Water sprinklers to AI

When I was in college, I spent a summer working in an ornamental lawn sprinkler company. My job wasn't to design anything. It was to stand on the line and do the thing. But I couldn't help myself. I started watching the flow. Where things bottlenecked. Where people waited. Where the same part got touched three times by three different hands before it moved forward.

I reorganized the assembly line. Doubled their capacity. I was 19 years old and I didn't know what "process optimization" meant. I just knew it was broken and I could fix it.

That instinct never went away.

From 2012 to 2017, I worked for an agency out of San Francisco. I built all of their automations. Their processes. Their systems. Their SOPs. Everything that could run without a human touching it, I made run without a human touching it. At the time, I was working mostly in Trello and custom scripts. That was the toolbox. Not much, but enough to turn a manual agency into something that ran more efficiently.

When I started Surefoot, I took it further. Built out every automation I could with Zapier, Make, Airtable, and Asana. Every manual process that took someone an hour a week got turned into something that took zero hours a week. A couple of years ago, we migrated from Asana/Airtable to ClickUp and kept going. I wasn't trying to be clever. I was buying back our time.

And now?

I've been using AI in my agency since 2021, and I've stayed at the forefront of how it's evolving. Now I've built AI agents that do what the automations couldn’t. An agent that monitors a/b test performance. An agent that pulls financials and forecasts cash flow. An agent that handles design briefs. An Agent that does exploratory reporting from BigQuery data. An agent that builds quick mockups to help convey ideas to my designers. Agents that report to me. Agents that report to various team members. Even agents that report to each other.

The architecture: three layers plus governance.

Layer one: inputs. Anything with an API or a login. Your ad platforms. Your store. Your ESP. Your analytics. Your support tickets. Every tool you're already paying for has a way to pipe its data into a system that can actually use it. MCP connectors make this easier than it's ever been. Most of what you use is already connector-ready.

Layer two: memory and data. Memory is more than markdown files. That's where everyone starts, and if that's where you stop, you're going to hit a wall. The agents I run use AI-enabled data tables that improve both performance and recall as they operate. Markdown files teach an agent how you think. Data tables give it persistent, structured knowledge that compounds over time. You need both.

There's also governance. Guardrails that control what an agent can and can't do. Which tools it can touch. Which actions need your approval. You don't just set agents loose. You define the boundaries, and they operate inside them.

Layer three: agents. The workers. A project manager that drops a brief into Slack every Monday. A finance agent that flags margin drift before it becomes a problem. A Google Ads specialist that monitors campaigns while you sleep. A legal reviewer that does the first pass on every contract so your attorney only touches the final 5%.

Each agent reads from your inputs, thinks through your memory and data tables, respects your governance rules, and does specific work for specific people on your team.

What you actually get.

Without the memory and guardrails, you have a brilliant intern who needs their hand held while they build context. When you add the files, the data tables, and the governance rules, you get a smart staff person who operates consistently and at speed, 24/7.

That's the difference. Not "AI helps a little." You can hand over real work to it and trust it'll get done the way it needs to be.

Why this matters right now.

Those reports you’re wishing you could have weekly instead of quarterly, done. That analysis that gets too confusing to build because it’s pulling from several data sources, done. The campaign you just launched and you want a holistic view on throughout the day, done. 

The iteration and learning loop is fast and not only takes work off you and your team’s plate, but it frees you up to do the really important stuff. The high-leverage work you struggle to find time for. 

I've been doing this for three decades, from a factory floor in college through Zapier/ Make/n8n, to Airtable/Asana/Clickup/Trello, all the way to AI agents that talk to each other. I've been running AI in production since 2021. I know what breaks. I know what to build first. I know which tasks are worth offloading and have a greater ROI.

If you want your brand wired up, book a call. Whether you just need your first agent or a full fleet, I'll get you from zero to running.

Book a call →

The companies that win the next five years are the ones that wire this up correctly right now. Don't be the brand waiting until Monday.

Looking forward,

Brian