There is a quiet race going on in mid-market boardrooms: the race to have AI. A licence here, a copilot there, a pilot someone can point to in the next board pack. It feels like progress. It usually isn't. Buying the tool is the easy 20%. The hard — and valuable — 80% is getting your organisation to actually work differently, and that is where most initiatives quietly stall.

The numbers bear this out. Gartner predicted in 2014 that at least 30% of generative AI projects will be abandoned after proof of concept by the end of 2025, citing poor data quality, weak risk controls, escalating costs and - tellingly - unclear business value [1]. This was true, by most empirical evidence.

Meanwhile BCG's now widely-cited "10-20-70" rule holds that only about 10% of the value from AI comes from the algorithms and 20% from the technology and data; the remaining 70% comes from people and process [2]. If that's right, then a strategy built around procurement is optimising the smallest slice of the prize.

A successful AI transformation isn't a transformation at all

The word "transformation" sets the wrong expectation. It implies a destination — a before and an after, a big reveal. Real change doesn't work like that. What actually delivers is a process: a continuous journey in which every initiative and every process change is underpinned by a clear, specific business reason. Not "we're doing AI," but "approvals take four days and cost us two lost orders a month — let's fix that."

That framing matters because it changes what you build and when you stop. When each step has to justify itself in business terms, you stop chasing impressive-looking pilots with no owner, and you stop funding the demos Gartner is warning about. The question shifts from "what can this tool do?" to "what is it worth to us, this quarter, on this process?" That single discipline eliminates most of the projects that would otherwise die in proof-of-concept.

Redesign the machine — in small steps, without breaking what works

Think of your business as a machine that already runs. The instinct with AI is to rip out a whole subsystem and drop in a shiny new one. Resist it. The goal is to redesign the machine — but incrementally, one connected improvement at a time, each contributing to the overall health of the business without breaking the parts that already work.

Small steps aren't a compromise; they're the safest path to compounding gains. A useful place to start is the seams between processes that people run manually every day. Take approvals — one of the most common and most overlooked. In most firms an approval and the work it triggers are two separate manual acts: someone approves a purchase, then someone else keys the order, updates the ledger, notifies the supplier. The approval still needs a human judgement. But everything around it — creating the request, routing it, and executing the downstream steps once it's granted — does not need to be created by hand. Connect the approval and its follow-on execution into a single automated flow, and you keep the human decision exactly where it belongs while removing the manual drudgery on either side of it. That's a redesign of the machine that breaks nothing and pays back immediately.

The people closest to the process should design the change

Here is the part most digital transformation programmes get backwards. They hand the redesign to a central team or an external vendor and then wonder why adoption is poor. The people who understand where the friction actually lives are the ones doing the work every day — and they are more ready than leaders assume. McKinsey's Superagency in the Workplace research is blunt about it: "the biggest barrier to scaling is not employees—who are ready—but leaders, who are not steering fast enough" [3]. In the same study, employees turned out to be three times more likely to be using generative AI at work than their leaders expected [3].

So involve them. Coach the people nearest each process to think in terms of automation — where does a task wait, get re-keyed, or get chased? — and then give them safe access to tools and time to try. You are not asking them to become engineers. You are asking them to spot the twenty small redesigns a consultant would take months to find, and giving them permission to test.

Governance vs. flexibility: the hardest balance right now

The catch, and it is the genuinely difficult part of this stage, is the balance between governance and flexibility. Too much control and you're back to a central bottleneck that kills the experimentation you just unlocked. Too little and you get "shadow AI," inconsistent data handling, and risk you can't see. The answer is not to pick a side but to set clear guardrails — what data can go where, which decisions must stay human, how a working experiment gets promoted into a real process — and then let people move freely inside them. Get that balance right and governance stops being the brake; it becomes the thing that lets you go faster safely.

None of this requires a moonshot. It requires clear business value behind every step, small changes that respect what already works, and the people who run the machine helping to redesign it. That is the whole game.

*At InfinityX Digital we help UK SMEs turn AI from something they "have" into something that changes how the business runs — one valuable, low-risk step at a time. If that's the journey you're on, start a conversation at [infinityxd.uk](https://infinityxd.uk).*

---

Sources & quotes

[1] Gartner — "Gartner Predicts 30% of Generative AI Projects Will Be Abandoned After Proof of Concept By End of 2025"** (press release, 29 July 2024).
[https://www.gartner.com/en/newsroom/press-releases/2024-07-29-gartner-predicts-30-percent-of-generative-ai-projects-will-be-abandoned-after-proof-of-concept-by-end-of-2025](https://www.gartner.com/en/newsroom/press-releases/2024-07-29-gartner-predicts-30-percent-of-generative-ai-projects-will-be-abandoned-after-proof-of-concept-by-end-of-2025)

[2] BCG — "The CEO's Guide to Maximizing the Value of AI" / 10-20-70 rule** (BCG, 2024).
[https://www.bcg.com/publications/2025/to-unlock-the-full-value-of-ai-invest-in-your-people](https://www.bcg.com/publications/2025/to-unlock-the-full-value-of-ai-invest-in-your-people)

[3] McKinsey & Company — "Superagency in the workplace: Empowering people to unlock AI's full potential at work"** (28 January 2025).

[https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/superagency-in-the-workplace-empowering-people-to-unlock-ais-full-potential-at-work](https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/superagency-in-the-workplace-empowering-people-to-unlock-ais-full-potential-at-work)

InfinityX Digital

Your value co-creators!

Start a conversation

London, UK — remote across Europe

(C) 2026 InfinityX Digital-LTD - Your value co-creators

Home-Work-Who we are-Services-Contact