An 82-Point Gap Between Access and Value
88% of organizations now use AI in at least one function. Only 6% qualify as high performers with measurable EBIT impact. Two-thirds are still in pilot purgatory.
The 82-point gap between access and value is the share of AI capability that evaporates between a license being purchased and an outcome being delivered. That difference deserves attention alongside adoption figures.
The gap can appear in a familiar situation: tools have been deployed and dashboards built, but a pilot is still active after six quarters. Someone has to decide what would justify putting it into normal use. The 88% means access is no longer the bottleneck; the 6% means most organizations have not converted that access into value at the level that registers in financial statements.
The instinct is to call it normal adoption lag. Rogers’ S-curve. Electricity took 30 years. The internet took a decade. ChatGPT reached one million users in five days — compressing the awareness curve from decades to days — while operational integration stayed on the traditional slow S-curve. Same story, faster.
Except it is not the same story. Electricity had a ceiling that adoption eventually reached. AI’s capability compounds, and the compounding rate keeps rising. With every model release, the gap between what is technically possible and what most organizations are doing on Tuesday morning gets wider, not narrower. That is not a transitional lag. It is structural dissipation — something the S-curve was never built to describe.
I call the space between the two curves the dissipation gap. Capability rises on a near-vertical curve. Diffusion rises on a flat one. The distance between them is where compounding economic advantage lives — and where most of the AI investment of the past three years has quietly evaporated.
Four Reasons Access Does Not Become Value
Four forms of inertia slow organizational adoption. They compound when simultaneous, and most large organizations are carrying all four at once.
Regulatory inertia. The EU AI Act is fully applicable on August 2, 2026. Large enterprises face $8-15 million in initial compliance investment and $1-5 million annually. That is a real cost. But the meaningful inertia is not the dollar figure — it is using compliance review as the reason to defer every initiative, including the governance preparation itself.
Organizational inertia. 45% of executives report AI ROI below expectations. The barriers were not technological. They were people, processes, and politics. Pilots stalled because the metric had no owner, the workflow was never re-scoped around the AI, and the legacy process had no retirement path. The technical deployment worked, but the organization had not adapted the workflow around it.
Cultural inertia. 87% of leaders cite psychological inertia as the main barrier to transformation; 68% of non-transforming organizations identify “insight inertia” — the unwillingness to accept that things need to change at all. That last category is the awkward one: the leader who in private says the company needs to move faster, and in public signs off on another evaluation cycle. The incentives help explain this behavior. The cost of acting is loaded onto them personally; the cost of waiting is diffused across the organization.
Trust inertia. This one is moving in the wrong direction. Trust in company-provided generative AI fell 31% between May and July 2025; trust in agentic AI dropped 89% over the same period. The pattern is not “trust takes time to build.” It is “trust is actively eroding while deployment accelerates.”
Trust often starts low for new technology and rebuilds as systems improve. ATMs were distrusted, and online banking faced years of resistance. Could the current gap be a similar transition rather than a structural problem?
The ATM example also shows the difference between technical readiness and organizational change. ATMs were technically reliable within a few years of introduction. Banks took more than a decade to restructure branch operations around them. Individual trust in the tool and organizational trust in the workflow run on different timelines. Leaders who would not stake a business outcome on an AI recommendation today are not, in most cases, waiting for better models. They are waiting for the day staking outcomes on AI feels like fiduciary judgment instead of a personal bet — and that day arrives later than the technology readiness suggests.
The Governance Paradox
Moving slowly on AI, the argument goes, is fiduciary responsibility. The EU AI Act imposes real compliance costs. 62% of enterprises deploying agentic AI without oversight experienced operational incidents within twelve months. Klarna replaced 700 customer service agents with AI and reversed course publicly when satisfaction cratered. Moving fast looks like the dangerous play, not the conservative one.
Klarna illustrates the risk of replacing roles without redesigning the work. In 2024, Klarna deployed AI chatbots across most of its customer service, cut roughly 700 positions, and publicly championed the move. The chatbot handled two-thirds of interactions in its first month.
Then customer complaints rose. Satisfaction declined. Internal reviews confirmed the AI lacked empathy and contextual reasoning for nuanced support. Klarna reversed course and committed to always offering a human option.
Klarna closed the dissipation gap the wrong way: eliminate the organizational layer instead of restructuring it. They treated AI as substitution, not transformation — and produced a trust deficit that made future adoption harder. The “go slow” argument seems vindicated.
It is not — because of what shows up underneath the prudence narrative. 72% of enterprises deploying agentic AI do so without formal oversight; only 1 in 5 maintains mature governance for autonomous agents, and governance maturity lags deployment speed by 18 to 24 months. Not every organization in that 72% is avoidant; some have legitimate constraints they have not yet resolved. But the broader pattern is hard to miss: the cohort citing regulatory risk as the reason for caution has, in most cases, not built a governance framework at all. In these cases, caution becomes a reason to defer the governance work that deployment requires.
This is where I think the adoption-curve explanation falls short. The objection says the 82-point gap is normal adoption lag. But the gap is not between organizations that have AI and those that do not — those numbers are 88% and 12%. The gap sits between organizations that have AI and organizations that have converted AI into value. That is a structural conversion failure, and it compounds in a way Rogers’ framework was never built to predict.
Compare to JPMorgan Chase. The largest US bank by assets allocated $17 billion to technology in 2025-2026 and deployed its proprietary LLM Suite to over 230,000 employees. Banking is one of the most heavily regulated industries on the planet. Most competing banks treated those constraints as reasons to defer. JPMorgan faced identical constraints and chose differently: build the compliance framework first, then deploy at scale on top of it. They exceeded their $1 billion AI value realization target ahead of schedule.
The paradox is that governance-first adoption moves faster, not slower. Deferring until clarity arrives creates the conditions for either inaction or a Klarna-style scramble. Investing in governance now is what makes safe deployment possible later. Different organizations read the same regulatory text as either a wall or a procurement specification. The choice is downstream of culture, not compliance.
Why Fast Followers Lose This Time
The other natural objection is the fast-follower argument. Amazon was not the first online bookstore. Google was not the first search engine. First movers eat the cost of market education and immature infrastructure; fast followers learn from the mistakes and deploy on cheaper, mature technology. Applied to AI, the logic looks airtight: wait, then move.
That logic holds in markets where the technology stabilizes and competition is about product. AI adoption is structurally different in two ways.
First, this is not a technology deployment. It is an organizational learning process. The companies pulling ahead are not installing better software — they are restructuring workflows, building internal expertise, creating data feedback loops, and developing institutional muscle memory for working with AI. Those capabilities cannot be procured from a vendor. They accumulate through use, like a craft.
Second, the academic record on AI specifically shows the opposite pattern from consumer technology. Peer-reviewed research on first-mover and second-mover strategies in AI/ML adoption demonstrates that first-mover innovation benefits widen over time as the technology matures, rather than narrowing the way they did for online retail or social networks. A fast follower can copy a product. They cannot copy two years of organizational learning in how to integrate AI into decision-making.
Each successful deployment builds trust and knowledge that accelerates the next. Each deferred one makes the next politically harder, because the organization is now staring at the gap between it and its competitors with fewer internal advocates who know what good looks like. The 6% high-performer separation looks early precisely because the gap is still forming. This is the window where the advantage compounds.
Where You Sit on the Dissipation Curve
Use four questions to assess whether access to AI is translating into business value.
1. Map your capability access. Inventory every AI tool and license your organization has deployed. Most score high here — access is no longer the bottleneck. If you score low, your problem is procurement, the simplest of the four to solve. The failure mode worth watching is confusing access with adoption. ChatGPT Enterprise licenses and a Slack channel for “AI experiments” are evidence that someone signed a contract, not that anything has changed.
2. Audit your four inertias. Score each 1-5: regulatory, organizational, cultural, trust. Sum the scores. Below 8 means low dissipation; above 15 means structural intervention. The mistake is treating all four as equally solvable. Regulatory and trust inertia need institutional investment over months; cultural and organizational inertia can shift faster with leadership commitment. Identify the one or two creating the most friction and start there. Fixing all four at once is the fastest route to four open initiatives and no movement.
3. Measure your value conversion rate. Classify every AI initiative as (a) delivering measurable business value, (b) active pilot with a defined production decision date, or (c) experimental with no timeline. Calculate the percentage of total AI investment in category (a). Below 20% means your dissipation rate exceeds 80%. Classify a pilot without a decision date in category (c).
4. Identify your trajectory. Compare your conversion rate quarter over quarter. Is it rising, flat, or falling? Is your time from approval to production getting shorter? If conversion is rising and velocity is accelerating, you are compounding. If either is flat or declining, your gap is widening even if absolute AI usage is growing. Doubling AI projects while keeping an 80% dissipation rate doubles the capability you are wasting — math that rarely makes it onto the dashboard, because the answer is uncomfortable.
The UK government’s i.AI incubator shows the trajectory problem at the scale of an entire institution. The team tested 57 AI pilot ideas across departments. 11 moved beyond initial testing — an 81% attrition rate, not a technology failure rate but a dissipation rate. The capability behind those 57 pilots was identical to what powered successful private-sector deployments. Four inertias compounded: regulation slowed every deployment, organizational complexity prevented scaling, civil service culture undermined adoption, and a bias scandal in the DWP fraud detection system destroyed the trust needed for future progress.
Track whether value conversion improves over time, alongside the number of active projects.
The Window Is the Advantage
The dissipation gap will not last forever. Eventually most organizations mature, vendors smooth the integration story, regulators converge on workable frameworks, and the 6% high-performer band widens. When that happens, the compounding window closes — not because the technology stopped advancing but because the asymmetry stopped paying.
While dissipation is high, the organizations that close their own gap are not catching up to a leader. They are becoming the leader, accruing data feedback loops, governance muscle, and workflow redesign that the next entrant cannot purchase retroactively. Neither doom nor boom is right about the speed of AI’s impact. The accurate read is uneven — exponential for capability, flat for diffusion — with the asymmetric returns going to whoever can pull their own diffusion up faster than their sector average.
Choose one barrier the team can describe clearly, and spend the next 90 days addressing it. That might mean agreeing on permissions, redesigning a handoff, measuring the whole job, or giving people training. The aim is to leave the next deployment with less of the same problem to overcome.
The competitive risk is falling behind organizations that are already learning from repeated deployments.
References
- McKinsey & Company (2025). “The State of AI in 2025: Agents, Innovation, and Transformation.” Survey of 1,993 participants across ~105 countries. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
- McKinsey & Company (2025). “The State of AI in 2025.” (Pilot purgatory / scale-to-production data.) https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
- Harvard Business Review (2025). “Overcoming the Organizational Barriers to AI Adoption.” (Trust collapse data: 31% generative AI / 89% agentic AI, May-July 2025.) https://hbr.org/2025/11/overcoming-the-organizational-barriers-to-ai-adoption
- Harvard Business Review (2025). “Overcoming the Organizational Barriers to AI Adoption.” (Executive ROI expectations; people-processes-politics framing.) https://hbr.org/2025/11/overcoming-the-organizational-barriers-to-ai-adoption
- Russell Reynolds Associates (2025). “Why Organizational Inertia Is Killing Your Transformation.” (87% psychological inertia; 68% insight inertia.) https://www.russellreynolds.com/en/insights/articles/why-organizational-inertia-is-killing-your-transformation
- LegalNodes (2026). “EU AI Act 2026 Updates: Compliance Requirements and Business Risks.” https://www.legalnodes.com/article/eu-ai-act-2026-updates-compliance-requirements-and-business-risks
- Deloitte (2026). “State of AI in the Enterprise 2026.” (Governance maturity, agentic deployment, operational incident rates.) https://www.deloitte.com/us/en/what-we-do/capabilities/applied-artificial-intelligence/content/state-of-ai-in-the-enterprise.html
- Pattnaik, D. et al. (2023). “Estimating the Innovation Benefits of First-Mover and Second-Mover Strategies in AI/ML Adoption.” Small Business Economics / Springer. https://link.springer.com/article/10.1007/s11187-023-00779-x
- Machulla, P. (2025). “The Compressed Diffusion Curve: How Generative AI Redefines Innovation Adoption.” Medium. https://medium.com/@pal.machulla/the-compressed-diffusion-curve-how-generative-ai-redefines-innovation-adoption-9becfc780806
- Deloitte (2026) and McKinsey (2025). Cross-referenced governance maturity and adoption growth data. https://www.deloitte.com/us/en/what-we-do/capabilities/applied-artificial-intelligence/content/state-of-ai-in-the-enterprise.html