The 30-hour question: what is your team still doing manually?

Ask a business owner where their team's time goes and you will get an honest answer that is usually wrong. Not dishonest — wrong in the way memory is wrong. The work people remember is the work that was difficult. The work that eats the month is the work that was easy, and happened eleven times.

So the question is not "what takes the longest?" It is "what happens most often?" Those are rarely the same task, and only the second one is worth automating.

Find it by counting, not by guessing

Here is an exercise that costs a week and no money.

Ask everyone to keep a single running list: every task they do that they have done before, with a tally mark each time. No timings, no categories, no forms — a page, or a note on a phone. Most people resist for two days and then find it oddly satisfying.

At the end of the week, collect the lists and sort every line into one of four boxes:

• Often, quick. Twenty times a week, two minutes each. This is the box nobody expects and where most of the month is hiding.
• Often, slow. The obvious candidates. Everyone already complains about these.
• Rarely, quick. Leave it alone. Automating this is a hobby.
• Rarely, slow. Usually a process problem, not an automation problem. Worth a conversation, not a tool.

The first box is the one to start with, and it is always a surprise. Forty minutes a day of copying details from one screen to another does not feel like anything while it is happening. Over a month it is more than a working week.

The four questions that separate a real candidate from a wish

Once you have a shortlist, put each line through these.

Does it have a clear trigger? Something has to start it: a form arrives, a payment clears, a date passes, a status changes. "When someone remembers" is not a trigger, and a task without one usually needs a process decision before it needs software.

Is the information already somewhere? Automation moves and reshapes information that exists. If the detail lives only in somebody's head or in a photograph of a handwritten note, that is the thing to fix first.

Is the rule writable? Say the rule out loud. "If the enquiry mentions a wedding, it goes to Ama; everything else goes to the shared list." If you cannot say it in one sentence, it is a judgement, and judgements should stay with people until the judgement itself has been settled.

Would anyone notice if it broke? This is the one people skip. An automation nobody is watching will fail silently on a Tuesday and nobody will find out until a customer complains in a fortnight. If no one owns it, do not build it.

Start at the handoffs

If you want a single place to look, look wherever work changes hands. Sales to delivery. Delivery to invoicing. Enquiry to whoever answers it. Handoffs are where things are retyped, where things wait, and where things get dropped — and they are the easiest to describe, which makes them the easiest to fix.

A typical first automation is not impressive and is worth doing anyway: a form submission that creates a record, notifies the right person, sends the enquirer a real acknowledgement, and sets a reminder if nobody has replied in a day. Four small steps. Removes the two failures that actually cost business — the enquiry nobody saw, and the enquirer who heard nothing and messaged a competitor.

The best first automation is one that removes a silence, not one that removes a person.

What to expect, honestly

Automating a task rarely gives you the whole time back. There is setup, there is a period of checking that it works, and there is a small permanent cost in maintenance — things change, and something that talks to five systems will break when one of them changes. Expect to keep perhaps two-thirds of the time you save, and expect the first month to be roughly break-even.

The compounding part is what makes it worth doing. The second automation is faster than the first, because the connections and the records are already in place. By the fifth, most of the work is deciding what should happen rather than making it possible.

The trap: buying the platform first

The common sequence is backwards. A business buys a platform, then looks for processes to put on it, then discovers the processes do not fit and builds around the tool. The cost is not the subscription — it is the year spent shaping the business to the software.

Count first. Pick one line off the list. Make that one thing work end to end, with somebody's name against it. Then ask what you actually needed, and buy that.