Live forecasts

Automated forecasting, for the metric you choose.

Choose the metric and horizon. Predict.ai evaluates candidate drivers, compares models on held-out history, deploys the strongest result, and keeps the forecast current as new data arrives.

Goal overview showing a live e-commerce revenue forecast with accuracy, drivers, and serving status — A live goal — forecast, accuracy, drivers, and serving status at a glance.

Driver discovery

Find the signals that improve the forecast.

Predict.ai searches across the workspace, not only the columns you select. Candidate inputs are ranked by lag, predictive strength, and confidence, with the tested and rejected relationships left visible.

Lagged relationships scored on historical evidence
Multi-step paths, such as fuel cost → shipping rate → daily sales
Relationship status and evidence strength visible for review

Relationship graph mapping how workspace signals correlate and which ones lead — The target at the center, drivers scored and lagged around it.

Model tournament

Compare models fairly. Deploy the strongest result.

Classical learners, foundation models, and any model you've built all train on the same discovered drivers and get scored on the same held-out folds. You can watch the leaderboard, or override the pick yourself.

Foundation, custom, and your own models — same tournament
Purged walk-forward folds, so scores don't leak the future
Manual override any time — autopilot is a default, not a cage

Model tournament leaderboard with per-fold scores and the champion model highlighted — Per-fold scores for every candidate, champion highlighted.

Scenarios & sensitivity

Simulate what-if scenarios before you decide.

Move a driver by a percent, ramp it over time, or inject an event — the live champion re-infers and shows you the delta, the attribution, and an honest caveat about what's simulated versus measured.

Natural-language what-ifs — describe the scenario, get the chart
Tornado analysis ranks which driver moves the forecast most
Event analogues show what actually happened last time, measured

Scenario builder with natural-language what-if presets and driver sensitivity analysis — Describe a what-if in plain English — the live model simulates it.

Alerts & anomalies

Get notified when the forecast crosses a line.

Attach a threshold, a change, or an anomaly condition to any goal. Every new forecast is checked against your rules, and you're pinged the moment one fires — with severities, cooldowns, and a live preview that shows exactly why.

Email, Slack, Teams, PagerDuty, or webhook — your channel
Preview an alert before it's live: "would this fire right now?"
Severity levels, cooldowns, and teammate routing built in

New alert dialog with a threshold condition and a live preview showing it would trigger on the latest forecast — Define the condition — the live forecast answers “would this fire right now?”

AI data analyst

Ask the forecast anything.

The workspace's AI analyst docks beside any goal and reads its real model, drivers, and live forecast. Ask why the number moved, what's driving it, or what happens if something changes — and get answers with charts, in plain English.

Ask why a forecast moved, what's driving it, or what happens if X
Charts drawn from the live model's actual output, not a guess
Answers cite the goal's real drivers, folds, and accuracy

AI analyst conversation grounded in the goal's deployed model and live forecast — A question, an answer grounded in the live model — not a guess.

Autopilot, with guardrails

Automated, with spending guardrails.

Set a monthly spend cap per goal — autopilot pauses itself and notifies you before it goes over, not after. Every automated action lands in a timeline with its trigger and its cost, so nothing runs unaccounted for.

Whole-workspace search

Discovery scans every signal you have, not just the ones you name.

Champion, not a guess

The model that wins the tournament is the model that deploys.

What-if, quantified

Every scenario returns a delta, a chain of attribution, and a caveat.

Notifications, not chores

Thresholds, changes, and anomalies — you hear about it when it matters.

Answers with context

The analyst combines the deployed model's output with the goal's drivers, history, and forecast.

Spend caps, honored

Automated goal spend stops at the cap you configure while monitoring continues.

Put your data to work.

Connect your own data or start with a sample. Explore automatic insights, ask questions, engineer features, and build a forecast in the same workspace.

Start with your data Talk to sales