Agents
Analysis that explains itself
The Analytics Agent reads across your governed data, works out what moved and why, and returns a clear summary with the numbers and definitions behind every statement.
What it does
Not just the number — the reason behind it
A metric on its own raises more questions than it answers. The Analytics Agent goes past the figure to the explanation. It reads across the governed data products relevant to a question, identifies what changed against the prior period or plan, ranks the drivers behind the change, and writes it up in language a decision-maker can act on — with each claim linked back to the data that supports it.
Reads across sources
Combines governed products from different systems into one coherent view.
Ranks the drivers
Attributes a change to its causes and orders them by contribution.
Writes it plainly
Produces a summary a reader can act on, not a table they must decode.
Inputs it accepts
What you feed it
A question or metric
A specific question, a metric to explain, or a period to review.
Governed data products
The published products it may read across, with their definitions.
Targets and prior periods
Plans, budgets or previous periods to compare the current state against.
Decisions it makes on its own
What it concludes without asking
When the data supports it, the agent draws and states conclusions: what moved, by how much, and which factors drove it. It uses agreed definitions so its figures reconcile with the rest of the business, notes the confidence behind each conclusion, and attaches the supporting query. It distinguishes a real signal from noise, and says when a change is within normal variation rather than manufacturing a story.
States conclusions
Says what changed and what caused it, rather than leaving the reader to infer.
Shows the evidence
Links each claim to the query and definitions behind it for verification.
Separates signal from noise
Flags when a movement is within normal variation instead of over-explaining it.
What escalates to a human
Where it defers to judgement
Low-confidence conclusions
When the data is thin or contradictory, the agent presents what it found and its uncertainty rather than asserting a driver it cannot support.
Decisions, not analysis
It explains what happened and why; the call on what to do about it stays with a person, with the analysis laid out to inform it.
Systems it connects to
Where it reads and writes
Data products
Reads across governed products under the same access policies as any client.
Semantic model
Uses shared definitions so its figures reconcile with everyone else's.
Reporting channels
Delivers summaries into the documents, decks and channels leaders already read.
A worked example
Why margin fell this month
Asked why gross margin fell two points this month, the agent reads across sales, cost and returns products. It finds three drivers and ranks them: a shift in product mix toward lower-margin lines accounts for most of the fall, a supplier price rise adds to it, and a small rise in returns explains the rest. It writes a short summary stating each driver with its contribution, attaches the supporting queries, and notes that the returns effect is close to normal variation — leaving the pricing decision to the team.
Questions
Frequently asked
- Can we trust its explanations?
- Each conclusion carries the query and definitions behind it and a note of confidence, so you can check the reasoning rather than take the narrative on faith.
- Does it make decisions?
- No. It explains what happened and why and lays out the evidence. The decision on what to do stays with your team, better informed.
- How does it avoid inconsistent numbers?
- It uses the same governed definitions as the rest of the platform, so its figures reconcile with your dashboards and reports instead of introducing a fourth version of the truth.
Explain a real movement in your numbers
Point the agent at a metric that surprised you and we will show you the explanation it produces and the evidence behind it.