People are freaking out over this stat:
MIT research says ~95% of enterprise AI pilots show no measurable impact on the business.
Here’s what I think:
Most organizations design pilots that were never going to scale in the first place.
Economist John List explains the trap in his useful book ππ©π¦ ππ°ππ΅π’π¨π¦ ππ§π§π¦π€π΅. A pilotβs purpose isnβt to prove that something works once. Itβs to test whether the results will hold when scaled into the complex reality of your org and market.
Yet thatβs not how most pilots are built — even before AI.
And now orgs are designing and executing doomed pilots even faster.
Leaders under pressure to show βweβre using AIβ push teams to launch something quickly. Those teams often lack the cross-functional capabilities (datasets, toolsets, skillsets, and mindsets) to design and deliver a pilot that reflects scale conditions.
The pilot collapses as soon as special conditions are overwhelmed by reality.
Poorly designed pilots is another example of why generative AI needs generative leadership.
Generative leaders…
β’ Define outcomes that matter: What is the future we are creating for our customers and/or team members? How will we define success?
β’ Convene cross-functional teams to innovate the workflow changes that adoption at scale requires.
β’ Create safety for team members to speak up “hey, I suspect this will break once we rollout to non-technical users — what can we do?”
For the healthcare leaders my team and I work with, the implications are especially urgent. Most care organizations do not have the bandwidth to fail at AI — and for sure, not 95% of the time.
Let’s give AI pilots every chance for success by designing them strategically, cross-functionally, and thoughtfully.
If you’re wondering how to kick off AI initiatives that beat the odds, reach out. Let’s talk.