Start Now
The big IT initiative playbook ran like this. The assessment. The design thinking session. The wall of post-it notes, dissecting every feature before anyone builds anything. Business cases. Then months, sometimes years, of development to deliver a product users marginally accept.
I am not interested in that anymore. I am interested in motion. Small motion. Controlled motion. Real motion on a real workflow, with a real operator at the validate gate, learning from what breaks.
Start first. Start small. Start now.
That is the whole idea.
Look at what starting produced
In 2022, four MIT computer science graduates founded Anysphere and built Cursor. Michael Truell, the CEO, is twenty-five. They started on a problem they understood, for users they knew, on codebases that were never clean. They shipped. They iterated. They kept the team lean. They did not wait for the enterprise to get ready for them.
Four years later, the company had surpassed three billion dollars in annual recurring revenue. In June 2026, SpaceX announced an agreement to acquire Cursor in a deal valued at sixty billion dollars.
I am not telling you to copy a startup. I am telling you to notice the shape. A small group started first. They learned in public. The market responded to the learning, not to a readiness deck.
That is available inside incumbent organizations too. Not at venture speed. Not with venture capital. But the same physics apply. Pick something. Build something. Run it. Improve it. The organizations that win are not the ones that waited longest. They are the ones that started earliest and iterated honestly.
What I learned on stage
A few years ago I stood in front of a room and quoted the statistic everyone quotes. Eighty-five percent of AI projects fail. I had used versions of that line before. It gets attention. It makes the room sit up straight.
My answer in that same talk was not to stop. It was to start small, start fast, and iterate in controlled scope. That was the counter then. It is still the counter. The difference is what I lead with now.
I lead with start.
The Gartner forecast behind the eighty-five percent figure, if you look it up, was about erroneous outcomes from bias and execution quality, not a scorecard of abandoned projects. The number got flattened over time. Other studies since measure other things. They are worth knowing. They are not a reason to stand still.
I take fear out of the equation by starting. Not by pretending risk is gone. By putting something small in production, putting a human on the gate, and learning faster than a deck can teach you.
What your operators already prove
Some consultants will pitch a data management engagement to get your data ready for AI. Clean it up first, then automate. It sounds responsible. It can also become the same playbook in a different costume: assess, harmonize, wait, and defer the workflow that would actually teach you what is broken.
If your data is not good enough for AI, how is it good enough for the people making decisions in your operation today?
Your teams work around gaps every day. They hold context. They call the colleague who knows the account. They stitch together three systems by hand. They know which exceptions are normal and which ones escalate. That is not a failure of data quality. That is how work gets done inside real systems.
AI should meet that same reality. Same inputs. Same standard. Same validate gate. Not a cleaner warehouse that does not exist yet.
Perfection is the enemy of progress. That does not mean skip accountability. It means do not let perfect become the reason you never learn what ninety days of building would show you.
How to start
Pick one workflow where delay is already costing you something. Scope it small enough to control. Connect to the systems you already run. Encode what good looks like at the validate step. Put your best person on the exceptions. Run it. Measure cycle time from signal to validated close. Improve from what you learn. Expand when the loop earns it.
Start first means the workflow comes before the enterprise program.
Start small means one team, one process, one gate.
Start now means the next ninety days teach you more than the next ninety pages of assessment.
Start first. Start small. Start now.
What would you know in ninety days if you started this week?