Small Business Automation: Fix the Process Before You Buy the Software

I'm in a discovery meeting with a construction owner a few weeks back and he tells me he spent 12 months hunting for the perfect accounting software. He's like, I put about $250,000 into it and got nowhere. Now he's still paying outside accountants to do the job. The software industry took a quarter of a million off him and he’s still none the wiser. 

When I asked what went wrong he didn't blame the product. His read is that most platforms are built for the biggest, simplest version of a business, so those companies slide onto them easily, and his firm isn't that. He'd shopped the whole market and concluded nothing fit. And he called universal adoption of any tool "near impossible," because every time staff turn over, the habits revert.

I used to think his story was an outlier. I've heard three more versions of it since the start of summer.

Everyone in the room has a scar

One came from a surgeon. His group's new records system doesn't recognise the terminology of his specialty when he dictates, and the cursor doesn't register where it sits, so notes land in the wrong part of the chart. The old patient records never merged either, so when an existing patient walks in, staff open the old system or wait on printed paper. The vendor sent three people onsite who, in his partners' words, fielded questions without helping. His workaround now is dictating into a Dragon license from years ago and pasting the text across by hand.

Another came from a restaurant owner who bought a direct ordering app on the promise of AI customer marketing. The pitch was lovely. A customer who orders dinner most Wednesdays misses one, so the app sends them $3 off the following Wednesday. The app is live, the card fees get clipped on every order, and that push has never once gone out. But the vendor gets paid either way, so who's in a hurry.

The third wasn't even a purchase yet. A hotel owner showed me a proposal from an enterprise data consultancy asking for north of $100,000: a week on site, thirteen weeks to production, and at the end of it an "AI database." I asked what the AI database would do, and he said that's exactly where they got confused, because nobody could explain how it helps or why it would be a good idea. Six figures for something no one in the building can describe!

They bought in the wrong order

I install this stuff for a living. But the software in these stories was mostly fine. Every one of those tools was chosen before anyone mapped the work it had to sit on, which is how a records system ends up in a surgical group without anybody checking whether it can take dictation from a surgeon. As one tech entrepreneur said, “The best process is no process”, and buying before mapping is how you end up automating a broken one. I say the first half of that in every first meeting, and I've watched it land with owners who've been pitched for years, because it's the first time anyone selling them technology has told them to spend less.

So my order is boring, and the boring is what saves the money. Map how the work happens, every step, including the ones only one person knows about. Delete next, because most businesses I walk through are carrying steps that shouldn't exist: reports nobody reads, approvals that approve nothing, numbers typed by hand into two systems that don't talk. Removing those costs nothing, and there's no license fee on it, which is why nobody sells it to you. Automation comes last, pointed only at the steps that survived. The first thing I point it at in most shops is answering a new lead inside five minutes, and I've written up why that number matters if you want it.

The 15 minute test

Before you buy anything, run it against the bar that the restaurant owner set for me. He solved his cost problem with no software at all. Every Monday and Friday he reviews his top five inputs by impact, about 15 minutes a session, and makes pricing decisions on the spot instead of finding them in a month end report. Then he looked at me and asked it straight: why would anything you build be better than that?

Fair question, and I gave him the only answer I'd want to hear in his seat. A tool earns its place doing what the 15 minutes can't. For him that's watching commodity prices around the clock, because beef moved 27% on him once and he heard about it from his supplier instead of ahead of him. His routine can't see that coming and software can. If a tool can't name the thing your manual version will never do, his 15 minutes wins.

One of these a month, free

I sit down with Bay Area owners every week, and once a month I write up the best of it: the pattern I keep seeing and the cheapest fix that holds. One email, no pitch in it. Subscribe to my newsletter and the next one lands in your inbox.

I'm Jay Waters. I sold to big companies at Meta for nearly 12 years, and now I install AI systems in person for small businesses around Marin County and the wider Bay Area.