Kairo
Product / Kai
The AI analyst

Kai has read your campaign data. That is the whole difference.

Ask a general chatbot why your LinkedIn CPCs went up and you will get a confident, generic answer built from blog posts. Ask Kai and you get an answer built from your own delivery data — with the query it ran shown underneath, so you can check it.

Kai · ABM analyst · Enterprise DACH
Which accounts on the DACH list moved up a stage this week, and what did they see?
Kai Nine accounts moved up. Seven went Aware → Engaged, two went Engaged → In market.

The two that reached In market — Ravensbourne Group and Kestrel Analytics — both had a second buying-committee member engage for the first time. Ravensbourne added a Head of Ops after eleven touches on the migration-cost ad; Kestrel added a VP Engineering.

Both saw the same creative first: SI-04 · “What a bad migration costs”. That ad is carrying 61% of stage progressions on this list at a £38 cost per account reached, against a £71 list average.
Push both to Salesforce and draft the note for the AEs.
Kai Done — both accounts synced with the stage change, the score history and the ads each contact saw. I have drafted a three-line handover for each AE; open the queue to review before it sends.
What to ask it

Four jobs, in roughly the order people use them

Campaign intelligence

“Which creative moved the most accounts from Aware to Engaged last month, and what did each movement cost?”

Account briefings

“Everything Ravensbourne Group has seen from us since June, in a paragraph I can read before a call at three.”

Attribution

“What opened in Q2 at accounts we had engaged twice before the opportunity was created?”

Reporting

“Write the quarterly summary for the DACH programme: coverage, movement, spend, and the two things I should change.”

How it stays honest

Four constraints we built in on purpose

  • It shows its query. Every answer expands to the filter and date range it ran. If the answer looks wrong, you can see why in one click.
  • It refuses small samples. Below the confidence threshold Kai says the sample is too small rather than producing a number that will get quoted in a deck.
  • It cannot spend money. Kai proposes budget and audience changes; a human approves them. There is no autopilot that quietly reallocates your quarter.
  • It stays in scope. Kai answers from your workspace data. It is not a general assistant, and it will tell you when a question is outside what it can see.
Kai · query shown for the answer above
list
Enterprise DACH (412 accounts)
window
2026-08-31 → 2026-09-06
filter
stage_change != null
join
creative_exposure (last 14d)
rows
9 accounts, 34 exposures
confidence
sufficient (n ≥ 8)
For your own agents

Kai is one client of the MCP server. You can be another.

On Pro and Agency, Kairo exposes an MCP server over the same account graph Kai queries. Point Claude, an internal agent or your own tooling at it and ask the same questions from wherever you work.

Ask Kai a question about your last quarter.

Connect an ad account, wait for the backfill, and put the first question to it. Most people start with “what did we waste money on”.