Canopy is the business context layer for AI agents.

The layer between your data and your AI systems that carries the organisational knowledge raw tables cannot: structural context (the metric tree and the org chart), operational context (RACI, workflows and approvals), behavioural context (the decision traces of what people actually did and what verifiably worked), temporal context (what was true when, plan against actual) and strategic context (the objectives, key results and funded initiatives the business is pursuing). Without it agents hallucinate rules, fragment into inconsistent answers across teams, and drift as the business changes. Canopy delivers all five, governed, at runtime.

No, and the difference is the point. Our own proprietary ML models and a battery of nightly statistical tests nominate candidate drivers with proper rigour, but nothing is marked proven by maths alone. Every task and initiative in KPI Tree declares the metric it intends to move and the expected direction before the result exists, so completed work doubles as a real-world experiment on a driver edge. Outcomes are measured by the same pipeline that calculates your actuals, and only edges that pass are marked verified. That is interventional evidence accumulated from work your team was doing anyway, and no tool that only observes your data can produce it.

Yes, and a particular kind. Most company brains index what your tools already hold, Notion pages, Slack threads and documents, so an AI can search them. Canopy ingests those sources too, but it curates them: every claim is checked against the numbers before it counts. A sales lead posting that twenty contracts closed is recorded as a claim; the warehouse's eighteen is the fact, and the item says so. Everything sits on a causal model that is statistically tested and re-tested nightly, attached to the metric it explains, weighted by who said it, with sensitive items kept with the right people. Aggregation is where it starts, not where it ends.

Yes. Anyone can add context in plain language: a launch date, an absence, a blocker, a decision and its rationale. Agents factor it in wherever the related metrics appear. Slack channels are read one at a time, and every item keeps where it came from. Every item records who said it, its weight follows your org chart, sensitive items can be restricted to a metric's owners and admins, and everything ingested is reviewed on the way in, so unreliable or contradictory claims are held back rather than served to agents.

No. It sits above it. Keep dbt, Looker or Snowflake Semantic Views as the source of calculation truth; Canopy adds the causal, ownership and outcome context they do not model.

For one-off numbers, do. But a direct connection re-derives the business on every question, cannot know what drives what with statistical confidence, who is accountable, whether the last action worked, or which objective and funded initiative a metric serves, and it burns roughly ten warehouse queries, and an order of magnitude more tokens, where Canopy serves one precomputed answer.

Anything that speaks MCP: Claude, ChatGPT, Gemini, Copilot, Databricks Genie, Snowflake Cortex, Cursor, VS Code, Windsurf, Gemini CLI and more.

By design KPI Tree does not store your raw data; cached data is encrypted at rest in an HSM-backed engine.