Not from the most painful task and not from the easiest, but from the work that grows with the catalogue and carries no judgement. One criterion, three questions, and why the first step is nearly always the same.
Start with the work that grows with volume, not with the number of judgements. Entering products, watching prices, keying the same data in a second place — these are tasks that will be larger tomorrow without anyone having got better at them. Do not start with the most painful thing if it requires judgement, and do not start with the easiest thing if it happens once a month.
One question filters almost everything: does the outcome change depending on who does it?
If two people would produce the same result, the task is mechanical — automate it. If the result depends on judgement, experience or a relationship, automation can prepare it but cannot replace it.
Entering a product from a price list is mechanical. Deciding whether to give this customer a discount is not. Transcribing a call is mechanical. Calling back is not.
How many times a week does it happen? A monthly task is not worth it, however unpleasant. A task that happens thirty times a day is worth it even if it looks small.
Does it grow with the business? If twice the products means twice the work, it will catch up with you. If it stays flat, it is a lower priority.
Can it fail quietly? Tasks where a mistake shows up immediately are safer to automate. Tasks where the mistake surfaces months later — a duplicated product, a wrong price — need a review step built in.
In catalogue retail the first step is usually getting products in. Not because it is the most unpleasant, but because it passes all three tests: it happens constantly, it grows in direct proportion to the catalogue, and its mistakes are discovered late.
There is a second reason, more important than the first. Product entry is where data enters the system. Once it is in and in order, everything else becomes possible: prices can be watched, quotes can be assembled, an assistant can answer from the catalogue. Automate quoting first and you will be quoting on bad data.
Each step stands on the one before it. You can skip, but you usually come back.
Approval. A mechanism with no acceptance gate produces volume. Review is the part that makes automation safe.
Negotiation. The discount, the deadline, the exception — that is the salesperson's work. The system can prepare everything for them and spare them the note-taking.
Sending. A quote, an order and a message to a customer go out after human confirmation. The difference between "prepared" and "sent" is the difference between a tool and a liability.
The platform is arranged in that order. Import brings products in and AI describes them, but the product waits for approval. The pricing module maintains prices by rules and records every change. Quotes are assembled from the catalogue and become an order with one confirmation. Communications capture and organise conversations. G1 answers customers from the same data.
It does not have to be rolled out at once. Each module works on its own — but it is worth starting from the bottom.
Which process should be automated first? The one that grows with volume and requires no judgement. In catalogue retail that is nearly always product entry — which is also where data enters the system.
Why not start with the most painful task? Because the pain often comes from a task that requires judgement, and such a task can be prepared but not replaced. Automating it gives a poor result and undermines confidence in everything else.
What is the test for whether a task can be automated? Whether the outcome changes depending on who does it. If two people would produce the same result, the task is mechanical.
Does everything have to be rolled out at once? No. The modules work independently. But the order matters — automating the output on top of disorganised data sends the work straight back to you.
What should not be automated? Approval, negotiation and sending. The system prepares, a person confirms.
How long does the first step take? It depends on the number of suppliers and the state of the catalogue. The practical approach is one supplier all the way through before adding a second.
Tell us what eats your team’s day. We will show you — on our own live system — how much of it can simply disappear.