Invite-only · November 2026
Atlan's Analytics Agent Accelerator: The Builder's Cohort
Build and improve the analytics agent your business believes. We just open-sourced the playbook our forward-deployed engineers use. Before anyone else runs it at scale, a small group of teams builds with us: one analytics agent on your own data, and the loop that makes it better every time it runs.
- Starts
- Nov 9
- Applications
- Open
- You build
- 1 agent + its improvement loop
- Open to
- Any team, Atlan customer or not
Admit one team3-minute application
ApplyFrom #data-panic to #data-calm.
The board meeting is in 20 minutes and nobody trusts the number. See how a governed analytics agent answers it, and how every miss becomes the context that fixes the next one.
Launch reel, 1 minute
Build it once. Then make it better every week.
Most teams stop at the demo. You'll leave with the agent and the habit that keeps it accurate after we're gone.
Board in 20 min. What's our Q3 sales pipeline?
Pipeline agent $18.4M"Confidence: very high." Finance has a different number, and nobody owns the metric.
- Govern the metricGoverned metrics
Q3 pipeline, defined once by RevOps. Stage 1 never counts.
- Certify the queryCertified queries
One trusted query, versioned in the Atlan Trident.
- Score a baselineEvals
Run the golden set: the questions your business actually asks.
- Then turn every miss into contextLearning loops
Miss, add the definition, rerun. Keeps running after the cohort.
Board in 20 min. What's our Q3 sales pipeline?
Pipeline agent $11.2MCertified. The number Finance runs on.
We know the science, and how to run a cohort that ships.
Our forward-deployed engineers spent a year building analytics agents with enterprise data and AI teams, hands-on with their own data. We know what moves accuracy, in what order, and how to run a cohort so every team leaves with a working agent and a score to prove it.
What moves an agent's accuracy
Each bar is the bare schema plus one kind of context, on the same 107 real questions our analysts asked, graded by code.- Bare schema31.8%
- Usage patterns43.9%
- AI-generated certified metrics and queries83.2%
Context moved the score more than the model did. No person wrote the context behind these results: our systems built every metric, past query and term from the query log, the catalog and existing documentation. Your numbers will be your own; expect the same order.
That was our last accelerator, Context Agents: a two-week sprint on real data, with 50+ teams in a single cohort. This one is deliberately smaller.
Source: atlan.com, Fear, trust and controlThe open-source playbook you'll work from, with an evidence register behind every number in it.
Read the playbookHear from our design partners on building analytics agents with Atlan
“It was a perfect harmony… between the four different types of teams… all working together and delivering this, I think in two weeks or so.”
“The fact that Atlan can generate so much on the fly, off the data that you give it… this is the fastest we've ever been.”
“Hats off. I'm very impressed by how well this is working: zero noise, zero delivery issues… thank you for doing this.”
A small room, on purpose.
You build with the team behind it, before anyone else does.
- The people who wrote the playbook
A small cohort means every team gets real time with the engineers who built it, not a webinar seat.
- First access
Build on the Atlan Trident before it opens up more widely.
- A say in what we build next
What trips you up goes straight to the product team building it.
Builders who want to get their hands dirty.
- You like to build and learnTitles don't matter. If you'd rather figure it out by building, you're who we want in the room.
- Serious about self‑serve analyticsThe business answering its own questions, with numbers it trusts, without a ticket to the data team.
- An idea of where to startA domain, or a few questions you'd like an agent to answer. It doesn't need to be polished.
- Atlan customer or notYou don't need the Atlan catalog to join.
Apply to the builders' cohort.
What you'll walk away with
A working analytics agent on one domain of your data
A baseline accuracy score on your own questions
The loop to keep improving it after the cohort ends
Early access to Atlan Trident
Trusted by the world's leading AI teams
Tell us what you're building
Questions
Do we need to be an Atlan customer?
No. The cohort builds on the Atlan Trident, which doesn't depend on the Atlan catalog.
What does it cost?
Your time and ours. You bring the domain and the questions; we bring the engineers who wrote the playbook.
How is this different from reading the playbook?
The playbook tells you what to build. Here you build it on your own data, then improve it, with the people who wrote it checking your work.
When does it run?
November 2026, starting November 9. We share the full schedule with the teams we invite.
What happens after I apply?
We read every request and invite a small group of teams. If you're in, we'll reach out with next steps.