Budget for the 70: why the largest share of AI value has no line item
BCG's 10-20-70 rule says 70% of AI value lives in people and process. That is exactly the part no budget template can hold — until the work is mapped and scoped.
See also: Target State – Issue 7
BCG has a name for the gap between what firms spend on AI and the value that comes out of it. It is the 70. Their rule for where AI value comes from is 10-20-70: 10% algorithms, 20% technology, 70% people and process. In The Widening AI Value Gap, their September 2025 study of more than 1,250 firms across 68 countries, only 5% of companies were achieving AI value at scale. Fully 60% were achieving no material value at all, despite substantial investment. The 5% that win do one thing the rest do not: they reshape core workflows end to end, redesigning how work gets done, rather than bolting AI onto isolated tasks. That is the 70. And it is the part that almost nobody funds, because the budget inverts the ratio.
What it is
BCG's own words: 70% of a business's strategic focus should be on the people and processes, 20% on the tech, and 10% on algorithms. Note the word focus. It is not a build order. It is a claim about where the value lives, and the claim is unfashionable. The algorithms improve monthly. The platforms are increasingly good. But value in a complex organisation is created by the way work flows through people, and the way work flows is decided by process, ownership and habit. None of that can be bought with a licence.
Why it matters now
Two things changed this year. First, adoption is no longer a differentiator: every firm in Northern Trust's 2026 survey of 300 asset management leaders reports deploying AI in some form. Every one. Second, the value gap is widening: BCG's future-built 5% generate 1.7 times more revenue growth and 1.6 times higher EBIT margins than the 60% stuck at the back. Universal adoption with almost no value at scale means the bottleneck has moved. It is no longer "should we do AI?" It is "who will fund the part where the value actually is?"
The 70 has no line item. A licence lands on a purchase order. "Two approval gates removed after a fight" fits nowhere in a budget template, so it becomes "change management" and quietly slips. Underneath that is a harder problem: you cannot price what was never structured. Nobody can scope the redesign of a workflow that has never been mapped, so the 70 gets waved at in the business case and dropped the moment the numbers are built. The ratio is not being disbelieved. It is being starved, because it is unscopeable.
What the 70 looks like on a Tuesday
The documented process and the way work actually happens are two different things. The real process is distributed across the people doing it: the workarounds, the exceptions, the duplicate steps and the informal handoffs that never reach the process map. One bank's "simple" ticket process turned out to have thousands of variants once anyone actually looked at what the systems recorded.
In a Lloyd's managing agent, that gap is the operating model itself: hundreds of Visio swimlanes and Word procedures, PDF'd, unsearchable, owned by nobody, and out of date the day the last change programme closed. Ask "who is responsible for this?" or "what breaks downstream if we change that?" and the answer is an email loop.
So before "where can we apply AI?" comes a prior question: do we actually know how this work gets done today? Automating a poorly understood process does not remove the problem. It makes the problem faster.
What firms should do
Make the 70 budgetable. That starts, unglamorously, with mapping: turn the documentation you already have, the swimlanes, the procedures, the manuals, into one connected operating model where every process has an owner, its systems and dependencies are traced, and the coverage gaps are visible rather than assumed. Once the workflow exists somewhere other than in people's heads, redesign stops being a hope and becomes scoped work: this process, these steps, these approval gates, this downstream impact. That is the 70, priced.
Then budget against the ratio. If your AI budget is 80% technology, you have funded the easy third and starved the part that pays. Before the next licence renewal, spend a few days finding out how your work actually gets done. Buy the 30. Budget for the 70.
The opmodal perspective
The 70 is where opmodal works. Not writing more process documentation, but taking the documentation firms already have and turning it into a single, connected, searchable model where ownership, dependencies and gaps become visible, and where workflow redesign can finally be scoped and costed. When the operating model lives in a platform rather than in people's heads, the 70 per cent stops being a campaign and becomes a line item. Map it first.
