People ask AI for faster horses. How do you build the car?
By Nicola Sirchia. Published . 2 minute read.
The short answer
When people ask AI for faster horses, the job is to find the problem behind the request and build the car. Most AI projects that disappoint did exactly what was asked: they made yesterday's work a little quicker. The value comes when you stop taking the request at face value and redesign the work.
What do people actually ask for?
People ask from inside the limits of what they know. So they ask for faster reporting, quicker summaries, better search and easier content creation.
Useful? Yes. Transformational? Not always. A company can use AI to optimise the way work is done today and never ask what the work should become.
Which questions find the real need?
Instead of "what did they ask for?", I ask four things:
- What friction are they experiencing?
- Which decision is too slow?
- Which work should disappear completely?
- Which outcome are we really trying to improve?
How do you get from the request to the car?
- 1Capturethe request
- 2Decodethe friction
- 3Findthe latent need
- 4Redesignthe workflow
- 5Prototypethe future state
- 6Measurebusiness impact
- 7Scalewhat changes the game
When is copying what works good enough?
Often. Looking at what others do, benchmarking and adapting saves time and reduces risk. It improves what already exists. It does not create anything new.
Going back to first principles is slower and harder. You ask which constraints are real and which are habits, legacy processes or "this is how it has always been done". A leadership team needs both: analogy to optimise, first principles to rethink.
What should a leader take from this?
People describe the pain. Leaders have to imagine the jump. The goal of AI is not a faster version of yesterday.
Sources
This piece comes from my own work with teams. It uses no outside data.