Find the bottleneck before you buy the AI
Automating the wrong step changes nothing. A simple four-stage check shows where AI will actually move your numbers.
Most AI projects fail for a plain reason. They automate a step that was never the problem.
A business moves at the speed of its slowest step
This idea comes from The Goal, a business novel Eliyahu Goldratt published in 1984. A system can only produce as fast as its most constrained step. Speed up that step and the whole business speeds up. Speed up any other step and nothing changes.
Here is an example. Say a company is good at marketing and gets plenty of leads, but it takes two or three days to reply to them. A better ad system will not help. The limit is the reply, so that is where the work belongs.
The four-stage check
For most companies the constraint sits in one of four places. Put a number on each.
- Reach. How many people see your business in a week?
- Acquire. How many of them call, fill out a form, or book a meeting?
- Convert. How many of those become paying customers?
- Expand. How long do customers stay, and how much does each one spend over time?
Now compare the steps. Suppose 1,000 people see you, 10 book a meeting, and 1 becomes a customer. Ten meetings from a thousand views may be healthy for your industry. One sale from ten meetings is where to look first. Every hour spent on more ads would be wasted until the close rate improves.
Then break the goal down
Once you know the stage, ask how you could improve it. Then ask how you would do each of those things. Keep going until you reach small, concrete actions: reply to each lead within five minutes, send a follow-up on day two, draft the proposal the same day as the meeting.
That list is your build plan. Some actions are a good fit for AI. Some are a setting in software you already own. Some need a person. The list tells you which is which before you spend anything.
Two warnings
- Measure first. If nobody tracks the numbers today, start tracking before you change anything. Otherwise you cannot prove the result.
- Expect the bottleneck to move. When you fix one step, another becomes the slowest. That is normal. Run the check again.
- Eliyahu M. Goldratt and Jeff Cox, The Goal: A Process of Ongoing Improvement (1984)