What Should I Automate First? The Bottleneck Rule
Short answer: automate the bottleneck, not the irritation. The bottleneck is the one point where work stops moving and everything else waits behind it. Speed up anything else and the queue just arrives at the jam faster.
- Find where work waits. Not where effort happens. Where things sit still.
- Find the single point of failure. The one person whose week off stops the business.
- Find the delay the customer feels and you do not. That is where money leaves quietly.
- Then measure before you build. Write the number down first, or you will never know if it worked. A 45-minute Opportunity Map does this properly in one sitting.
Every owner I have asked this question has an answer ready. It is always the job they hate. The invoices, the quotes, the chasing, the spreadsheet nobody enjoys.
That instinct is understandable and it is usually wrong. Annoyance measures how a job feels, not what it costs. And the jobs that cost the most tend to be quiet, which is precisely how they survive so long.
The bottleneck rule
This one is borrowed from manufacturing. In any process, one step sets the pace for the whole thing. That step is the bottleneck. Improving anything upstream just means work arrives at the jam sooner. Improving anything downstream means idle capacity waiting for the jam to clear.
Only improving the bottleneck improves the output of the business.
It is unglamorous advice, and it is the difference between automation that shows up in the bank and automation that just feels modern. Most first projects fail this test. They speed up a step that was never the constraint, produce a genuine time saving for one person, and change nothing about what the business actually delivers in a week.
Three questions that find it
You do not need a consultant for this part. Sit with a coffee and answer honestly.
One: where does work pile up? Not where the effort is. Where things sit. A stack of quotes waiting to be written. Jobs waiting on a price. Invoices waiting on somebody to remember. If work queues in front of a step, that step is a candidate.
Two: who is the single point of failure? If one named person went off for a week, what stops? That person is a bottleneck whether or not they feel like one, and usually they are the most capable person you have, which is why everything routes through them.
Three: what does the customer wait for? This is the one owners miss, because the delay is invisible from the inside. It might be a quote, a booking change, a start date, an invoice query, or an enquiry that sits too long. Two days feels normal from your side of the desk and feels like being ignored from theirs. You never hear about the ones who went elsewhere, so the cost never lands on your desk. If it turns out to be enquiry response for you, we have worked that arithmetic through.
If the same step shows up in two of those answers, you have found it.
Why the annoying job is so rarely the right one
Do the sums. A task you hate that takes ten minutes a day is roughly forty hours a year. Real, worth fixing eventually, and about a week of work.
Now the quiet one. Say some of your enquiries go cold because nobody replied fast enough. Do not take a number off me, use your own: enquiries a month, average job value, how many you close, and how long a reply currently takes. Whatever that comes to, it is invisible, because the people it affects simply went somewhere else and never told you.
The loud problem costs you about a week a year. The quiet one can cost considerably more than that. Fix the quiet one first, then absolutely fix the annoying one, because morale is a real input and doing work you hate has a cost that never shows up in a spreadsheet.
A worked example
A small firm is convinced the answer is invoicing. It is fiddly, it is monthly, everyone hates it.
Run the three questions and a different picture appears. Work piles up at quoting, not invoicing. One person writes every quote. Customers wait several days for a price, and some have already booked someone else by the time it lands.
Invoicing was annoying. Quoting was the bottleneck. Automating invoicing would have delivered a genuine improvement in how the month felt and precisely zero improvement in how much work came in.
This shape repeats constantly. The thing that irritates the owner and the thing that constrains the business are usually two different steps, and only one of them is worth doing first.
Once you have found it, resist doing everything
The temptation, having found the constraint, is to build a system that solves it completely. Do not. Build the smallest thing that would prove the constraint is real and that moving it changes something.
Three rules keep first projects honest.
- Weeks, not quarters. If you cannot prove it inside a month, you will have committed to more of it before you know whether the first bit worked.
- Write the number down first. Average response time today. Quotes out per week today. Whatever it is, capture it before you build, because memory is generous and will tell you it improved.
- Keep a person in the loop where it matters. Anything touching a customer or money should be drafted by the system and sent by a human, at least until you have watched it be right for a while.
If you want the longer version of how this fits together, the AI workspace guide covers what gets installed and in what order. And before spending anything, the checklist is worth ten minutes.
The honest verdict
Most businesses do not have an AI problem, they have a sequencing problem. There is no shortage of things worth automating. There is a shortage of clarity about which one to do first, and doing them in the wrong order is how people conclude that AI did not work for them.
Find the bottleneck. Build the smallest thing that moves it. Measure it against a number you wrote down. Then do the next one.
Read next:
- AI workspace for small business: what an installed workspace actually contains.
- Too many tools, nothing joined up: the failure mode this advice is designed to avoid.
- The Opportunity Map: the 45-minute session that answers this question properly.
Frequently asked questions
What should a small business automate first?
The bottleneck: the point where work stops moving and everything else waits. Rarely the job that annoys you most.
Why not just automate the most annoying task?
Annoyance measures feeling, not cost. A ten-minute daily irritation is about forty hours a year. A slow reply to enquiries can cost far more, silently, because nobody tells you they went elsewhere.
How do I find my bottleneck?
Where does work pile up, who is the single point of failure, and what does the customer wait for. If a step appears in two of those three, that is your answer.
How long should the first build take?
Weeks, not quarters. Small enough to prove or disprove quickly, measured against a number recorded before you started.
Who wrote this. Jody Murfit spent 30 years in construction, 20 of them co-founding Grocott & Murfit, which grew from 2 people to 70. He now builds AI workspaces for UK small businesses. Jody Murfit on LinkedIn.
Not sure where yours is?
The Opportunity Map runs 45 minutes. We walk your week task by task and score each one, so you leave with a ranked list rather than a hunch. You keep it whether or not you build anything with us.
Book a 45-minute Opportunity Map