These are ranked by how little they disturb your existing process, not by how impressive they sound. The best first automation is the one nobody has to change their behaviour for, because adoption is where most of these die.
The six
-
1. Supplier invoices into the accounts system
The most common job we are asked to remove, and the easiest to justify. The document arrives, gets read, gets typed in. Extraction and matching handle most of them, the rest queue with both documents side by side.
-
2. The same twelve customer questions
Order status, opening hours, stock, returns. Answered from live systems rather than a script, on whichever channel your customers actually use. Removes inbox volume without changing how anyone works.
-
3. The Monday morning report
Assembled by hand from two or three systems every week by someone senior enough that it is expensive. Builds itself overnight and lands before anyone arrives.
-
4. Quote and follow-up chasing
Quotes go out and are never chased consistently. A schedule with sensible stopping rules recovers work you already won and forgot about. Almost no process change required.
-
5. Onboarding document chases
The same four documents, requested by hand, every time. A sequence that asks, reminds, and tells you what is outstanding.
-
6. Compliance and certificate expiry
Nothing is required until something expires quietly. Continuous collection rather than a scramble before an inspection.
Why this order
The first four touch documents and messages that already exist. Nobody has to log into a new system or learn a new habit; the work simply stops arriving. The last two require a little more cooperation, which is why they come after you have something working and people trust it.
What to do this week
Ask the three people who do the most admin to write down every recurring task and roughly how long it takes. Do not analyse it. Just list it. The candidates will be obvious within an hour, and you will usually find one nobody had thought of.
Then run the arithmetic on the automation calculator and start with whichever pays back fastest.
Questions we get asked
Should we start with the biggest problem?
Usually not. Start with the one that pays back fastest and touches fewest systems. A quick win buys the goodwill to tackle the big one properly.
How many should we do at once?
One. Get it live, watch it for a fortnight, then start the next. Parallel first projects tend to produce two half-finished ones.
What if our process is a bit chaotic?
Then that is the finding. Automating chaos produces faster chaos. Sometimes the honest output of scoping is a tidier process and no automation at all.
How long until the first one is live?
Two to four weeks for a single workflow once scoping is done, depending on how many systems it touches and how clean the inputs are.