How it works
From symptom to root cause.
Inversion is a guided root-cause investigation. Instead of handing you a one-shot answer, it builds a picture of the problem and works through it with you until the evidence closes on the cause. Here is the method.
The method
Four stepsDescribe the effect
State the problem or outcome you want to trace — a failure, a result, a finding. Add whatever context and stakes you have; more signal means a sharper starting graph.
Review the causal graph
Inversion generates the plausible causes as a directed causal graph — every node a real, testable cause, with the branches worth chasing — so you can see the whole problem at once, not a wall of text.
Investigate with the agent
An AI investigator works the graph with you, one question at a time. It asks for the evidence that would confirm or eliminate each branch, logs your findings with a strength rating, adds causes as facts emerge, and rules out the branches the evidence kills.
Confirm, plan, export
When the evidence settles, it marks the confirmed root cause, drafts a corrective action plan with safety checks, and assembles an editable report you can export to PDF, PowerPoint, or Word — or share with collaborators.
What makes it different
Why it holds upIt moves on evidence, not plausibility
Branches shift when there's a hard finding behind it — measured, observed, or reported — and every finding is logged with its strength so the reasoning is auditable.
Ruled-out stays visible
Eliminated causes aren't deleted; they stay on the graph, struck through, so anyone reading the report can see what was considered and why it was dropped.
It weighs every branch
Under the hood it continuously re-weighs each branch to decide what to ask next — so the investigation spends your time on the causes that actually matter.
Any domain
The method is domain-agnostic. The same investigation works for a fired-heater trip, an off-spec batch, a clinical finding, or a business result you can't explain.
Three modes
Pick the depthThe flagship investigation — the causal-graph method above, end to end.
OpenTurn the model into the specialist you need; it interviews you before it answers.
OpenA fast go-to chat that pulls live web sources when the facts matter.
OpenRun your first investigation.
One real Inversion investigation and 150 chat messages on the free plan — no card required.