Build the
systems your
agency keeps
meaning to build.

I trained as a mechanical engineer. Spent the first part of my career drawing parts in CAD that other people would eventually have to make, sell, and ship. The work was fine. What I noticed, year after year, is that the part almost never failed because of the part. It failed because the team that owned it didn't have the right system around it. The bottleneck was upstream of the geometry — and most of the time nobody was looking there.

So I stopped designing parts and started running the systems that surround them. Operations Manager at a B2B development agency, where the title said process and project management but the work that actually moved the needle was different. Finding the one constraint that, removed, made the next three months obvious. Hiring eight people into eighteen. Watching a side-product go from $1M a year to $12M because we stopped trying to fix everything at once and fixed the right thing. Most of what I know about systems-that-make-money I learned in those two years.

What I came to love is the boring answer. The exciting one — pivot the strategy, build the new product, hire the senior leader — almost never works. The boring one — find the one place the chain is binding, build the smallest thing that unbinds it, refuse to build anything that doesn't — almost always does. Manufacturing engineers have known this for forty years and called it theory of constraints. Most software companies, agencies, and service businesses haven't translated it across the gap yet. That gap is the work I keep coming back to.

It's also where I think my view is a little different from most people doing this work. Most people building "AI automation" right now came from software, sales, or growth. I came from a shop floor — where you ship product under cost, time, and regulatory constraints, and where the systems either work or the line stops. The discipline that gives you is the same discipline you need to make AI useful inside a real team. And the absence of it is exactly why most AI projects right now are theatre. A lot of the field is allergic to the boring answer. I'm allergic to the exciting one when it isn't earned.

That's what Sumaq is. AI and automation systems for B2B agencies and service teams who already know they have a problem on a spreadsheet — too many manual hours, too few qualified replies, reporting that takes three days a month — and want someone who'll find the actual constraint, build the smallest thing that fixes it, and not pretend the rest of the work is interesting. The operations layer that disappears when it works. The kind that doesn't get a press release.

What gets me out of bed is the moment a number actually moves because someone took the boring answer seriously. That, and the fact that we're early in the most interesting decade for this kind of work since the internet shipped. Most of what's possible right now hasn't been built yet, and a lot of what has been built shouldn't have been. Closing that gap, one team at a time, is the work I want to still be doing in ten years.

I write in public on LinkedIn, partly to think and partly because most of what I know I learned from someone else writing in public five years ago. Pay it forward.