Next Webinar →Build a self-serve agent, live | 20 MAY
WebinarThe Self-Serve Agent Series · Part 1 of 3

Build a self-serve agent, live

Oliver Hughes builds a live self-serve AI agent from scratch, explaining key architecture trade-offs and fast iteration you can apply with any tools.

Join 4,000+ data pros who already learn with us
AccentureBumbleCleoDoceboInsurelloJustParkKitLeadIQMoneyboxMUBIWriter
AccentureBumbleCleoDoceboInsurelloJustParkKitLeadIQMoneyboxMUBIWriter
AccentureBumbleCleoDoceboInsurelloJustParkKitLeadIQMoneyboxMUBIWriter
About this session

What to expect, and why it’s worth your time.

Self-serve AI agents are no longer optional, but most teams are still stuck debating architecture instead of shipping.

In this session, Ollie builds a working self-serve agent from scratch, live, and walks through the trade-offs at every fork in the road. You'll see the architectural choices that matter and how fast iteration is the key to success.

We’ll be using Count to build, but the principles are tool-agnostic. If you walk away and build your agent somewhere else, we hope we have helped.

This is Part 1 of a three-part series. It works as a standalone session if you only attend one - but the series gets more valuable as you go.

Frequently asked questions

  • What if I can’t make the live session? Sign up anyway - we’ll send the recording within 24 hours, plus all the resources.
  • Do I need to be a Count customer? No. We use Count to demonstrate, but the session is tool-agnostic.
  • Will this be a Count product demo? No. We’ll be building inside Count because we have to build inside something - but the principles apply whether you build it yourself or buy a different tool.
  • Can I attend just one part of the series? Yes. Each session stands alone. But signing up to one gets you all three on the calendar automatically.
What you’ll leave with

Concrete takeaways, not a think piece.

01
The three architectural patterns
Build vs. buy, where context lives, how data flows in - the trade-offs that matter at the start.
02
Scoping your first agent
How to define an agent scope that has a real chance of working on Day 1.
03
Three ways to build the data model
Raw tables, a semantic layer, or something in between - with the trade-offs of each approach.
04
Watching it fail in real time
The iteration loop is the product. We fix failures live and show you how to speed up the cycle.
05
The infrastructure questions that actually matter
Semantic layer? MCP? What context goes where? The questions you need to answer before you start.
Your host & who this is for

Led by someone who’s actually done the work.

Oliver Hughes

Oliver Hughes

CEO · Count

Ollie is the CEO and Co-founder of Count. He started Count to help data teams spend less time building dashboards and more time driving decisions that actually change the business.

Who this is for

  • Heads of Data and Analytics leaders
    Evaluating whether to build or buy a self-serve agent and want a grounded view of what it actually involves.
  • Data engineers and analytics engineers
    About to start their first agent project and want to see the full picture before diving in.
  • CDOs and data leaders
    Who want a working mental model before approving the spend.
  • Anyone who’s been told ‘we should have an AI agent’
    And isn’t sure where to start.

Secure your spot.
It’s free.

Oliver Hughes builds a live self-serve AI agent from scratch, explaining key architecture trade-offs and fast iteration you can apply with any tools.

Registration open
20 May 2026
3:00 pm BST / 10:00 AM EDT
Oliver Hughes
Oliver Hughes
CEO