Most organisations can now say, with confidence, that AI has made people faster. Far fewer can say what that speed is actually worth. That gap, between individual productivity and organisational value, ran through almost every conversation at a private dinner we hosted recently on where AI and agentic transformation are actually creating value.
We brought together senior technology and operations leaders to talk openly about where the value in AI transformation actually sits, and why so much of it isn't showing up where organisations expect it to. The conversation stayed grounded in real experience, candid and constructive in equal measure.
Why doesn't AI speed translate into measurable business value?
The room split cleanly into two camps. On one side, hands-on doers: experimenting, shipping, moving faster than they ever have. On the other, the people who have to justify the spend, caught between heavy upfront investment and genuine uncertainty about where token costs are heading.
Nobody doubted the speed AI unlocks, or the pace at which it's being adopted. What they doubted was whether the return holds up across the full investment period, and whether rising token economics quietly erodes it.
We’ve all seen the research and reports on the struggle to realise value at the enterprise level. CF/BEYOND is centred around pragmatic application of AI, modelled to scale and measured throughout. It goes beyond vanity usage metrics, focuses on total cost of ownership and ensures value of human integration is maintained.
Why is cost-saving still the default lens for AI investment?
Operational leaders in the room described mandates built almost entirely around cost reduction. Technology leaders framed it as efficiency: doing more with less. What was missing across both groups was any real maturity around value uplift, the question of what new value the organisation creates, not just what it strips out.
This showed up again in a more personal way. Technology professionals feel the difference day to day. They're more proficient, faster, less blocked, if occasionally more isolated. But there was a shared, slightly uncomfortable recognition in the room that this personal productivity isn't translating upward into value at the organisational level. The strategic payoff isn't showing up yet, even where the individual gains clearly are.
Where does value actually sit in the business, and who should decide?
Rather than asking where AI could be applied, leaders need to use their organisation’s own business logic to identify where value actually sits, then test whether AI can reach it. That identification isn’t a simple task, you have to remove the historical layers that have accumulated to overcome legacy constraints.
That shift moves value definition out of IT, where it's traditionally been measured as a technical output, and into the business, where it becomes a strategic exercise. It also explains why measurement is maturing: usage and adoption metrics are giving way to genuine outcome tracking, because usage was never the point. (add in a point around why Chief AI Officers might contradict this approach).
The discussion connected this directly to a harder question most leaders haven't fully answered: where does your business logic actually live? Not value captured one person at a time through individual productivity, but value designed deliberately into the operating model itself. Until organisations can answer that, they're managing activity, not value.
Every durable initiative discussed in the room started from that same business problem, not from a piece of technology looking for a use case. First-principles thinking still has to underpin activation and deployment, and skipping that step was named as the single most common root cause of pilots that never scale beyond the team that built them. Organisations that treat AI as the starting point for a system, redesigning strategy, product, engineering and operations together, tend to see that value compound. The ones that bolt AI onto existing structures tend to see isolated wins that plateau.
Why is institutional knowledge becoming the real AI differentiator?
There was strong agreement in the room that the differentiator has shifted away from the model itself and toward the proprietary context wrapped around it. Leading organisations are actively harnessing institutional knowledge and embedding it into agents, turning tacit expertise into a reusable, scalable asset.
This is what separates a generic AI deployment from one that compounds value over time: an organisation that captures what its best people know and makes it permanently available, rather than one that re-solves the same problem in every team, every time it comes up.
Why haven't operating models caught up with AI's capability?
Across the evening, one theme kept resurfacing in different forms: technology adoption is outpacing organisational change. Most businesses haven't redesigned operating models, processes or ways of working to absorb AI's capability. They've bolted it onto old structures instead.
Genuine, measurable value shows up where process redesign and organisational design happen alongside deployment, not after it. That's a harder, slower thing to do than shipping another pilot, which is exactly why so few organisations have done it. Redesigning the organisation is understandably the part that makes leaders most nervous, but the same principles that make AI delivery work apply here too: segment the change, deliver it in pieces, and let each one prove its value before moving to the next. You don't have to eat the elephant whole.
Regulation showed this most clearly. There was real debate about whether regulation slows the cost-saving thesis some are banking on. KYC came up repeatedly as the example: enormous automation potential, sitting inside a regulatory environment that moves nothing like as fast. Where the technology is ready but the regulatory structure isn't, the ceiling on value sits with the structure, not the AI.
What this means for organisations scaling AI
Most leaders are still finding a firm grip on how AI delivers value beyond cost, efficiency and productivity. That's the maturity gap CF/BEYOND is built to close: treating value definition as a strategic exercise, redesigning the operating model alongside the technology, and building an Intelligence Engine that turns institutional knowledge into a compounding asset rather than a one-off pilot win.
Speed on its own isn't the advantage it looks like. Real progress belongs to the organisations that can point to exactly what their AI investment is worth, and that are prepared to redesign how the business works to capture it, rather than simply adding more tools.
FAQs
Meet the author
Paul O'Beirne is Director of Transformation at CreateFuture. With a consulting background across strategy, operating model design and technology, he specialises in turning strategic ambition into executable change, most recently through his work building native AI operating models.
Industry Insights
Explore the latest thinking from our industry and tech experts.
What is AI native?
Why do AI experiments fail and how do you scale AI-native delivery?
CreateFuture expands its Anthropic AI capabilities as Version 1 Group achieves Preferred Services Partner status