An agency invoices for effort. Umbodi bills for work performed, does it under a governance setting you choose, and records who changed what — across search rankings and what the AI answer engines cite.
Every act is logged with its actor. To know a change actually worked, a tenth of fixes are withheld as a control group.
What changes when an agent does the work instead of a retainer.
A monthly plan with a credit allowance, and you are never billed for work that fails.
Five categories, capped by plan. Publishing to your own site is its own deliberate setting.
A tenth of fixes are withheld as a control group, so lift is attributed rather than asserted.
We put your questions to five answer engines and record whether your domain is among the cited sources. A citation is counted, never inferred — and an answer that cites nobody is recorded as exactly that, not as a zero.
The modes differ by autonomy ceiling, not by feature: it prepares the change for someone to apply, it opens a pull request your team merges, or — only if you switch it on — it merges on green, watches the deploy and puts the change back if it did not take. By default every act ends with you.
Referring domains gate every other number, and no code earns a backlink. Umbodi counts who links to you today, finds the roundups already ranking for your terms that omit you, reads the byline off the page and drafts the approach — then stops. It will not mail a stranger on your behalf.
The governance is in the product, not in a policy document.
Every applied change writes an entry naming the actor and the page it touched.
Row-level security, per-account budgets and a circuit breaker on runaway spend.
Three of the four presets keep publishing with a person. Autopilot is a decision, not a slider position.
Umbodi replaces the recurring execution half of an SEO retainer: the audits, the meta and schema work, the content briefs and the follow-up. It does not replace strategy, brand or commercial judgement, and it says so rather than implying otherwise.
Autonomy is set per task across five categories and capped by the account plan, so a user cannot grant themselves more than the account allows. Every applied change writes an apply-log entry naming the actor, the page and the outcome, and the log is queryable.
Umbodi deliberately withholds a tenth of eligible fixes as a control group and compares the two arms, so a reported gain is attributed rather than assumed. Measured rankings and projected lift are kept strictly separate, and where there is not enough data the product reports that instead of a number.
Register and run it on one real property. The audit costs nothing.