The most useful thing we sell is a no.
That sounds like a line, so let me make it concrete. Here are three projects we were asked to build, were happy to build, and did not build — and what the math said instead.
One: the quarterly board packet
A regional developer with about thirty employees asked us to automate the board packet. Every quarter, their controller pulled numbers from three systems, reconciled them, built twelve slides, and sent it to the CEO for a round of edits. She hated it. He hated reading it. It felt like exactly the kind of thing a machine should be doing.
So we priced it. Sixteen hours of her time per quarter, at a loaded rate of about $55. Add four hours of the CEO's time at a much higher rate. Call it $6,000 a year, generously.
The build — connecting three systems with genuinely messy data models, handling the reconciliation logic, generating a formatted deliverable that a board would accept — was going to run north of $30,000. Even amortized, that's a five-year payback on a document format that would probably change twice in that window.
We told them not to build it. We also told them the real problem wasn't the packet; it was that the three systems didn't agree with each other, which was a data problem worth solving on its own terms someday. They fixed the slide template instead and moved on. Cost them nothing.
Two: the AI receptionist that nobody would have called
A boutique brokerage wanted a voice agent to answer inbound calls after hours. Very fashionable. Very buildable.
We asked how many after-hours calls they got. They didn't know. Nobody ever knows — that's normal, and it's the first thing to go check. So they pulled the phone logs.
Eleven calls in ninety days. Nine were vendors. One was a wrong number. One was a real prospect, who also emailed, and who they closed anyway.
There was no problem. There was a fear of a problem, which is a completely different thing and cannot be automated away. The correct build was a $0 build: a better voicemail greeting that told people to text.
What's worth noticing here is that a build shop with a quota would have taken that project. It was funded, it was scoped, the client was enthusiastic. It would have been a perfectly successful engagement by every metric except the only one that matters — whether it made the business better.
Three: the workflow that was about to disappear
A property management company asked us to automate their utility bill allocation across units. Painful process, real hours, defensible number. The math actually cleared, which made this one harder.
But in discovery we found out they were nine months from migrating to a new property management platform — one that, we confirmed, did utility allocation natively. They knew this. They'd just stopped connecting it to the thing that was annoying them every week, because the annoyance was loud and the migration was quiet and far away.
So the honest answer was: don't build this, wait nine months, and if the new platform's allocation module is as bad as most vendors' allocation modules are, call us then. They waited. The module was fine. We never got a dollar from them and we get referrals from them constantly.
Why we do this
Not out of virtue. Out of arithmetic.
We guarantee a minimum 2x annual return on what a client spends with us. That guarantee is not primarily a marketing device — it's a filter that points inward. It means every project we take, we have to be able to defend on a spreadsheet before we ever open an editor. It means the fastest way for us to lose money is to build something that shouldn't exist. So we've built a discovery process whose main job is to kill bad projects early, and we've accepted that this will cost us revenue in the short run.
It also does something to the relationship that a discount never could. When we say "this one is worth building," the client believes us, because they've heard us say the opposite.
What this means for you, whether or not you ever hire anyone
Before you buy any automation — from us, from a SaaS vendor, from your nephew who's good with computers — make somebody argue the other side.
Ask the person selling it: under what circumstances would you tell me not to do this? If they can't answer, or if the answer is a soft-focus "well, if you're not ready culturally," you're not talking to an advisor. You're talking to a salesperson with a demo.
Then check the three killers yourself. Is the volume real, or is it a fear of volume? Go count. Is the process about to change anyway — new system, new regulation, new hire? Wait. And is the payback inside two years on honest numbers? If not, walk.
There is enormous, unglamorous, provable money sitting in the workflows your team does every single week. You do not need to reach for the marginal projects. You do not need to automate everything.
You need to automate the things that free your people to do work that a machine can never do — and you need somebody in the room who will tell you, out loud, which ones those aren't.