Most AI conversations in SMEs go wrong in the same place. Someone brings in a strategy, the strategy has a 12-month horizon, and the business is being asked to spend money now on a return that shows up next financial year. Leadership nods along and quietly decides not to.
The engagements that actually get traction start somewhere else. They start with one workflow that is obviously broken, fix that, and let the number do the talking. Here are three from the last year.
1. The site survey that took all afternoon
A construction services business sent people out to survey sites. Drive there, walk the site, take photos, take notes. Then drive back and spend the rest of the day turning the photos and notes into a report the client could act on. Four hours per site, and the report-writing half was the part everyone hated.
We built them a tool that takes the photos and the field notes and produces the first draft of the report. The surveyor checks it, fixes anything the tool got wrong, and sends it. The report-writing half of the job went from roughly 3 hours to under 1.
Four hours became one. Same people, same vans, same clients. The team can now do 2 to 3 times the number of surveys in a week, which in a business where surveys are the front end of every job means 2 to 3 times the pipeline.
The tool took under 4 weeks to build and cost less than a junior surveyor's salary for a quarter.
2. The 60-minute task that 10 people did every day
A professional services firm had a watermarking and document-prep step that every person on the team did before sending work out. One hour a day, per person, ten people. Nobody questioned it because it had always been done that way and it was not anyone's job to fix it.
Sixty minutes became 3. The tool handles the formatting, applies the watermarks, checks the output, and hands back a file ready to send.
The arithmetic on that one is brutal in the best way. Ten people, 57 minutes saved, every working day. Roughly 1,300 hours a year handed back to the business. At that firm's billable rate, that is about $250,000 of capacity they now have and did not have before. They saw it in the first month because the tool was live inside a fortnight.
Nobody was made redundant. The team just stopped doing the thing that made them want to quit and started doing the thing the clients pay for.
3. The admin role that was about to walk out the door
A mining company had a senior administrator leaving. She had been there a long time and a frightening amount of how the operation actually ran lived in her head and her inbox. Nobody had written it down because she had always just done it.
We had a few weeks. We sat with her and mapped every process she touched. What comes in, what she does with it, who she sends it to, what breaks if it does not happen. Then we documented it properly, flagged the parts that were candidates for automation, and built the plan to keep the operation running after her last day.
The immediate value was zero disruption through a role exit that could easily have cost them months. The longer-term value came out of the mapping itself. Once you can see every process on a page, you can see which ones are double-handled, which ones exist because of a system that was replaced 3 years ago, and which ones a tool could do in seconds. That map became the optimisation roadmap.
What these three have in common
They do not look alike on the surface. Field operations, document prep, admin knowledge capture. Different industries, different problems.
Underneath they are the same shape.
Each one was a task that was done often, by people who were paid to do something more valuable, and that nobody had questioned in years. Each one had an obvious before-and-after number: hours, reports, dollars. And each one could be fixed in weeks with something specific to that business, rather than in months with a platform.
That is the pattern. Frequent, tolerated, measurable, and fixable with a tool rather than a transformation.
How to find yours
Ask three questions around your own leadership table this week.
What do our most expensive people spend time on that does not require them? If your surveyors are writing reports and your consultants are formatting documents, you have found it.
What process only works because one person knows how it works? If the answer makes anyone uncomfortable, that is the one to map first.
What do we do every day that nobody has questioned since we started? The hour-a-day watermarking job sat there for years because it was small enough to tolerate and nobody owned fixing it.
You will have a shortlist in 20 minutes. Pick the one with the clearest number and start there. Do that one well and the strategy conversation gets much easier, because you are no longer asking anyone to believe in AI. You are showing them last month's numbers.
That is what Howll Advisory does. If you want help finding yours, the first conversation is 30 minutes and there is no deck.