AI training that ends with working agents.

An onsite workshop, then four weeks building real agents with your team. Learn from AI founders directly.

Michael Ducker on stage at an AI training session, talking through the agenda slide behind him: the hook, AI basics, coworking with AI, and always-on agents
“We’d each been working with AI on our own. Valet helped us turn that into something collaborative, giving us real ways to share what we were building across the team.”
“We were trying to keep up with AI innovation on our own, but we didn’t know where to start. In a single day, Valet got everyone fluent in the tools and then built with us on real problems inside our organization. We walked out with agents and automated workflows actually running against one of our biggest bottlenecks. It moved AI from something we talked about to something we use.”

A workshop, then tech-therapy.

Most AI training stops at the demo. We stop to understand.

The first section covers how these tools actually work. Then we listen and understand what makes your job hard. We leave armed with the insights to build solutions that make your day easier.

The day

How it works

The model, the harness, and how to leverage it.

Research

We dig into the problems your team faces getting work done.

Next 4 weeks

We build agents with you that make your job easier.

How to actually use Claude and ChatGPT.

We start with what is in the box: how the model predicts, where it breaks, and how to tell a good answer from a lucky one.

Then the tools your team already pays for. Claude Code for shipping work. ChatGPT and Claude for daily cowork. And how to connect them to your Slack, Drive, and CRM so they see real context.

What the model sees

"The cat sat on the ___"

mat
46%
couch
17%
floor
12%
windowsill
8%
keyboard
6%
roof
4%
other
7%

It does not know the answer. It holds a distribution and samples from it. Same prompt, different day, different word. Once your team sees this, the magic turns into something they can reason about.

Learn the Blueprint Method.

Every agent comes down to four concepts: what is its goal, what skills can it use, what triggers it, and how does it communicate with you?

Write those and you have a blueprint. We'll learn how to build these together.

The blueprint

Outcome

Every Monday at 8am, a pipeline digest: deals moved, deals stalled past 14 days, three follow-ups to send.

Skills

Reads the CRM, ranks deals by stall risk, and drafts follow-up emails in your voice.

Trigger

Monday 07:30, and on demand when someone mentions the bot.

Channel

Posts to #sales-pipeline. Never sends a DM.

The engagement

We build real agents with you.

Once you've taken our workshop, you're in our community. We'll support you over the next month to turn your ideas into production agents your business can rely on. We'll build examples for you, work with your team to deploy them, and help you on your own exploration.

A team working together around a table on laptops, building agents alongside a whiteboard of plans

Meet your instructors.

You learn from the founders building Valet, engineers who have scaled infrastructure at Chime, Mercury, Stripe, and Heroku.

Michael Ducker

Michael Ducker

Co-founder, Valet

Previously VP Product at Mercury. Invented the Credit Builder card category at Chime and led their technical product function, driving company velocity at scale.

@miradu
Jamu Kakar

Jamu Kakar

Co-founder, Valet

Staff Engineer at Stripe on payment infrastructure, ran application platforms at Heroku, and built Chime's money movement platform: systems processing billions in transactions.

@jkakar

Bring us your team's real work.

Book 30 minutes. We'll answer any questions and get you scheduled. Limited spots available, because you know, we run a startup too.

Schedule a call

Or email ducker@valet.dev