Most organisations are now doing something with AI. They have pilots, tools, working groups, vendor demos, productivity experiments, board papers and, in some cases, a small army of enthusiastic people quietly using personal accounts to get work done faster than the official process allows. From a distance, this can look like progress.
The problem is that AI activity is being mistaken for AI progress.
That distinction matters because the gap between AI adoption and AI value is now one of the clearest signals in enterprise transformation. McKinsey’s 2025 global AI survey found that 88% of organisations report regular AI use in at least one business function, yet only 39% report any enterprise-level EBIT impact from AI. The small group of high performers is not simply using better models; they are rewiring workflows, governance and leadership attention around value. (McKinsey & Company)
BCG’s AI Radar 2025 makes a similar point from a different angle. It found that only around a quarter of executives report significant value from AI initiatives, despite rising investment, and BCG continues to frame successful AI at scale through its 10-20-70 model: 10% algorithms, 20% technology and data, and 70% people and processes. (BCG Global)
That is the bit many leadership teams still underweight. AI does not create enterprise value because people have access to tools. It creates value when work is redesigned, decision rights are clarified, data is made trustworthy, risks are governed, people are equipped, and leaders change the way the organisation operates. In other words, AI adoption is an operating model change, not a technology rollout.
The last two years have created a misleading sense of momentum. AI usage has spread quickly because the tools are accessible, impressive and, in many cases, genuinely useful at individual task level. People can draft, summarise, analyse, classify, search, code, compare, ideate and automate parts of their work with far less friction than traditional enterprise software required. It is no surprise that adoption has accelerated.
But individual usefulness is not the same as enterprise value.
The enterprise question is more demanding. Has AI reduced cost in a way finance can see? Has it improved revenue conversion, customer retention, risk management, quality, speed or margin? Has it changed the work, or simply helped people cope with inefficient work faster? Has it been embedded into core processes, or does it sit beside them as a helpful assistant with no formal role in the operating model?
The MIT NANDA “GenAI Divide” report captured this tension sharply, arguing that despite substantial enterprise investment, a very large majority of organisations are seeing little or no return from generative AI initiatives, with the core issue sitting less in model quality and more in enterprise integration and learning. (MLQ)
That finding should not be used as a lazy argument against AI. The technology is useful and getting more capable. The lesson is more practical: AI value does not appear automatically because the tool is clever. It appears when the organisation becomes capable of using that tool inside real work, with real governance, against real outcomes.
The model may not be the problem. The organisation around the model often is.
One of the most important things to understand about AI is that it reveals organisational quality. It exposes unclear processes, weak data ownership, inconsistent decision-making, poor governance, fragmented systems, overloaded teams and leadership misalignment. It does not necessarily create these problems; it makes them harder to ignore.
If customer data is inconsistent, AI will amplify that inconsistency. If process ownership is unclear, AI will struggle to sit inside the workflow. If risk appetite has not been defined, teams will either move too fast or freeze. If legal, compliance, technology and operations are not aligned, the organisation will create a patchwork of local experiments that cannot scale. If leaders have not agreed what AI is for, the business will measure activity, not value.
This is why many organisations feel busy but remain stuck. They have an AI committee, but no operating model for AI. They have use cases, but no prioritisation logic. They have tools, but no clear ownership of adoption. They have enthusiasm in pockets, but no enterprise-wide mechanism for turning experiments into sustained performance improvement.
That is not a tooling gap. That is a transformation gap.
At Arqvera, this is the central premise behind AI.ccelerate: AI adoption has to start with readiness, operating model impact, governance, value and human adoption. The tool choice matters, but it is rarely the first question. The better first question is whether the organisation is ready to convert AI capability into changed behaviour and measurable value.
Pilots have a role. They help teams learn, test feasibility, build confidence and identify use cases worth scaling. The problem is that many organisations now have too many pilots and too little value discipline. They are running experiments because experimentation feels safer than making operating model choices.
A pilot should answer a specific question. Can this use case improve a measurable outcome? What workflow would need to change? What data is required? What risk controls are necessary? Which people need to adopt it? What cost-to-value profile makes it worth scaling? What would we stop doing if this works?
Too often, pilots answer a weaker question: can the technology do something interesting?
The answer is usually yes. That is not enough.
The real test is whether the organisation can industrialise the use case. That means taking it out of the innovation corner and placing it into governed, repeatable, supported work. It means defining ownership, process impact, exception handling, model oversight, training, measurement and benefit tracking. It means deciding whether the use case belongs in the flow of work or remains a productivity aid for individuals.
This is where AI programmes either cross the value gap or become part of the corporate theatre of progress. The organisation can produce case studies, dashboards and internal showcases while the underlying economics remain largely unchanged.
Most leadership teams know that employees are using AI tools informally. The instinctive response is often to see this mainly as a security, privacy or compliance risk. It can certainly be that, and those risks need to be governed properly. Sensitive data in unmanaged tools, unclear intellectual property treatment, unapproved processing and weak auditability are not minor concerns.
But shadow AI is also a management signal.
It tells leaders that people see value the formal organisation has not yet made safe, available or usable. It may show that approved tools are too slow, policies are unclear, enterprise platforms are poorly integrated, or teams are under pressure to deliver more than current processes allow. When people work around the official system, the response should not only be control. It should be curiosity.
What work are people trying to improve? Where are current processes too slow or too manual? Which tasks are repetitive enough to attract AI use? Where does the organisation need approved capability, clearer guidance or better workflow redesign?
The answer is not to allow uncontrolled usage and hope for the best. Nor is it to ban everything and pretend workarounds will disappear. The better answer is governed enablement: clear policies, approved tools, training, use-case pathways, risk tiers, data rules, human oversight and a practical route for frontline ideas to become enterprise capabilities.
Trust is a system. AI trust requires more than reassurance that the tool is safe. It requires clarity about how AI is used, who is accountable, how outputs are checked, what data is permitted, which decisions remain human, and how value is measured.
The most valuable AI opportunities are rarely achieved by dropping tools onto existing workflows and asking people to be more productive. That may create local efficiency, but it seldom creates enterprise-level value. The bigger opportunity comes from rethinking how work should happen now that certain tasks, decisions and handovers can be supported differently.
For example, in customer operations, AI may reduce the time required to summarise case history, classify issues or recommend next actions. But the value comes only if the service model, escalation rules, quality controls, team roles and customer measures are adjusted around that capability. In finance, AI may support variance analysis or forecasting, but value depends on data quality, management cadence and decision ownership. In sales, AI may help with account research, proposal drafting or pipeline analysis, but the benefit depends on adoption, coaching, CRM discipline and commercial leadership.
The same is true in professional services, software delivery, supply chain, HR and legal. AI can reduce friction, but the organisation has to decide what happens to the freed capacity. Does it reduce cost? Improve quality? Increase throughput? Shift people into higher-value work? Improve customer experience? Shorten cycle time? Without that management choice, productivity gains remain vague and often disappear into more activity.
This is why organisations do not realise value from AI tools. They realise value from changed work.
AI governance is sometimes treated as the thing that slows innovation down. That is usually because governance has been designed too late, too heavily or too far away from the work. The answer is not to avoid governance. The answer is to make it useful.
Good AI governance defines risk appetite, decision rights, use-case approval routes, data standards, model oversight, security requirements, human accountability, legal and compliance guardrails, vendor responsibilities and value measurement. It should make safe progress easier. It should help teams know what they can do, what they cannot do, when they need review and how to move from idea to scaled use without reinventing the process every time.
Poor governance does the opposite. It creates confusion, delay and workaround behaviour. Teams either avoid AI because they cannot navigate the rules, or they proceed locally because the formal route feels unusable. Both outcomes are avoidable.
This is where Trust Arq is relevant. AI initiatives need assurance around readiness, governance, delivery confidence, partner selection, risk and value. Independent challenge is especially useful before major commitments are made, because the AI market is full of confident claims and uneven delivery capability. Confidence before commitment matters.
A great deal of AI commentary talks about humans in abstract terms: workforce impact, reskilling, augmentation, displacement, adoption. These are important, but they can become bloodless. In practice, AI changes how people experience work. It can help them, unsettle them, expose capability gaps, change status, alter judgment boundaries and raise legitimate questions about trust, accountability and future roles.
If leaders ignore that, adoption will become performative. People may attend training, use approved tools occasionally and say the right things in surveys, while continuing to rely on old routines when the work really matters. That is not because people are resistant by nature. It is because they need confidence that the change is useful, safe, supported and aligned with how performance will actually be judged.
Human-centred change is not about making AI adoption comfortable at all costs. It is about making it real. People need to understand the purpose, the new expectations, the boundaries of human judgment, the skills required, the support available and the consequences of not changing. Managers need to know how to coach teams in the new work. Leaders need to model the behaviours they expect, not simply approve another tool.
That is the role of Change Studio in AI adoption: communication, engagement, behaviour change and adoption designed into the transformation rather than bolted on after the technology decision has already been made.
One of the quiet reasons AI programmes underperform is that value ownership is unclear. The technology function may own the platform. A transformation team may own the programme. A data team may own the models. An innovation team may own the pilots. But who owns the business outcome?
If the answer is not clear, the initiative will drift.
An AI use case intended to reduce customer handling time needs an operational owner who can change service processes, measures and team routines. A use case intended to improve sales effectiveness needs commercial leadership ownership. A use case intended to improve finance productivity needs finance leadership ownership. A use case intended to reduce risk needs accountable ownership from the risk, compliance or operational function it affects.
Value cannot be delegated to the tool. It cannot be delegated entirely to IT. It cannot sit in a benefits register that nobody uses after the pilot ends.
Arqvera’s Value Compass exists to keep this connection intact: outcome, baseline, owner, measure, adoption and benefit. This matters because AI value is often cumulative and operational. It appears through repeated use, better decisions and redesigned work, not through a one-off deployment milestone.
Leaders do not need to wait for perfect certainty before acting on AI. They do, however, need to stop treating AI as a technology rollout.
First, define the business outcomes before selecting or scaling tools. Cost reduction, revenue growth, customer improvement, risk reduction, quality, cycle time and decision speed are different objectives and require different operating choices.
Second, assess readiness honestly. That includes data quality, process maturity, leadership alignment, governance, workforce capability, risk appetite, vendor dependency and benefits ownership. Arqvera’s AI Readiness Self-Assessment is a useful starting point for leadership teams that want to understand whether confidence is evidence-based or merely enthusiastic.
Third, prioritise fewer use cases with clearer value logic. The aim is not to have the most pilots. The aim is to scale the few use cases that matter.
Fourth, design workflow change around the AI capability. Ask what decisions, roles, handovers, controls and measures need to change for the benefit to appear.
Fifth, govern AI as part of the operating model. Policies and principles are useful, but the real test is whether teams can move safely from idea to adoption with clear accountability.
Finally, track value after deployment. If the benefit depends on changed behaviour, keep measuring until the behaviour has changed and the operational result is visible.
AI is not simply changing work. It is exposing how well organisations already work.
Ready organisations will create value from AI because they will treat it as a transformation of work, governance and capability. They will start with outcomes, design the operating model, build trust in data, equip people, manage risk and track value. Tools will matter, but tools will not be asked to compensate for weak organisational discipline.
Unready organisations will continue to accumulate AI activity. They will run pilots, buy licences, publish principles, attend webinars, set up committees and wonder why the board is still asking where the value is. They may blame the model, the vendor, the users or the pace of regulation. Sometimes those factors will matter. More often, the issue will be simpler and harder to admit: the organisation invested before it was ready to change the way it works.
AI should amplify human capability, not bypass judgment or remove accountability. The future belongs to organisations that combine human expertise with machine capability inside a well-governed operating model.
The technology may be new. The leadership challenge is not.
Is an AI and technology transformation consultancy and advisory.
We help organisations shape business cases, projects, deliver excellence, and realise change and outcomes that stick. We support organisations before, during, and after projects with an end-to-end service where our domain specialization comes to life.
Before (Inception): We work with you to clearly define the idea, vision, strategy, and business case for change, as well as help select the right partners, and establish governance
During (Execution): We help deliver project and change objectives while keeping implementation under control through structured governance and assurance to realise intended outcomes.
After (Value Realisation): We ensure outcomes deliver measurable value and embed continuous improvement from successes and learnings.
Arqvera is led by industry veterans in the UK and USA with 100+ years of technology delivery intelligence across global consulting, digital transformation, and mission-critical projects and programmes.