BCG’s formula is straightforward: strategic clarity plus applied AI creates transformative impact.
Strategic clarity means choosing where AI can create competitive advantage, concentrating investment around those opportunities, establishing clear accountability and tracking value into the P&L.
Applied AI means embedding AI into how the business actually operates. BCG breaks this into three connected capabilities: an agentic enterprise, an agentic workforce, and agent ready architecture.
The agentic enterprise requires an AI first operating model with appropriate controls. The agentic workforce requires redesigning work around the respective strengths of people and agents. Agent ready architecture requires trusted data, shared operating context, enterprise intelligence and common infrastructure that enables agents to operate safely across workflows.
When these capabilities work together, the difference in value is substantial. Companies with strong strategic clarity and mature applied AI realise around 4.5% of revenue in AI value, compared with around 0.9% where both are weak, a fivefold difference.
The report also marks an important shift in how BCG views AI maturity. The question is no longer simply whether an enterprise has cloud infrastructure, good data or access to models. Those capabilities are increasingly table stakes.
The differentiator is whether AI is embedded into the operating model itself.
That means deciding which processes should be redesigned, what agents are permitted to do, when human judgement remains necessary, who owns each agent, what context the agent can access, how performance is evaluated, where escalation happens and how the enterprise measures value.
For WORK SELF, the workforce section is particularly relevant. BCG explicitly describes agentic readiness as an organisational shift, not a technology upgrade. Companies need to decide where agents should take on work and where human judgement, relationships and accountability should remain differentiating sources of value.
That moves the discussion from AI adoption to human and agent operating model design.
Key stats and quotable claims
BCG surveyed 1,330 CxOs and senior executives who are AI decision makers.
Respondents came from more than 60 countries.
Almost 50% of companies are now generating measurable value from AI.
Companies with strong strategic clarity and applied AI realise around five times as much AI value as companies weak on both dimensions.
7.5% of companies qualify as future built, up from 5% in 2025.
41% are scaling.
47% are emerging.
Just 4.5% remain in the stagnating category, down from 14% in 2025.
Future built companies generate 2.4 times as much top line growth as laggards.
They record 2.3 times the three year TSR.
Their EBITDA growth rate is approximately three times as high.
They report 4.3 times the cost reduction and 6.6 times the revenue increase associated with AI compared with laggards.
Agentic AI represented 17% of AI value in 2025, 22% in 2026, and is projected to reach 39% by 2030.
Future built companies are 3.5 times as likely as laggards to establish a dedicated AI transformation programme.
95% of future built companies use clear KPIs or directly track P&L value from AI.
Future built companies expect nearly 70% of AI value to come from reshaping and inventing business processes rather than simply deploying better AI tools.
42% of all companies expect agents to act autonomously by 2030.
Only 5% currently have all six relevant agent controls deployed enterprise wide.
Companies with all six controls generate roughly three times as much agentic AI value as companies with only one.
55% of future built companies conduct strategic workforce planning compared with 17% of laggards.
89% of respondents expect AI to generate new work.
Only 11% expect AI primarily to replace existing jobs.
Companies expect an overall workforce reduction of roughly 10% to 15% from AI, with a greater effect expected in middle management and frontline leadership.
66% of future built companies expect agents to operate with bounded or full autonomy by 2030, versus 32% of laggards.
95% of future built companies are conducting a data transformation, compared with 74% of laggards.
70% of future built companies run that transformation enterprise wide, versus 43% of laggards.
More than two thirds of future built companies are working towards a common enterprise wide AI platform.
Future built companies have an average of 20 AI workflows implemented at scale, compared with four for laggards.
62% of future built companies’ initiatives involve reshaping or inventing workflows rather than simply deploying AI into existing work.
BCG’s transformation model allocates 70% of effort to people, organisation and processes, 20% to technology and data and 10% to algorithms.
Overall summary
BCG’s central argument is that the enterprise AI problem is changing.
The first phase of AI transformation centred on access to models, cloud infrastructure, data foundations and pilots. Those capabilities still matter, but they are becoming less differentiating as adoption spreads.
The new question is whether organisations can embed AI into the way the business operates.
BCG finds that companies generate significantly more value when two conditions exist together. They have strategic clarity about where AI should create competitive advantage, and they possess the organisational, workforce and technology capabilities required to apply AI at scale.
Strategic clarity prevents AI from becoming a collection of unrelated pilots. Future built companies concentrate investment on a smaller number of high value opportunities, establish clear executive ownership, run AI as a multiyear transformation programme and track outcomes against financial value.
Applied AI turns that strategy into the operating model.
BCG identifies three elements.
The first is the agentic enterprise. Organisations need controls that allow increasingly autonomous agents to operate safely. The report identifies memory and context, scope and oversight, evaluation and rollback, tools and integration, ownership, and security as six essential control areas.
The second is the agentic workforce. Instead of simply adding agents to existing jobs, future built companies redesign workflows around agents and people. The starting question becomes which work should move to agents, where people remain in the loop and where humans remain in the lead.
The third is agent ready architecture. Agents require more than access to databases. They need current and trustworthy data, shared semantics, business rules, operating context and enforceable permissions. BCG describes this broader enterprise intelligence layer as increasingly central to competitive advantage.
The economic model also changes.
BCG argues that organisations should increasingly think about cost per successful outcome, rather than simply cost per token. The cost of an agentic workflow includes model usage, context, reasoning, retries, escalation and the human effort required to review or redo the work.
That connection between technology economics and human effort is particularly relevant. A cheaper model is not necessarily cheaper if employees repeatedly have to correct its outputs.
The report therefore reinforces a theme running through the other Judo research: the next stage of AI value is not primarily about giving employees more tools.
It is about redesigning workflows, roles, decision rights, accountability, context and human intervention around increasingly capable agents.
Deep dive
AI value is becoming real
For leadership teams, AI value can seem frustratingly elusive. But that is starting to change. Almost 50% of companies are now generating measurable value from AI, even as the biggest share of that value remains concentrated among a relatively small group of top performers.
Now, agentic AI has arrived with the promise of even greater rewards. It dramatically increases the potential value from AI, and a clear playbook can enable companies to capture that value, regardless of their current level of AI maturity.
The playbook embraces a simple formula: strategic clarity plus applied AI can translate AI ambition into transformative impact. Strategic clarity is about getting the big choices right to strengthen competitive advantage, applied AI means embedding AI into how the business operates day to day, and transformative impact is the enterprise value created when those elements come together.
This formula is the foundation for the rest of BCG’s report. The difference between AI leaders and laggards is not simply access to more sophisticated technology. It is the combination of clearer strategic choices and an operating model capable of applying AI at scale.
Keys to unlocking AI value
When strategic clarity meets applied AI, companies unlock five times as much AI value as when both are lacking. BCG’s Applied AI Index, based on a survey of more than 1,300 CxOs and senior executives across 20 plus sectors, offers empirical support for that approach.
Strengthen competitive advantage with strategic clarity
To succeed in agentic AI transformation, companies need to get the big choices right. The most AI mature companies in BCG’s sample, those that it calls future built, show exceptional strategic clarity in doing so, maintaining an average index score of 60 across functions compared with an average of 36 for laggards.
They concentrate AI into a focused, multiyear programme that aligns with the company’s biggest opportunities and business priorities, with clear accountability and a mechanism for tracking value to the P&L. In the survey sample, future built companies are 3.5 times as likely as laggards to set up a dedicated AI transformation programme.
Future built companies know what AI initiatives in their programme are worth because 95% of them use clear KPIs or directly track the P&L value from AI. Most other companies have far less visibility, with 47% of laggard companies measuring AI value directionally at best.
BCG’s argument is therefore not that enterprises need more experimentation. The higher maturity organisations are increasingly selective about where they place their ambition and more disciplined about measuring what the investment produces.
Build the capabilities for applied AI at scale
To get applied AI right, companies need to build capabilities across three pillars.
Shift to an AI first operating model with agentic controls
Future built companies expect nearly 70% of AI value to come from reshaping and inventing business processes, not just from deploying better AI tools. As these processes become increasingly agentic, companies will need to have the right controls in place to scale them safely.
Only 5% have all the relevant controls in place today, although 42% expect agents to act autonomously by 2030.
This creates a clear operating model gap. The level of autonomy companies expect to give agents is rising much faster than the governance infrastructure required to control that autonomy.
Reinvent the workforce around humans and agents
AI transformation is fundamentally an organizational shift that requires deliberate strategic workforce planning to determine which roles should exist in an agentic world.
Such planning is practiced by 55% of future built companies versus 17% of laggards.
It is especially important because AI is likely to reshape work rather than simply eliminate it: 89% of survey respondents expect AI to generate new work, versus just 11% who expect it to primarily replace existing jobs.
That distinction matters. If AI primarily reshapes work, workforce transformation cannot be reduced to a headcount exercise. Leaders need to understand what work moves to agents, what remains human and what new work emerges around the new operating model.
Build the agentic platform, powered by data
The technology advantage in AI is shifting from basic cloud and data foundations to agent ready architecture, the data and platforms needed to embed agents in the enterprise’s operations.
Future built companies are already pulling ahead: 95% are undergoing a data transformation, and more than two thirds are committing to a single enterprise wide AI platform with a common architecture and control plane instead of a single vendor stack.
The report is therefore moving the technology conversation above infrastructure alone. The question becomes whether enterprise architecture can provide the data, context, permissions, orchestration and controls an agent needs to participate safely in work.
AI leaders emerge in every sector
Technology, banking and telecom lead BCG’s maturity ranking. But the report finds that a company’s sector does not determine its state of AI maturity. All 20 plus sectors in the sample included at least one future built organisation.
The value gap across sectors is markedly consistent. Future built and scaling companies generated two to three times as much value from AI as laggards, taking into account both cost reduction and revenue growth.
BCG also finds that AI value is concentrated in a relatively small number of core functions. Across the sectors analysed, 54% to 78% of AI value rests in core business functions, with technology and IT and R&D ranking as the first or second driver virtually everywhere.
The exceptions matter because they demonstrate why BCG argues against generic AI rollouts. In banking, customer service is one of the major sources of value. In power and utilities, the value shifts towards energy generation, distribution and maintenance. In technology businesses, R&D and customer facing work become more important.
The conclusion is straightforward: concentrate investment in the two or three areas where the economics are strongest for the specific business.
Applied AI maturity is rising
Since BCG conducted its 2025 analysis, global applied AI maturity has increased across all industries. The group of stagnating companies has shrunk significantly to just 4.5% of the total, less than one third of last year’s 14%.
Companies are moving up to the emerging category, which now accounts for 47% of organisations in the analysis, and to scaling, which accounts for 41%. As a result, more organisations are starting to see value from AI.
Scaling organisations have adopted two to three times as many critical practices as laggards, but they still struggle to realise the full potential value from their AI investments. BCG describes their problem clearly: too many AI workflows in production and too few that are visible in the P&L.
Overall, 7.5% of the companies in the analysis are future built, a 50% increase from last year’s 5%. These companies apply the combination of strategic clarity and applied AI and demonstrate materially stronger performance.
Future built companies generate 2.4 times as much top line growth as laggards and 2.3 times as high a three year TSR. Their EBITDA growth rate is three times as high, supported in part by cost reductions where they outperform laggards by a factor of 4.3 and increased revenue by a factor of 6.6.
BCG does note in its methodology that overall company performance has multiple drivers beyond AI. The financial comparisons therefore show an association with higher AI maturity rather than proving that AI alone caused the differences.
Agentic AI is becoming a larger source of value
Agentic AI is serving as a catalyst. Its share of AI value climbed from 17% in 2025 to 22% in 2026 and BCG projects it will reach 39% by 2030.
On that trajectory, agentic AI would overtake nonagentic AI solutions to become the largest single source of AI value.
But BCG is clear that agentic AI does not eliminate the need for the underlying foundations. Laggards still need to build capabilities across technology, data and the operating model.
What changes is the payoff for companies that develop those foundations successfully.
For companies in the stagnating and emerging tiers, BCG recommends building the foundations and operating model for AI across the enterprise: prioritising project delivery, redesigning the operating model, developing a portfolio of high value areas and upskilling the workforce.
Scaling companies seeking to become future built need to industrialise their AI initiatives and agentic readiness by embedding AI into products, tracking value, augmenting humans with agentic AI and giving agents greater autonomy under suitable controls.
What sets future built companies apart
Future built companies focus their ambition on bigger bets in the areas that matter most.
They report that 41% of their revenue and costs are addressable through AI. In contrast, laggards report that only 23% of their revenue and 26% of their costs are addressable.
Across the overall sample, about 68% of AI value is generated in business units, 21% in enabling functions and 11% in technology and IT.
Leading companies also move faster. Factoring in full process redesigns, future built companies reach impact in around 11 months, compared with 14 months for scaling companies and 17 months for laggards.
They also outspend on AI, not simply on IT. Enterprise IT budgets are roughly flat relative to the previous year, even as AI maturity has increased. What has changed is dedicated AI spending, which BCG says has moved from approximately 1% of revenue in the prior analysis to 3.3% in the latest study.
A growing share of the investment sits outside enterprise IT. Across surveyed companies, AI spending outside IT is around five to six times as high as AI spending within IT, at 2.8% of revenue versus 0.5%.
This includes AI talent, governance, transformation programmes, operational AI and AI embedded in customer facing products.
The message is that where the organisation spends matters as much as how much it spends.
Make AI economics visible
As the share of AI cost generated outside IT increases, BCG argues that companies need transparency over total AI spending.
Companies benefit from virtual P&Ls that compare AI benefits against underlying workforce and token costs.
This reflects a broader evolution in the role of technology leadership. As AI development and spending extend across the enterprise, technology leaders need to maintain visibility over AI economics and link costs directly to business value.
Future built companies also manage token spending deliberately rather than by default.
Half actively encourage adoption to maximise usage, while another 22% set usage limits or controls to manage spending. The specific approach can vary, but nearly three quarters of future built companies have made an explicit decision about token spending and manage it towards an explicit return.
The corresponding discipline is far weaker among laggards.
Five moves to accelerate AI value generation
BCG brings its findings together into a framework consisting of strategic clarity, applied AI and transformative impact.
Applied AI itself has three subcomponents: agentic enterprise, agentic workforce and agent ready architecture.
Strategic clarity: get the big choices right
Strategic clarity is about getting the big choices right to strengthen competitive advantage. It requires connecting initiatives across the enterprise and understanding how technology, functions and people fit together.
It also requires foresight about where advantage may shift, conviction to concentrate resources on a few high value plays that reshape processes end to end rather than spreading effort across many small deployments, and discipline in tracking the outcomes.
Future built companies bundle their AI efforts into a single programme owned by the C suite with a multiyear timeline.
They are 3.5 times as likely as laggards to commit to a multiyear AI effort, 61% versus 17%, instead of funding short, stop and start initiatives.
C suite leaders in future built organisations are also 20 percentage points more likely than laggard leaders to hold the AI budget.
Accountability is effectively universal among future built companies: 100% have clear ownership of the AI agenda.
Value tracking is paramount
Future built companies track AI value directly into the P&L, which gives them a clear sense of what is working and how well while providing a source of funding for subsequent AI investment.
Among this group, 95% measure value on the P&L or through a proxy.
The same rigour is missing at many other organisations. BCG cites separate research finding that while nearly nine out of 10 CEOs already see cost or revenue benefits from AI, only 14% have clearly defined the P&L impact of every initiative.
Future built organisations have an average of 20 AI workflows implemented at scale, compared with an average of four for laggards. And when they set targets, they are more than 50% more confident of achieving them.
BCG gives the example of a global investment bank that chose to run AI as a single funded company wide programme. Leadership identified fewer than 10 high value use cases, developed targets for end to end processes and established a programme office reporting to senior leadership to track execution and value directly to the P&L.
The organisation aims to generate financial upside of $500 million by 2027 and $1 billion by 2030.
Shift to an AI first operating model with agentic controls
BCG argues that agentic is where a large share of tomorrow’s AI value sits.
Among future built companies, 44% say they already realise full value from agentic AI, compared with 2% of laggards.
Future built companies also expect nearly 70% of their AI value to come from reshaping and inventing business processes, not simply deploying better AI tools.
But bringing more agentic workflows into production creates the risk of agent sprawl: duplicated agents, conflicting policies or permissions, unclear ownership and monitoring gaps.
The deeper shift comes from agents combining data, tools, permissions, memory and other agents across workflows. Risks can therefore emerge dynamically and compound at machine speed.
BCG argues that applied AI consequently requires both the enterprise risk operating model and the underlying control architecture to evolve.
Six essential agent controls
Companies that combine central risk and control guardrails with federated implementation across business and technology functions bring three times as many workflows to scale as companies relying on fully centralised steering.
BCG identifies six essential controls.
Memory and Context. Configure how agents use, store and recall memory and context. Memory can compound learning across agents, but it can also propagate errors, sensitive information or manipulated context across workflows.
Scope and Oversight. Define each agent’s purpose, scope, autonomy and required human oversight. Authority should be bounded not only by which systems an agent can access, but by the actions it is permitted to take, for which purposes and in what context.
Evaluation and Rollback. Establish evaluation criteria, release gates, continuous monitoring and rollback mechanisms. Controls should detect drift, anomalous behaviour and emerging risks after deployment rather than merely determine whether an agent was safe at launch.
Tools and Integration. Implement standardised tool integrations, APIs and agent registries such as the Model Context Protocol registry. Shared standards provide visibility into which agents are operating, what they can access and how they interact.
Ownership. Assign ownership across business, technology and risk functions. Every material agent should have a named owner accountable for its use and outcomes, including clear authority over who can change, suspend or stop it.
Security. Set runtime security controls, audit trails and cost guardrails. Controls should constrain agent actions as they occur and preserve enough evidence to reconstruct what an agent did, under whose authority, using which data and tools, and which controls were triggered.
The report’s finding here is particularly important: controlling agents does not diminish their value. It increases it.
Companies with all six controls deployed enterprise wide generate around three times as much agentic AI value as organisations operating with only one.
Autonomy has to be designed at the process level
Across the full sample, 42% of companies expect agents to act autonomously by 2030, including making decisions without human approval, yet only 5% currently have the full set of relevant controls in place.
BCG gives the example of a large global business process outsourcing provider that redesigned its model around increasing levels of AI autonomy.
The organisation created a ladder moving from insights, where AI diagnoses the work, to copilot, where AI assists employees, to autopilot, where AI agents execute tasks end to end.
Critically, autonomy is set at the process level, not at the level of individual tools.
The process defines which steps exist, where decisions occur and when a human must be involved.
This is a significant operating model principle. The question is not simply whether a particular AI tool is autonomous. The organisation has to decide which parts of the workflow can operate autonomously and where human involvement is structurally required.
Reinvent the workforce around humans and agents
The next element of agentic readiness involves reinventing the workforce.
Laggards take a baseline view of agentic AI, focusing on tools that help existing employees execute current processes somewhat more efficiently.
Future built companies redesign the organisation around agentic AI, reshaping processes and workflows to capitalise on agents.
BCG describes this as an organisational shift, not a technology upgrade.
Companies must make an explicit strategic choice about where agents should take on work and where human judgment, relationships, accountability or other forms of contribution should remain a source of differentiation.
Rather than simply agentifying today’s organisation, companies need to work backward from strategy to determine what to delegate to agents, where humans should remain in the loop or in the lead, and what roles and skills that implies for the human workforce.
That degree of strategic workforce planning capability is a major differentiator.
Strategic workforce planning happens at 55% of future built companies and just 17% of laggards, a 38 percentage point gap.
Future built organisations already have 13% of their employees working in dedicated AI roles, a higher proportion than laggards expect to reach by 2030.
BCG argues that these workforce changes make CHROs the architects of a hybrid workforce of humans and agents. They need to address HR fundamentals while simultaneously redesigning roles and embedding agents into the organisation’s ways of working.
AI creates new work as well as changing existing work
Most companies, 79%, expect technology costs to rise through 2030, while 49% expect labour costs to fall.
But BCG says the bigger story is that AI creates new work rather than simply eliminating existing work.
Fully 89% of companies expect AI to generate new work, versus 11% that expect it primarily to replace existing jobs.
Overall, companies expect a workforce reduction of roughly 10% to 15% from AI. The expected contraction is concentrated more heavily in middle layers of the organisation, particularly middle management and frontline leadership.
The implication is that leaders will not fully automate the workforce.
Instead, they will make choices about where AI can create a step change in productivity and where human interventions should be preserved, elevated or reinvented because they create strategic value.
Today, more than five times as many future built companies as laggards, 38% versus 7%, say that AI fully executes tasks independently. Another 33% of future built companies say AI augments expert decisions and improves output quality.
Future built companies are also much more likely to let an agent run an entire task from start to finish with no human step in between and are more likely to use AI for substantial multistep work rather than quick one off questions.
By 2030, 66% of future built companies expect agents to operate with bounded or full autonomy, compared with 32% of laggards.
Adoption alone does not change the workforce
BCG describes a global cloud and technology company with high AI adoption across a 1,000 person engineering team but only around 10% productivity improvement.
The problem was not the tools. It was employee behaviour.
Engineers used AI infrequently and mainly for basic tasks.
The company responded with tailored six week sprints that coached engineers to use AI more effectively at work. Metrics tracked depth of use, moving towards end to end feature development; breadth of use, extending from coding into the full software development lifecycle; and frequency of interaction.
The number of AI power users increased from 5% of engineers to 50%, and those employees became around 80% more productive than before the transition.
The example reinforces the wider point from the report: providing technology is not the same as redesigning how people work with it.
Customer service as a human and agent workflow
A large US regional health insurer identified customer service as a lighthouse project.
About 80% of incoming calls concerned benefits and coverage, while the average call lasted 16 minutes, nearly twice the industry average. Customer service employees needed to search more than 20 documents per call on average.
The organisation deployed three agentic AI solutions powered by a platform linking internal documents, structured data and live APIs.
The first large scale pilot went live within four months.
Agentic chatbots can now handle approximately 60% of incoming call types, reducing average handling time by 35% and freeing human employees to handle more complex requests.
The company projects that the first implementation wave will reduce contact centre costs by more than $60 million, with broader AI use cases projected to reach $200 million in run rate savings.
The workforce point is again not that the human disappears. Routine requests move towards agents while human attention is redirected towards more complex interactions.
Build an agentic platform powered by data
The technology gap is no longer mainly about basic data or cloud, which BCG says are increasingly widespread.
Instead, it has moved up the stack into data, enterprise intelligence and platform layers that allow agents to understand context, take governed action and operate reliably at scale.
Basic data structuring and access controls are now relatively common across maturity levels.
The sharper divide concerns whether organisations can rely on the semantics of their data: whether information is current, consistently defined, trustworthy and managed with clear ownership.
BCG calls the new advantage agent ready data.
Future built companies are 1.4 times as likely as laggards to give agents live access to company data so answers remain grounded in current information.
They are 1.7 times as likely to have explicit and enforceable controls over what each agent can see and do.
Among future built organisations, 95% run a data transformation compared with 74% of laggards, while 70% run it enterprise wide compared with 43%.
But BCG is clear that readiness requires more than adding a technology layer.
Companies need to define trusted data in a way that users can understand and implement consistently across the organisation. They also need a shared understanding of semantics and operating context and need to redesign information flows and controls for agents.
A common control plane, not a single vendor stack
More than two thirds of future built companies concentrate their AI effort around a company wide platform target, compared with around one fifth of laggards.
BCG explicitly says the goal is a common architecture and control plane, not a single vendor stack.
The organisation should build shared capabilities for security, access, orchestration, observability, evaluation and agent registration once, and then reuse them across the enterprise while allowing teams flexibility across different models and providers.
This becomes increasingly important as agents operate across more enterprise workflows.
Own and protect the enterprise cortex
BCG argues that traditional technology lock in can evolve into cognitive lock in.
As AI systems become part of how companies make decisions and execute work, organisations risk becoming dependent not only on a vendor’s technology but on the reasoning processes and operating context embedded within it.
Companies should therefore keep their enterprise cortex, proprietary knowledge and IP, shared ontology, business rules, decision logic and operating context, in a governed enterprise intelligence layer under their control.
That layer should remain technically separable and portable.
Around that core, organisations should retain flexibility across models and providers, supported by common routing, evaluation and fallback logic.
The goal is to use the best technology for different tasks without surrendering the company specific knowledge that makes the AI useful.
Make AI economics visible and manageable
Agentic systems introduce a new variable cost into the technology stack.
That cost depends on the model used, the amount of context supplied, the reasoning effort required and the number of steps an agent takes to complete a task.
BCG argues that traditional infrastructure cost management is therefore insufficient.
Increasingly, the relevant unit is not cost per token, but cost per successful outcome at the required level of quality, latency and risk, including the human effort needed to review and operate the workflow.
Architectures need workflow level visibility into consumption and value.
Organisations also need levers such as using deterministic software where a model is unnecessary, routing tasks towards the smallest model capable of meeting requirements, and setting context reuse, caching and stopping rules.
BCG makes an especially useful observation here: the cheapest model is not always the lowest cost option.
A more capable model that completes a task correctly on the first attempt can cost less overall than a cheaper model that retries, escalates or produces work a human needs to redo.
Economics therefore becomes a question of architecture and accountability rather than procurement alone.
Every material workflow needs a named owner, an intended outcome and explicit budget guardrails.
Continuously renew the technology landscape
Agentic AI can create complexity as quickly as it creates capability.
New agents, interfaces, dependencies, machine generated code and controls can accumulate on top of existing systems.
BCG argues that companies consequently need to treat modernisation as a continuous discipline rather than a periodic programme.
AI can accelerate legacy migration, ERP modernisation, documentation and code generation, but only within disciplined architecture, engineering and lifecycle controls.
The objective is to prevent today’s rapid experimentation from becoming tomorrow’s rigid and difficult to maintain agentic estate.
Transformative impact comes from scaling workflows
Future built companies win through depth and speed.
They have fewer workflows stuck in planning and run five times as many workflows at scale, 20 versus four for laggards, from a portfolio only 1.5 times as large.
They also place greater emphasis on process redesign.
BCG argues that commercial AI tools can be deployed into existing tasks and workflows, but significantly more value comes from reshaping and inventing business processes and workstreams end to end.
Among future built companies, 62% of AI initiatives involve reshaping and inventing rather than simply deploying AI solutions, compared with 57% for scaling organisations and 44% for laggards.
Future built companies do not invest everywhere.
They concentrate resources where the value is greatest, scaling a relatively small number of high value functions such as R&D, maintenance and customer service.
BCG finds that when companies take a core workflow to scale, examples including credit decisioning, insurance claims fraud detection and consumer demand forecasting can generate double digit gains in productivity, revenue and cost reduction.
A functional view of AI maturity
BCG reiterates that creating value from AI requires both strategic clarity and applied AI.
Companies strong in both see approximately five times the impact on AI value, and around three quarters of this group are classified as future built.
But the report also adds an important qualification.
The real value of work with AI does not arise within individual functions but emerges from end to end reshaping of workflows and processes that span multiple units and functions.
Functions remain a useful starting point for companies at lower maturity levels, but they are not the final unit of transformation.
Strategic clarity is the strongest differentiator across the 14 functions BCG examines. Future built companies average a score of 60 compared with 36 for laggards.
Scale also matters more than pilots.
Within IT, future built companies move a much larger share of workflows into deployment and scale than laggards and generate more value from each.
The capability gap is widest in R&D, followed by customer experience and operations, which again reinforces BCG’s finding that leaders concentrate on the parts of the enterprise where value is greatest.
Playbook: how to become an agentic future built company
There is no shortcut to becoming a future built company and no single measure is right for every organisation.
But BCG identifies one principle that applies broadly: becoming future built is primarily an organisational shift, not only a technical one.
In BCG’s 10 20 70 model, companies focus just 10% of their AI efforts on algorithms, 20% on technology and data and the remaining 70% on people, organisation and processes.
Future built companies follow this weighting closely.
Nearly three quarters of their most important priorities fall into the people, organisation and process category, with approximately 19% coming from technology and 9% from algorithms.
Standout capabilities include human AI augmentation, AI productisation, research and ideation, AI operations, AI delivery and the agentic AI ecosystem.
The implication is that model quality and infrastructure are necessary but account for a minority of the transformation challenge.
Most of the work lies in changing how the organisation operates.
A structured approach to implementing AI
BCG describes a multinational consumer company that wanted to reinvent core processes across functions using AI.
It followed a clear sequence: map the work and tacit knowledge in detail; redesign each process, balancing AI and human tasks; deploy fast with low code AI tools that embed enterprise knowledge; and scale the best ideas to achieve P&L impact.
The journey began in marketing, where the company introduced GenAI tools across the entire function.
The tools automated tasks accounting for 30% to 40% of employee time while the company reinvented workflows end to end.
Human work time spent on routine activities fell by as much as 70%, while output quality improved twofold.
The organisation then applied the same approach to R&D and procurement, supported by upskilling and value tracking.
The sequence is important.
The company did not begin with the tool.
It began with the work and the tacit knowledge required to perform it.
Rebuilding workflows for AI rather than simply automating them
BCG also describes a global insurer seeking to transform its core business with agentic AI across business units and group functions.
Rather than adjusting existing workflows at the margins, the company rebuilt those workflows from the ground up as agentic first.
It developed a quantified business case and implementation roadmap for each workflow, created a minimum viable product and then scaled deployment through a central office.
The insurer expects several hundred million dollars in total value from the group wide transformation.
Again, the difference is between putting AI into the current workflow and redesigning the workflow around what AI now makes possible.
The priorities depend on the company’s starting point
BCG distinguishes between organisations trying to move from laggard or emerging status into scaling, and those already scaling that want to become future built.
Companies building the foundations should bundle AI efforts into a single multiyear programme, establish clear C suite ownership and value measurement, keep humans in the loop as final decision makers and owners where appropriate, establish guardrails and escalation routines, strengthen talent and workforce planning, and improve enterprise data and platform foundations.
Companies moving from scaling towards future built need to industrialise.
They should extend direct P&L value tracking, concentrate on end to end workflows, give agents real autonomy within bounded domains, enhance decision rights across workflows, redesign work around humans and agents, build agentic architecture at scale and tie technology spending directly to business outcomes.
The destination therefore changes as maturity increases.
Early maturity is primarily about building the foundations required to use AI safely.
Higher maturity is about increasing autonomy, redesigning work and scaling the workflows that create the most value.
Conclusion
Companies at lower AI maturity levels often think that the goal is to simply deploy more tools, fund more pilots or chase the newest technical breakthrough.
BCG argues that this approach will not work.
Becoming an agentic, future built organisation requires a disciplined agenda anchored in strategic clarity and applied AI. The organisation must embed programmes into the way work gets done, govern them with confidence and measure them by the value they create.
Agentic AI raises the stakes because it transforms AI from an assistant at the edge of the enterprise into a participant in its core operating model.
BCG’s final message is that sector, legacy and starting point do not determine whether a company can become future built. The important variables are leadership choices: where the organisation concentrates its efforts, how it empowers people, whether it builds agent ready foundations and whether it can scale the workflows that matter.
What the report means for workforce transformation
The workforce section is one of the strongest parts of the report because BCG moves beyond the question of how many jobs AI removes.
The report explicitly says that 89% of companies expect AI to generate new work, while only 11% expect it primarily to replace existing jobs.
This means the larger workforce problem is likely to be redistribution of work.
Some activities move entirely to agents. Some remain with employees. Some become collaborative. New activities appear around supervising, governing, improving and operating agentic workflows.
The role therefore needs to be redesigned at the level of tasks, decisions and accountability.
BCG’s wording is particularly important: companies need to work backwards from strategy to determine what to delegate to agents, where humans should remain in the loop or in the lead, and what roles and skills that implies for the human workforce.
That requires much more than training employees to use AI.
Leaders need to understand which roles change, which activities disappear, which activities become more important, what new work emerges, where human judgement creates strategic value and which employees can realistically transition into the new model.
The governance section reinforces the same requirement.
Scope and oversight must define the agent’s purpose, autonomy and required human oversight. Ownership must identify who remains accountable. Evaluation and rollback determine when the system should stop. Memory and context determine what the agent knows. Security defines what the agent can do.
Each of those technical controls has a corresponding workforce decision.
The human and agent operating model cannot therefore be designed separately from workforce transformation.
What this means for WORK SELF
This report is particularly close to the problem WORK SELF is addressing because BCG effectively connects agentic AI value, workforce reinvention, human oversight and enterprise context in one operating model.
There are several direct connections.
First, BCG says organisations need to determine what should be delegated to agents and where humans should remain in the loop or in the lead.
That requires employee and role level visibility rather than aggregate workforce planning.
Second, BCG places memory and context first among its six agent controls. Agents increasingly need more than raw enterprise data. They need the operating context that determines how work should happen.
Third, BCG explicitly says agents need defined scope, autonomy and required human oversight. That aligns closely with the need to define who owns the work, when human review is required and what should trigger escalation.
Fourth, BCG argues that every material agent needs a named owner accountable for its use and outcomes.
Fifth, the report says the new technology advantage sits in an enterprise intelligence layer containing proprietary knowledge, shared ontology, business rules, decision logic and operating context.
That is extremely close to the Human Context Infrastructure problem.
Maya Enterprise sits at the point where those requirements become operational. Existing enterprise systems, AI models and agents can remain in place. The missing layer is governed human context around the workforce, roles, ownership, review, autonomy and escalation.
The report also gives a strong economic justification for that layer.
BCG says companies should increasingly assess cost per successful outcome, including the human effort needed to review and operate the workflow.
That means rework is not a peripheral workforce problem. It becomes part of the economics of the agentic architecture itself.
An agent that is cheap to run but repeatedly requires a person to correct its output may be more expensive than a higher quality system with better context and fewer human correction cycles.
The commercial connection for WORK SELF is therefore not simply that human context improves employee experience.
Human context becomes part of the infrastructure required to make agentic AI economically viable at scale.
Key takeaways
AI value is increasingly real, but it remains concentrated among companies that combine strategic clarity with the ability to apply AI at scale.
Agentic AI moves AI from a productivity tool into the core operating model.
Future built companies expect most future AI value to come from reshaping and inventing workflows, rather than deploying more tools into existing work.
Governance is an enabler of agentic value. Companies with all six controls generate around three times as much agentic AI value.
Memory and context, scope and oversight, ownership, security and escalation become operating infrastructure rather than policy documents.
Workforce transformation is not simply about eliminating jobs. 89% of companies expect AI to create new work.
Companies need to explicitly decide what agents do, what humans do and where humans remain accountable.
Strategic workforce planning is already much more common among future built organisations, 55% versus 17% of laggards.
Agent ready data requires shared semantics and operating context, not merely access to databases.
The right economic unit increasingly becomes cost per successful outcome, including human review and rework.
Scale matters more than pilot volume. Future built companies run five times as many workflows at scale.
BCG’s 10 20 70 model makes the broader conclusion clear: the largest part of AI transformation sits in people, organisation and processes, not algorithms.
The bottom line
BCG’s report marks a useful shift in the enterprise AI conversation.
The question is no longer simply whether organisations can deploy AI.
Almost half are already producing measurable value.
The harder question is whether they can redesign the organisation around increasingly autonomous agents.
That requires strategic clarity about where value sits, workflow redesign rather than isolated tool deployment, agent ready data and context, governance that defines autonomy and accountability, and a workforce model that determines where agents act and where humans remain in the loop or in the lead.
The most important line for WORK SELF is therefore not that agents are becoming more capable.
It is that AI transformation is fundamentally an organisational shift.
As agents move from assistants at the edge of the enterprise into participants in the core operating model, companies need a governed way to connect those agents to the people, roles, context and accountability structures around them.
That is the layer where agentic AI becomes an operating model rather than another technology deployment.
