Behind on the work, and deciding whether to hire or automate

The month the question arrives

It is usually a good month that starts it. Orders are up, everyone is staying late, two things slipped that should not have slipped, and the owner is doing admin at eleven at night because that is the only hour nobody interrupts. Somebody says the obvious thing, which is that we need another person.

Then somebody else says the other obvious thing, which is that half of this could be automated. Both statements are reasonable and both are answers to a question nobody has actually asked out loud, which is what kind of work is piling up.

Hiring and automating are not competing purchases. They fix different failures. A business short of judgement is not helped by a rule, and a business short of hands to apply a rule it already agreed on is not helped by another opinion in the room.

Why the budget framing gets it wrong

Compared on cost alone, a person and a system look like two prices, and the argument becomes a guess about which one is cheaper over two years. That comparison hides the thing that actually decides the outcome.

A person can run a process that has never been written down. That is genuinely their advantage: they absorb the exceptions, they ask when something looks odd, and they carry the reasoning in their head. It is also the trap, because as long as a person is holding the process, the process never gets described. Two years later the knowledge lives in one head and leaves the building when they do.

A system cannot start until the rule exists. That feels like a disadvantage during the build and turns into the main benefit afterwards, because writing the rule down is the part your business was missing. Plenty of automation projects pay for themselves in the specification stage, before anything is switched on, simply because somebody finally wrote what was supposed to happen.

So the honest question is not which is cheaper. It is how much of the work that is drowning you follows a rule anybody could apply, and how much of it needs somebody who knows the business.

The free fix: two weeks and two columns

This costs nothing and it settles the argument better than any spreadsheet of salary comparisons. Give the person who is most overloaded a page with two columns and ask them to log what they do for two weeks. Not timed to the minute. One line per task, in one of two columns.

Column one is for anything that followed a rule. If the answer was already decided, and a careful person with a written instruction could have done it identically, it goes here. Typing an order into a second place. Sending the standard price. Chasing a delivery status. Filling the same form.

Column two is for anything that needed a decision. If the right answer depended on knowing this customer, this supplier or this business, it goes here. Whether to give a discount. Whether a complaint is genuine. Whether to take an order you might not be able to deliver.

Keep a third mark in the margin for interruptions, meaning any time the person had to stop what they were doing because a message or a call arrived. That mark matters more than most people expect, and it is why the tally is worth doing rather than guessing.

Reading the tally

The shape of the two columns gives you a straight answer, and it is usually not the answer people assumed on the first day.

One caution about the arithmetic. Automating work in column one does not shrink your headcount on the day it goes live. It changes what your existing people spend the day doing, and if you promised yourself a saving instead of a capacity gain, you will read the outcome as a disappointment even when the business is clearly better off.

• Column one much heavier: a hire will spend most of their week on rules you already know, which gets expensive quickly. Write the rules and build
• Column two much heavier: hire. A system asked to make judgement calls it was never taught produces confident wrong answers, which cost more than the delay did
• Interruption marks dominating both columns: neither. Your problem is that work arrives in an unstructured channel and gets handled at random, so fix intake first
• Roughly even: the common honest answer is a part time person plus removing the three heaviest items in column one
• Column one heavy but the rules keep changing: wait. Building against a process you are about to replace is effort you will throw away

When hiring is plainly the right answer

There are situations where automation is the wrong purchase and no amount of enthusiasm changes that. If the process changes materially every month, the specification will be out of date before it is finished, and a person adapts for free. If the work is mostly relationship work, where the value is that a specific human being remembers a specific customer, a system replaces the thing you were being paid for.

If you cannot describe the rule on one page, that is not a reason to try harder with software. It is information: the work contains judgement you have not noticed yet, and installing a system will surface it as a series of unexpected wrong answers in front of customers.

And there is a prerequisite people skip. Every system needs an owner inside the business, somebody whose job includes noticing when it stops behaving and knowing who to call. A business with nobody available to hold that responsibility should hire first and automate second, in that order, because an unowned system fails silently and silence is the expensive kind of failure.

The reverse case is worth stating too. If you are about to hire your third person to do work that is entirely in column one, you are buying capacity at the highest available price and adding management to your own week. That is the moment the tally usually pays for itself.

A person can run a process nobody has written down. That is exactly why, five years later, nobody has written it down.

After the two columns, what building the automate side involves

If the tally comes back weighted towards the rule-applying column, the shape of the work is short and published rather than estimated. The audit takes the first week and turns that column into a written procedure, which is the real deliverable. One piece is live inside a fortnight. The second and third arrive across month two, by which point you know whether you still need the hire and, more usefully, what kind of hire it should be.

The judgement column does not disappear when the other one is automated. It concentrates. Anything about a price, a spend, a public statement or a complaint waits for human approval, and a case outside the agreed rule hands it to a person with the record attached. That is a job description on its own, and it belongs in the advertisement rather than in an assumption. The person you hire after this is doing the exceptions, which is a better job and usually a more expensive one.

Whatever gets built sits in accounts under your own logins with its documentation, so you own the system and neither a change of supplier nor a resignation restarts the work. Doing the first one in-house is realistic where somebody already automates things and will still be there next year, and it carries an advantage nobody outside can offer: the person applying the rule by hand is the person writing it down.

The published version of the rule-applying half is Culligan Pakistan, delivered through East River, where an AI powered CRM cut response times and automated lead replies at 140 leads in one month, and Big Texas Land Buyers, where voice agents took over from a cold calling team at more than 500 calls a day. Ibrahim owns build and workflows at Wobble, which works from Karachi, bills month to month and carries 25 engagements across six countries.

Common questions

Is automation cheaper than hiring someone?

Sometimes, and the comparison is less useful than it looks. A system has a build cost, a running cost and an owner who maintains it, while a hire has a salary, a training cost and a management cost. The better question is which of the two can actually do the work in front of you, because buying the wrong one cheaply is still the wrong one.

Will automating mean I have to let staff go?

Not usually, and planning for it tends to disappoint. What normally happens is that the same team stops doing the repetitive half of their week and takes on volume they could not previously handle. If your plan depends on a headcount reduction, say so honestly at the start, because that changes what you should build and how people will treat it.

How do I know a task is ready to be automated?

Write the instruction for it on one page and hand that page to somebody who has never done the task. If they can complete it correctly without asking you a question, the task is ready. Every question they ask is a rule you had not written down, and each one would otherwise have become a wrong answer in production.

Should I hire someone to run the automation?

You need an owner, though rarely a dedicated one at first. Somebody already in the business should be responsible for noticing when a workflow stops behaving and for knowing who to contact. Without a named owner, a system fails quietly and the failure is usually discovered by a customer.

Can I do both at once?

You can, and it works best when they are pointed at different columns. Hire for the judgement work and build for the rule work, rather than hiring someone and then asking them to also implement a system in their first month. Splitting a new person between learning the business and building infrastructure usually produces neither.