Most AI budgets fund exactly one leg of a three-legged stool. MIT's AI Strategy and Leadership Program has a name for what happens next: it tips over — no matter how good that one leg is.
Ask most executives what their AI investment consists of and they'll describe a platform, a licensing agreement, a model, maybe a vendor. That's one leg of the stool. It's also, almost always, the only leg anyone funded.
The framing comes from MIT's AI Strategy and Leadership Program, in a session built around a deceptively simple idea: an AI investment doesn't stand on technology alone. It stands on three legs — technology, organizational architecture, and leadership — and if you invest heavily in one while ignoring the other two, the whole thing tips over. Not eventually. Predictably.
Technology is the easy leg to build. It's a line item. It has a vendor, a contract, a go-live date. Procurement understands it. Finance can model the ROI on a spreadsheet. It's the leg that gets a press release.
It's also the leg that, on its own, does almost nothing. A model that no one has the authority to act on, deployed into a reporting structure that hasn't changed, championed by no one above the project-manager level, produces exactly what you'd expect: a tool nobody quite adopts and nobody can quite explain why not. I've written about what that looks like in practice in The AI Execution Cliff: Why Adoption Is Rising and Results Aren't — the technology leg was never the problem there either.
The second leg is structural. Who actually has the authority to make the decision the AI is informing? Who's accountable when the model is wrong? Does the org chart, the approval chain, and the reporting structure reflect the fact that a machine is now doing part of the work a person used to do end-to-end?
Most organizations never touch this leg. They bolt AI onto an org chart designed for a world where every decision had a single human owner, and then wonder why decisions stall in ambiguity — nobody's sure if the AI's recommendation, the analyst's judgment, or the manager's sign-off is the actual decision. I covered this exact failure mode in AI Isn't Taking Your Job — It's Changing Who's Responsible: AI doesn't just change what work gets done, it changes who owns the outcome. If nobody redraws that map, the second leg is missing entirely.
The third leg is the one that's hardest to buy and easiest to skip: leadership that actively, visibly moves with the change instead of announcing it and stepping back. Not a memo. Not a kickoff meeting. Sustained, visible reinforcement — leaders modeling the new behavior, absorbing the early friction publicly, and treating the rollout as their problem to own rather than IT's problem to execute.
MIT's own research on how AI is reshaping leadership itself is worth pairing with this: as I wrote in AI Isn't Coming for Leadership. It's Coming for Half of It., the leadership behaviors AI actually automates are the transactional ones — status updates, scheduling, first-draft synthesis. What's left, and what becomes more important, not less, is exactly the sustained, human reinforcement work this third leg requires. Leaders who treat AI adoption as someone else's project are, functionally, choosing not to build this leg at all.
If you invest in AI but haven't addressed organizational architecture and leadership, the investment will almost certainly fail. All three legs have to move together.
The instinctive fix leaders reach for is "we need to be more nimble." It's worth being precise about what that word means, because the three terms get used interchangeably and they aren't the same thing. Lean is about removing waste from a process. Agile (capital A) is a specific software development methodology. Nimble — the organizational version, drawing on Deborah Ancona's research on distributed team autonomy — is about empowering distributed teams to make decisions without routing everything through a central approval chain.
In the end, all three are really about the same underlying thing: empowering people and reducing bureaucracy. But nimble is the one that matters most for AI adoption specifically, because AI's value shows up fastest at the point of use — the analyst, the case manager, the account rep — not at the point of central approval.
Here's the caveat most "be more nimble" advice skips: nimble isn't right for every organization, and it's actively dangerous in some. Regulated environments, operations-heavy businesses, anywhere a safety-critical procedure exists for a reason — airlines, healthcare operations, financial controls — can't hand teams full autonomy to improvise around AI outputs. The fix isn't to force the whole organization to become nimble. It's to run a parallel innovation track alongside core operations, where a smaller team has room to experiment, while the operational core keeps its procedures intact. Done carelessly, this creates an "us vs. them" dynamic between the innovation track and everyone else — which is itself a leadership-leg problem, not a technology one.
The other practical tool from the same MIT session is a stakeholder map — a simple 2x2 grid plotting influence against impact for everyone the AI rollout touches. It answers three questions before you're mid-rollout and guessing: Who do you actually need on board? Who's a likely detractor, and why? What can you offer each group to bring them along, rather than assuming goodwill will carry them?
This is a five-minute exercise that most AI rollouts skip entirely, and it's usually the cheapest fix on this list. It belongs in organizational architecture and leadership both — it tells you where the structural gaps are, and it tells leadership exactly who needs direct, personal attention instead of a company-wide email.
Before your next AI investment gets approved, ask one question of the budget: which leg is this funding? If the honest answer is "just technology" — again — the investment is already structurally at risk, regardless of how strong the model is. The fix isn't a bigger technology budget. It's deliberately building the other two legs with the same rigor the technology leg already gets.
That's not a harder problem to solve. It's just a different one — and it's the one most AI budgets never touch.
The AI Efficiency Audit maps all three legs of your AI investment — technology, organizational architecture, and leadership — and shows you exactly where the gap is before it costs you the rollout.
Start with an audit →Sources: MIT xPRO. (2026, May 7). Module 8 office hours: Nimble organizations, systems thinking, and stakeholder mapping [Live session]. MIT xPRO AI Strategy and Leadership Program, Massachusetts Institute of Technology.