The 70% problem
THE VALUE GAP · No 3
The 70% problem
Seventy per cent of AI’s value is people and process — and almost nobody has budgeted for it.
———
“Where there is no standard, there can be no improvement.”
— Taiichi Ohno, architect of the Toyota Production System
Somewhere in your organisation there is a process that was designed for a filing cabinet. It was adapted for email, patched onto a shared drive, wrapped in a workflow tool, and it still carries the shape of the cabinet — the extra approval that exists because someone, years ago, lost a form. Now there is a plan to put AI on top of it. The model will read the request, draft the reply, route the sign-off. The process underneath will not change at all. This is the most common AI project in the enterprise today, and it is also among the most likely to disappoint.
The reason is not mysterious, and the clearest account of it comes from an unlikely place — a consulting firm’s own delivery data. Boston Consulting Group describes the anatomy of a successful AI effort as a 10-20-70 split: roughly 10% of the work is the algorithms, about 20% is the technology and data plumbing, and the remaining 70% is people, process and organisational change.¹ Read that ratio slowly. The part everyone buys — the model — is the smallest tenth of the job. The part almost no one budgets for is the other seven-tenths. Most programs invert the ratio. They spend the money and the excitement on the 10%, treat the 20% as an integration ticket, and hope the 70% takes care of itself. It never does.
So the model arrives, capable and expensive, and meets a workflow built for a slower world. It speeds up one step in a chain of eleven, most of which still wait on a human to notice, a queue to clear, an approval that was never necessary. The people doing the work feel the friction first. They were promised help and handed a faster way to reach the same bottleneck.
———
Bolting AI onto old workflows is a frustration engine.
———
The evidence that this is the failure mode, rather than a failure of the technology, is now hard to argue with. McKinsey’s State of AI, published in November 2025, found that redesigning workflows is the single factor most strongly tied to AI’s bottom-line impact — and that only around 21% of organisations have fundamentally redesigned even one workflow for AI.² The capability is everywhere; the reinvention of the work is almost nowhere. Deloitte’s 2026 Human Capital Trends puts a number on the cost of getting the order wrong: organisations that lead with the technology rather than the work are about 1.6 times more likely to miss their expected returns.³ Leading with the tool is not neutral. It is a measurable way to lose.
The vendors, to their credit, have begun to concede the same point in their own data. ServiceNow’s 2026 AI Maturity Index reports that organisations which actually reorganise around AI — restructuring roles and processes, not merely installing the platform — see an average AI return on investment near 160%.⁴ It is a vendor claim, and should be read as one; a firm that sells the platform has every reason to celebrate the customers who commit hardest to it. But the direction of the finding lines up with everyone else’s. The return does not come from the software. It comes from the willingness to change the work the software runs on.
There is a shortcut being taken across the market that deserves naming, because it looks like value and is not. If the work is not redesigned, the return is hard to prove — so a proxy is reached for instead: the headcount line. But a layoff proves only that you spent less; it does not prove the work got better, the customer noticed, or the outcome moved. Cutting your way to a flat result closes the very door — redesign — that would have produced a real one.
This is the argument of the book this series accompanies, arriving from a new direction. The value gap is not a technology problem; it is an approach problem. AI amplifies whatever you already have — it accelerates a well-designed process and, with equal enthusiasm, accelerates a badly designed one. Drop a powerful model onto a workflow full of unnecessary steps and you do not get fewer unnecessary steps. You get them faster, with more confidence, at greater scale. The technology is a multiplier. It has no opinion about what it multiplies. And this is not really an AI insight; it is a platform one. The 70% — the redesigned process, the lifted IT maturity, the automation that finally runs end to end, the improved experience for the person doing the work — is where technology of any kind turns into value. AI simply makes the tenth you buy more powerful, and the seventy-tenths you skip more expensive.
Which is why the distinction that runs through the whole book matters most here. A tool deployment is run as an IT project: scope it, install it, test it, close the ticket, report adoption. A work redesign is run as a business transformation: start from the outcome, ask which steps actually earn their place, and rebuild the process so the AI is doing work worth doing rather than work that should not be done at all. The first produces a green dashboard. The second produces a result. They are not the same undertaking, and they cannot be led by the same person under the same governance.
So the question the book ends on is the one to ask before a single model is switched on: who owns the outcome of this work — not the deployment of the tool, but the outcome — and are they empowered to change the work itself, or only to install software on top of it? If the answer is that a project team will roll out the platform and move on, you already know which tenth of the effort you have funded, and which seven-tenths you have skipped.
So the 70% is not the overhead on the value. It is the value. Fund it as the job, and the model finally has work worth doing; skip it, and you have bought a faster way to run a process you should have retired.
———
THE VALUE GAP — The full argument, and the BRIDGE framework for reinventing the work rather than merely automating it, is in Bridge The Value Gap, out now at rodneyhobbs.com.
———
References
¹ Boston Consulting Group, the 10-20-70 principle for AI value (10% algorithms, 20% technology and data, 70% people, process and organisational change), 2024–2025.
² McKinsey, The State of AI, November 2025 — workflow redesign as the factor most correlated with AI’s bottom-line impact; approximately 21% of organisations have fundamentally redesigned at least one workflow.
³ Deloitte, 2026 Global Human Capital Trends — technology-first adopters approximately 1.6 times more likely to miss expected returns.
⁴ ServiceNow, 2026 AI Maturity Index (vendor claim) — average AI ROI near 160% among organisations that reorganise around AI.
⁵ Gartner, research on post-AI workforce reductions, 2026 — headcount cut at nearly the same rate whether measured AI ROI is high or flat.