Not subscribed? Sign up to get it in your inbox every week.

Deuce’s Slack pinged on a random Tuesday afternoon.

"Just found a great nugget in here for the sales call next week."

The nugget came from their deep research bot, an AI agent that continuously scans public brand strategy signals, scores them against Recess's ideal customer profile, and flags high-intent opportunities. The bot had picked up that a personal care brand was pivoting away from exclusively digital marketing toward "trial-driving activations."

Activations is Recess's core business. The bot scored the lead, ranked it as a high-priority target, and drafted a pitch. No human researcher involved. No one Googling brand strategy shifts and reading trade press for an hour.

Nine months earlier, that opportunity doesn't exist. It's buried in a trade publication nobody on the sales team has time to read.

This is how a brand activation platform started deploying AI agents that find, score, and prep client pitches while the team sleeps. But the interesting part is how it all came about.

Recess is the marketplace connecting major CPG brands with experiential sampling opportunities. A brand wants its new energy drink in the hands of gym-goers? Recess finds the venues, manages the logistics, runs the campaign, and measures whether those free samples turned into purchases. One campaign drove a 338% increase in purchase intent after sampling. Another converted 1 in 4 samplers to purchasers.

Two sides to the market: brands who want to sample (demand), and venues and events where sampling happens (supply). The constraint is finding enough of the right opportunities on the supply side to sell against. That means knowing about every event, meetup, club, and concert city by city

"You're basically going to websites and Googling 'events in this area' and trying to figure it out," Deuce said. "That is a huge pain point."

Recess had a head start most companies don't. Before any AI projects started, they'd already piped their core operational data (MongoDB, HubSpot CRM, QuickBooks) into BigQuery. Not every data source, but the bones were there: sales data, customer data, and financial data in one queryable place.

When they partnered with 3rdBrain, we embedded an automation specialist named Leo directly into their team after a kickoff strategy meeting.

Deuce expected something different. "At first I thought I would be working with you guys more like a consultancy or a dev shop and less of a staffing agency," he said. "That was something I didn't quite understand at the beginning."

The first project was deep research: building AI agents that could continuously scan public information about brands from earnings calls, trade press, and other announcements or interviews online to surface insights relevant to Recess's sales team. That required that existing data layer to work against so the tool could score and rank opportunities against Recess's actual successes.

That deep research project is still running nine months later. "...probably one of the most useful sources of information and data that we have," Deuce said.

With the research layer producing results, Leo expanded into marketing and sales automations. He took over the Clay workflows, built and managed n8n automations, and started connecting more data sources into the BigQuery layer. Call transcripts went in. Additional marketing data. The data layer still isn't perfect but it was functional. "It's got the core pieces that I need, pretty much."

Then came the event aggregator. This was the twenty-hours-a-week pain point and an AI agent was not the solution.

Leo built an n8n workflow for it. The team submits a form: "We need events in Los Angeles for fitness studios." The workflow identifies which event aggregator websites cover that category. It grabs their sitemaps, pulls listings matching the target geography, sends them through an LLM to extract the messy unstructured data (contacts, dates, event details) into clean records, deduplicates against the master database, and stores everything. The next query for LA only pulls in new events.

"Something that took us 20 hours now takes about an hour," Deuce said.

Automations and lowcode were the start, but things got interesting after Deuce spent a weekend hacking through his tool backlog with Claude Code.

He started building internal tools and shipping working prototypes in a weekend.

The tools were buggy but they worked.

Deuce and Leo fell into a rhythm: Deuce builds the prototype and scaffolding, then hands it off to Leo for the polish and reliability work that turns a hack into something the team can rely on. "All the effort comes in the last 20%," Deuce said. "I don't have the time to ever get it working perfect."

They scoped a KPI dashboard on a Wednesday. Leo had a working prototype Thursday.

Leo built a programmatic SEO tool to mass-generate hundreds of landing pages, dynamically placing photos and social media posts for every market and partner type in Recess's repertoire.

"You can do anything now," Deuce said.

The engineering team watched this happen for weeks. The non technical co-founder shipping weekend prototypes. The lowcode automation guy turning out tools at a pace that made sprint planning look like a speedbump.

The AI bug caught them too.

"People are like, all right, this is legit," Deuce said.

Deuce and Leo built "Recess Brain” an internal Claude plugin that connects the company's tech stack decisions, SOPs, metrics, playbooks, skills, and agent configurations. Any team member can query it. Any employee can build an agent, deploy it, and share it with the rest of the org.

"It's basically a unified data layer but for skills," Deuce explained. "Everything about the company tech stack and decisions and skills and agents and playbooks and SOPs and metrics."

This quarter they ran their first company-wide AI agent training. Everyone built their first agent.

The Props channel lit up:

"Major props to Leo. We worked together this morning to get the Recess Brain downloaded and it is amazing. Going to be such an unlock for all of us. I can't wait for our training next week. Raise the bar, stay curious."

Deuce's message was clear: "You don't need to ask anymore, you just need to do it. Don't ask me to do anything anymore, just do it yourself."

Leo went from building n8n workflows to building full web apps. In nine months, he 5x'd his capabilities because the pace of work demanded it. 

"He's leveling up the team, for sure," Deuce said.

They tried to automate LinkedIn content creation and over-engineered it. Too many prompts, too much context data, too many personas. They tried to build a system that could post as multiple people in different voices instead of doing one voice well.

"We clobbered it down with too many prompts and fed it too much data," Deuce said. "We tried to make it create too many different personas instead of making it just good and simple."

Anyone who's tried to build AI automations has hit this exact wall. The instinct is to make the system do everything. The reality is that simpler is almost always better: one input, one task, one output.

Most companies start with the tool. Which AI platform should we use? Which vendor should we evaluate? How do we build an AI strategy?

Recess never answered those questions. Here's what happened instead:

The data layer came first. BigQuery was connected to their core systems before any AI projects started. Every major AI workflow they've shipped, from the deep research bot to the event aggregator to the deal-scoring algorithm, works with it. Without that data layer, each project starts with a three-month migration and cleanup instead of a working prototype by Thursday.

They embedded a person, not a platform. Leo didn't arrive with just an implementation roadmap or a deck of recommended tools. He arrived with curiosity, learned the company from inside the team, and started building. The first project generated a visible win which bought credibility for everything that came after.

Leadership made AI impossible to ignore. When Deuce started shipping weekend prototypes of features on the quarterly roadmap. It was a COO who got excited about Claude Code and couldn't stop building. But the effect on the engineering team was more powerful than any training mandate: the non-technical co-founder was outpacing them, and the only rational response was to start using the same tools.

AI adoption spread organically. Leo led the first agent training. People built their first agents. The Recess Brain gave everyone access to the institutional knowledge and tools they needed. The culture shifted from "AI is something Deuce and Leo do" to "AI is how we work here."

What happened fast and a little messy: clean enough data, one hungry person embedded in the team, a founder who couldn't help himself, and an org that caught the bug by watching it happen.

The result is new tools in record time, hundreds of hours saved, and an AI-infected team.

Adoption doesn’t start with a seminar, it starts with competition.

3rdBrain embeds AI native talent directly into operator-led teams. One ravenous autodidact who grows with your team and builds until the wins compound. 3rdbrain.co

Searching for your next role? Fill this form out, and we’ll intro you to the best companies in the world

Hiring? just respond to this email! We’ve got dozens of vetted operators standing by.