At a twelve-person title agency, there is a workbook called MASTER_v4_USE_THIS_ONE.xlsx.
It lives on one person's desktop. It has eleven tabs, four of which are hidden. It contains a macro that somebody's brother-in-law wrote in 2021 and that no one has been able to modify since. It pulls a list of files, checks them against the recording status, flags the ones that are aging, and produces the Monday morning list that the whole office actually runs on.
It is not in the software you pay for. It is not in the process documentation. It is not on any org chart. And if the woman who maintains it took a two-week vacation, the agency would feel it by Wednesday.
Every company we walk into has one of these. Usually several. And they are, without exaggeration, the most valuable strategic asset in the building — not because they're good, but because of what they tell you.
Shadow automation is a map of your pain
Nobody builds a duct-taped spreadsheet for fun. Every one of them exists because a person hit a wall between what the business needed and what the systems could do, and rather than escalating, they solved it themselves on a Tuesday night.
Which means each one is a marker. Somebody, at some point, was in enough pain to build a workaround on their own time. That is a signal you paid nothing for and that no consultant could have generated, and most owners have never once gone looking for it.
So go looking. Ask a simple, non-threatening question at your next team meeting: what have you built for yourself to make your job work? Not "what's broken" — people are diplomatic about what's broken. Ask what they've built. The spreadsheets, the Zapier chain, the folder of email templates, the checklist in their notes app, the text macro, the Access database from 2014 that finance still depends on.
Write them all down. That list is your automation roadmap, and it cost you fifteen minutes.
Why it beats any roadmap you'd build top-down
Three reasons, and they're all about risk.
The workflow is validated. If someone built a workaround and kept using it for two years, the pain is real and durable. You are not guessing at demand. You are looking at revealed preference.
The logic already exists. The hardest part of automating a process is usually extracting the rules — the exceptions, the edge cases, the "well, unless it's a cash deal, then we skip that step." In a shadow spreadsheet, that logic has already been written down, painfully, by someone who understood it. It's ugly, but it's there. You're not designing a process. You're upgrading one.
And the adoption problem is pre-solved. The single biggest reason automation fails at small companies is that nobody uses it. But if you rebuild the thing your team already built for themselves — properly, so it doesn't break, so it doesn't live on one laptop, so it doesn't die when she goes to Cabo — you are not asking anyone to change their behavior. You're removing the thing they hate about the behavior they already have.
That is a much, much easier sale than a new platform.
The conversation to have carefully
There's a delicate thing here and it's worth naming.
Some people will be nervous to show you their spreadsheet. They may feel that it exposes them — that they've been doing something unsanctioned, or that if you see how the sausage is made you'll take it away from them, or worse, that automating it means automating them.
Get ahead of that, out loud. The message is simple and it has the advantage of being true: you built a tool to survive a bad process, and instead of asking you to keep maintaining it forever, we're going to build the real version so you never have to think about it again. Nobody is losing a job because they were resourceful. That would be an extraordinary thing for a company to punish.
And then follow through, which means being explicit about what she does with the reclaimed time. In the title agency's case, the woman with the workbook is the one with eleven years of underwriting instinct — the one who can look at a chain of title and feel that something is off before she can articulate why. That instinct is worth vastly more than her ability to maintain a macro. She should be curing title issues, training the newer closers, and being the person the agents call when it's complicated.
She should not be spending Sunday night patching a spreadsheet because a column header changed.
Start here, not with a strategy
There's a version of the AI conversation that begins with a strategy offsite, a maturity model, and a two-year roadmap. It produces a beautiful deck and, in a twelve-person company, roughly nothing.
Here's the alternative. Find the shadow tools. Price the workflows underneath them. Rebuild the top one or two properly, with the person who built them in the room. Ship in weeks, not quarters. Measure the hours back.
The strategy will emerge from that, and it'll be a better strategy than the offsite would have produced, because it'll be made of things that are actually true about how your company works.
Your team has already told you where the problems are. They just told you in Excel.