AI or Another Hire? How to Decide for a Growing Business

How to run the comparison on paper

Most hire or build arguments happen with no numbers in the room. Write the job down instead. List every task the new person would do in a normal week, put an honest hour estimate beside each one, then mark each task as judgment or repetition.

Cost the hire properly while you are there. Salary is the smallest part. Add the shortlisting weeks, the manager hours lost to training, the software seats, and the two or three months before anyone works at full speed. Compare that over twelve months rather than one, since a setup cost happens once and a salary does not.

If more than half the hours on that list are repetition, the job as written is two jobs, and you were about to pay one salary for both.

Can you split the job instead of choosing one?

The hire or build question usually has a third answer. In most growing businesses the split inside a single role lands near half and half, sometimes worse. Build the repetition first, then hire for what is left, which is a smaller and better defined role, easier to fill and cheaper to keep.

Pay attention to the grey middle while you mark up the job description. Plenty of work needs a person to approve it but not to prepare it. Let a system prepare the quote or the report, and let a human press send.

When hiring is still the right call

Automation has a floor. A task that runs four times a month rarely earns back the cost of building and maintaining something to do it.

One case is worth naming plainly. If you have never had a person in that function, hire the person. Automating a sales process you have never run yourself means encoding guesses. Let someone do it by hand for a few months, watch what they repeat, then build that. The Wobble view is that the second hire in a function deserves the harder question, not the first.

• The task runs a handful of times a month, so the build outlives the benefit.
• Nobody describes the process the same way twice.
• The customer needs to hear a voice before they commit.
• A bank, an auditor or a regulator requires a named person to sign.

Who owns the system once it is running?

The fair worry about building instead of hiring is that you end up depending on something nobody in the room understands. The fix is ownership. Every system running part of your business needs a named owner who knows what it does, where to look when it stops and who to call. That person does not have to be technical, only accountable.

Ask any vendor a plain question before you sign. If we stop working together next month, what do we keep? Wobble builds so the answer is all of it. The accounts are yours, the workflows sit in your systems and the data stays with you. Renting output suits a campaign. For the part of the business that repeats every day, owning it is worth more.

Every system that touches a customer needs a name beside it.

Where the split lands, and what building the repetitive half involves

The third answer above, splitting the role, has a timetable behind it. The audit takes the first week and produces the two columns rather than a recommendation. The repetitive half is live inside a fortnight, and only then does the advertisement get written, because the role you were about to advertise is not the role you will need. Whether it is worth doing is settled by the arithmetic already described: frequency multiplied by hours, set against what a build costs once and what it costs to keep.

It does not always work, and the failures have a recognisable shape. A task that runs four times a month cannot repay a build, whatever the enthusiasm in the room. A process that changes materially every month cannot either, because the specification is out of date before the testing finishes. And where the manual step is really a quality check nobody has written down, removing the person does not remove the mistakes. It sends them to the customer instead.

Whatever gets built, the stop is a person's job and a real one. Anything about money, a price, a public statement or a complaint waits for human approval, and a case outside the rules hands it to a person with the record attached. That hour a week is what the ownership section above is actually describing. It can be your own team's hour too: where somebody in-house already automates things and will still be there next year, building the repetitive half internally is cheaper than either option in the title.

The published version of splitting a role rather than filling it is Quillon, where producing audit-ready training packs stopped being a hand build and became a single audited automation of 34 AI nodes, and Patel Group, where leads were managed informally and follow ups were missed until an AI driven CRM was built around exactly those bottlenecks. Haad owns growth and client solutions at Wobble and does the diagnosis that produces the two columns in the first place.