Forecasts with their own error bars, the reasons behind a number, what-if on your own inputs, alerts that learn your normal, and the recommendation that follows — each one computed from your rows, never invented.
Every step is a question a business owner actually asks. Basics first; the advanced steps are one click further.
Your files become a workspace the platform can reason about.
Rows are profiled, addresses geocoded, columns typed — automatically.
Revenue is revenue, a customer is a customer — the assistant, reports and recommendations all rely on it.
Plain English → a chart, a table or a map from your own rows.
'Now only Canada', 'same for last year' — the assistant remembers the thread.
A number checked on a schedule; you are told when it is unusual.
Replenish, win back, fix margin, plan capacity — with the expected impact.
Every geocoded record, live, with the trade area around it.
Your data as a live page on your own URL — the map with a card for every location, or a chair's Today — with an Ask block over exactly that data.
Computed numbers, AI prose, your brand — a PDF for the board.
When the basics are in place.
Contracts, price lists and invoices become passages the assistant quotes with page numbers.
Orders joined to customers, stores, facilities — questions across tables.
Where a number is heading, with how wrong the method was last time.
What happens to margin if freight falls, the odds of hitting the target, a price change, demand vs capacity.
A URL, sheet or QuickBooks entity re-read on a schedule; every import kept as a version you can go back to.
Predict a number or a label from your columns, or project a measure — named, saved, reusable next quarter, with honest holdout metrics.
Invite by email with a role; keep a project to its members; leave notes on data, reports and recommendations.
Every morning your data changed, the analysis runs and the PDF lands in the inbox you choose — after your approval, or straight away once you trust it.
A key scoped to the APIs you hold; metered per call.
Each has its own page.
Which columns move an outcome, measured on held-out rows — and whether a change did anything.
Drivers and why →What happens to margin if freight falls 20%? What are the chances we reach the target next year? What if we raise prices 5%? When do we run out of capacity if demand grows? What does my saved model say happens if an input changes? What happens to margin if the aluminium can costs 10% more?
What if →Saved questions checked on a schedule; actions that leave the workspace wait for your approval.
Watchers and actions →What will we run out of? Who stopped buying? Where are we losing margin? When do we hit capacity?
Recommendations →Contracts, price lists and PDFs searchable beside your tables; passages are quoted, never obeyed.
Documents →Last night golden checks passed (95%) against a demo company whose true numbers are known.
Evaluation and trust →Single sign-on, roles, an audit trail, human approval of outbound actions, and your own AI keys if you want them.
Security →