"Lead with AI" is a mistake
Starting with the technology is the most common way an AI project fails inside an expertise-led company. Start with the constraint that is costing you money.
Contents
Leading with AI means choosing the technology before naming the problem. It is the most common way these projects fail in a company built on expertise. The work that returns money starts somewhere else. It starts with the constraint that is already costing you capacity, margin or clients, and with the judgement your people apply to it.
What does leading with AI actually look like?
It usually starts with a tool, a demo, or a board question about what the company is doing. A pilot follows. There is a working prototype, and nobody can say which number it moves.
The pattern is easy to spot once you know it. The project has a technology in its name and no owner outside the technology team.
Why does starting with the technology fail?
Because it inverts the order of the decision. A technology choice is cheap to reverse. A commercial commitment is not. Teams that start with the tool spend the next year defending the tool.
There is a second cost, and it is quieter. The people whose judgement makes the work valuable are asked to review something they did not shape. They disengage. The system then loses the only input that made it worth building.
What should come first?
Leading with AI, as Mayfly Ventures defines it for expertise-led companies, is selecting a technology before naming the constraint it is meant to remove.
The first question is narrower and harder to answer. Where does capacity actually run out? In most expertise-led companies it runs out in one of three places: growth that depends on hiring, judgement held by a few people, or value locked inside service delivery.
How do you know which constraint is yours?
Three signs, and they rarely appear together:
- Growth costs more than it returns. Every new client needs another senior hire. Revenue climbs and margin sits still.
- Work queues behind the same names. Two or three people appear on every difficult job, and nothing hard moves while they are away.
- Methods never leave delivery. The company owns the knowledge and still sells it by the hour.
Each one points at a different first move. None of them is a technology decision yet.
What changes when you start with the constraint?
The scope gets smaller and the argument gets easier. You are removing a specific limit rather than buying a capability, so you can say what it is worth before anyone writes code.
Technology still arrives. It arrives later, as the mechanism, once the constraint is named and the judgement is captured. That order is the whole difference.
Where to start
Pick the constraint you can describe in one sentence. Write down what it costs you this year. Then ask what would have to be true for it to stop being a constraint.
If the honest answer is a tool, keep going. If the answer is captured judgement, you have found the work.
Keep reading
Growing without hiring at the same rate
Revenue climbs, margin stays flat, and every new client needs another senior hire. What breaks first in a services business, and what to change before it does.
When judgement lives in five heads
In most expertise-led companies the hardest decisions queue behind two or three people. What that costs, how to see it early, and how to move it.
Venture studio, consultancy or agency?
Four models, four different jobs. What a consultancy, an agency, an in-house team and a venture studio are each good at, and how to choose without guessing.