Anywhere9 min readUpdated 2026-10-17

What can actually be automated in my business?

Work is worth automating when it passes three tests: it happens often, it follows the same steps each time, and the information it needs already exists in writing somewhere. That covers most of what happens between a message arriving and a job being booked: reading and answering enquiries, qualifying them, sending a quote, chasing the follow-up, raising the invoice, copying the same details into a second system, and assembling the weekly report. It does not cover judgement with no rule behind it, or work whose inputs only exist in somebody’s head. The practical test is whether you could write the steps down for somebody starting on Monday. Whether a tool could do it matters less.

Most answers to this question are a list of tools, which is the wrong end to start from. A tool can only automate a job that already has a shape, and the useful work is finding which of your jobs have one.

So the list below is jobs, not products. Where the job is on it, the tool is a detail and usually one you already pay for.

Which jobs qualify?

Three marks, and a job needs all three. One or two is how a business ends up with an automation that has to be supervised, which costs more attention than the work it replaced.

Often
Enough times a week that somebody would notice it stopping. Once a quarter is not worth building for.
Same steps
The same decision, on the same inputs, reached the same way. Exceptions are fine if you can name them.
Written inputs
The information needed is already in a message, a form, a sheet or a system. Not in a conversation nobody recorded.

A job that fails the third test is not a lost cause, but the first piece of work is writing the inputs down, before anything is automated. That is usually a form or a field in a system, and it is cheap compared to the alternative.

What does automating an enquiry actually mean?

It means the message gets read, answered, priced where a price can be given, and put somewhere it will be seen again, without anybody needing to be free at the time. The reply is the part people picture, and it is the smaller half.

Broken into steps, one enquiry is six jobs, and each one is separately automatable:

  1. Read what arrived, in whichever language it arrived in, and work out what is being asked for.
  2. Decide whether it is real work, a supplier, or somebody who has the wrong company.
  3. Answer it, with an actual answer rather than a confirmation that it was received.
  4. Price it where the price follows a rule, and say what the price depends on where it does not.
  5. Offer a time, and write the booking into the calendar that the team already uses.
  6. Follow up on the ones that went quiet, on a schedule, until they answer or the schedule runs out.

Steps three and six are where the money is, and both fail for the same reason: they depend on somebody being available. The guide on enquiries that go nowhere covers what that costs and how to see it in your own numbers.

What else is on the list?

Everything here passes the three tests in most businesses. Which ones are worth doing depends entirely on volume, which is the first thing worth measuring.

  • Quotes and invoices. Generated from the details already captured, sent, and chased when unpaid. The chase is the part that gets dropped.
  • Onboarding a new client. The same eight messages, folders, logins and reminders, triggered once instead of remembered eight times.
  • Moving data between systems. The same details typed into a second place because the two do not speak. This is the most common one and the least interesting to describe.
  • Scheduling and reminders. Confirmations, the day-before reminder, the reschedule, and the no-show follow-up.
  • The weekly report. Pulled from the systems that hold the numbers, assembled, and sent. Nobody enjoys this one and everybody does it by hand.
  • Document handling. Reading what arrived as a PDF or a photograph and putting the fields where they belong.
  • Review and feedback requests. Sent after the job rather than when somebody remembers, which is what makes the difference to a Google Business Profile.

What should be automated first?

Whichever one is both high-volume and currently failing. Not the most interesting one, and not the one that would be most impressive to show somebody.

The order that works in practice: start where work arrives, because that is where losses compound. An enquiry answered late is often an enquiry lost entirely, whereas a report assembled by hand is an hour spent, every week, and the hour is recoverable later. Fix the leak before the inefficiency.

That is why the work we sell is three fixed workflows rather than an open brief: an instant reply that answers and prices, a booking, and a follow-up. They are the three that sit on the arrival path, they are the same three in nearly every business, and building them once from a template is what keeps the price where it is. The automation page sets out what each one covers.

What is not worth automating?

Four kinds of work. They are worth naming because getting this wrong costs more than a wasted build: you end up with an automation that does the wrong thing for a month before anybody notices.

  • Judgement with no rule behind it. If two experienced people would decide differently and both be right, there is nothing to encode.
  • Anything that happens rarely. A process that runs twice a year will have changed by the second time, and nobody will remember how the automation worked.
  • Work whose inputs are a conversation. If the information only exists in somebody’s head or in a call nobody wrote up, write it down first.
  • The irreversible step. Sending money, signing something, cancelling something. Automate the preparation and leave the last click to a person.

The fourth one is a design rule rather than an exception. Almost every process has one irreversible step, and the right shape is an automation that does everything up to it and then waits.

Does this mean replacing people?

In the businesses we work with, no, and the reason is arithmetic rather than reassurance. The work being automated is mostly work nobody is doing: enquiries answered a day late, follow-ups that never went out, reports that arrive on Wednesday for a decision made on Monday. There is no person to remove from a job that is already not getting done.

What changes is what the team spends attention on. It does also remove a reason to hire: the next person you would have taken on to keep up with volume is the hire that gets deferred.

Where does the automation actually run?

In your own accounts, under your own logins, on the tools you already pay for or on something self-hosted. That is a design constraint rather than a preference, and it decides two things that matter later.

The first is cost. Tool subscriptions for this kind of work run AED 1,500 to 4,500 a month, so a design that reaches for a new product per workflow costs more in subscriptions than in fees. The second is what happens if you stop working with whoever built it. When the workflows live in your accounts, they keep running and somebody else can take them over. When they live on a vendor’s platform, leaving means starting again.

Ask whoever you are talking to which of those two they are proposing. It is the single most useful question in the conversation and it is rarely volunteered.

Questions people actually ask

How do I work out whether I have enough volume to bother?
Count one week. How many enquiries arrived, how many were answered the same day, how many got a second message when they went quiet. Most businesses are surprised by the second and third numbers, and those two are the case for doing anything at all. If the weekly count is in single figures and all of them were answered within the hour, you do not need this yet.
Does it need to be AI, or is this just automation?
Most of it is ordinary automation and has been possible for years: a trigger, some steps, a system updated. The genuinely new part is the step that reads something unstructured (a message written any which way, in Arabic or English, or a document) and turns it into fields the rest of the process can use. That step used to require a person. Everything downstream of it did not.
What happens when something changes in the business?
Someone has to change the workflow, which is the part that gets left out of the sales pitch. Prices change, the team changes, a tool gets replaced. Our monthly fee is for exactly that: keeping the three workflows working as the business moves. If you are buying from somebody who quotes a build and no monthly, ask who does this.
Can you automate something in a system we built ourselves?
Usually, if it has an API or a database we can reach. If it has neither, the options are narrower and sometimes the answer is that the integration is its own project rather than part of a fixed workflow. That is worth establishing before anything is quoted, not after.
Where should I start if I only want one thing?
The instant reply, in whichever channel most of your enquiries arrive on. It is the one with the shortest path between building it and seeing a difference, because the problem it fixes, a message sitting unanswered for hours, is happening today and can be measured today.
All guides

Tell us what your business does

A few sentences is enough for us to price it. If it is not a fit, we will say so in the first reply rather than book a call about it.

Send an enquiry