Dr. Jonah Tebaa has sat in enough board meetings to notice a pattern that rarely makes it into the minutes: companies approve an AI budget as if the number itself were the strategy, then never formally decide how that number gets divided across projects with very different risk profiles.
In one advisory engagement he describes, a board had approved a six-figure AI budget for the year without incident. When the spend was broken down by category, it turned out to split almost exactly 90/10 — ninety percent toward low-risk automation projects like document processing and ticket routing, and just ten percent toward the one initiative with real potential to change what the company sold. No one had chosen that split deliberately. It had simply emerged, because automation projects arrive with clean ROI cases attached, while the more ambitious bet was harder to quantify and easier to underfund.
The Gap Between a Budget and a Strategy
Tebaa's argument is that a budget figure only tells a board how much it is willing to spend — not what it is buying, when it expects a return, or who is accountable for the riskiest and potentially most valuable bet in the portfolio. Boards that treat "approving the number" as the strategic act, he argues, end up funding whatever is easiest to justify on a spreadsheet, which is almost never the initiative capable of reshaping the business.
He points to a recurring failure mode: ambitious language about transformation at the leadership level, paired with a funding structure that quietly starves the one project that would deliver it. In more than one case he has advised on, the transformative bet was the first line item cut when budgets tightened mid-year — precisely because no one owned it and no one had agreed in advance what its early-stage success should look like.
A Three-Horizon Framework
To correct this, Tebaa asks boards to sort AI spend into three horizons before debating the total figure: Automate (well-understood processes, a payback window of months, tracked in routine operational reviews), Augment (AI embedded into judgment-heavy work like forecasting or underwriting, reviewed quarterly with a named sponsor), and Transform (AI-native bets on new products or business models, with a longer payback horizon and success measured early on by learning rather than revenue).
His starting allocation — roughly 60 percent Automate, 25 percent Augment, 15 percent Transform — is offered less as a fixed formula than as a working hypothesis a board can adjust. What he insists on more strongly is the process around it: each horizon needs a single accountable owner, and the split itself should be revisited every quarter rather than set once a year and forgotten.
Why the Quarterly Re-Split Matters
In Tebaa's framing, an annual "set and forget" allocation defeats the purpose of separating the horizons in the first place. Markets and internal learning move faster than the yearly budget cycle, and a transformation bet that looked adequately funded in January can look badly under-resourced by the third quarter — or, just as legitimately, may deserve to be scaled back once the board has real evidence in hand. The point, in his view, is that the decision gets made on purpose, by named people, on a fixed schedule — not by default.
The board from his example moved, within two quarters of adopting the framework, from a 90/10 split to something closer to 65/20/15, with a named executive sponsoring the transformation initiative and a short monthly update — even in months where the update amounted to little more than documented learning.
Tebaa's closing point to boards is a simple diagnostic: if you cannot say, off the top of your head, how this year's AI budget splits across automate, augment, and transform, that gap is itself the finding. It usually means the split happened by accident — and the bet most likely to matter in three years is running on whatever budget was left over.