Every person and agent acts under its own identity. Govern checks each action against your systems of record, sends regulated calls to a named approver and logs every step for replay.
Email the Tavrenta speaker dinner invitation to the 300 Northeast neurologists on the campaign list.
Agent proposalSend one email, with one invitation piece per recipient. Checks ran first.
Check pack: HCP communicationStep: Execute email and other digital channel sends
Check
Source of truth
Fails closed when
Result
Identity
Veeva Network
No hard-key match and no MDM-confirmed link
Passed
Consent
OneTrust, read live at send time
The live read fails, or consent is missing or withdrawn
41 removedNo email consent.
Territory and country
CRM alignments and the country rule pack
No alignment, or a rule bars the action
3 reroutedState rules bar the dinner.
Approved content
Vault PromoMats, read live
The approval has expired or been withdrawn
BlockedExpired piece replaced.
Patient privacy
Mask serve gate
A patient-level value would be served
Passed
Safety and medical routing
Safety screen and intake
The screen finds a hit
PassedNo hits.
Audit
The ledger
The ledger is unavailable
Passed
Send to 259 HCPs: 256 get the dinner invitation, 3 get the webinar version.Autonomy: Propose and wait. Decider: Brand manager, Tavrenta.
Ask Ontelis anything…
Trusted by 2 out of the 5 largest life sciences companies
A named person makes every regulated call.
Each step names a decider, a backup and a deadline. Agents never sign, approve content or assess an adverse event.
Every agent has an identity and a kill switch.
Each agent is registered with an owner, a purpose and a version. When it acts for a person, it gets only the permissions you both hold.
Decide what each agent may do alone.
Pick one of four levels for each step, from Act to Human only. The default follows how hard the action is to undo, and spend limits start at zero.
Checks read the source before anything runs.
Each check reads its system of record, such as OneTrust for consent or Vault for approved content. A stale or missing source stops the action. Prescribing data never decides what an HCP receives.
Replay any action exactly as it happened.
Every action, by an agent or a person, is written to an append-only, hash-chained ledger. Replay shows what the agent saw and which rules applied.
Some rules no one can turn off.
No setting, contract term or admin role skips Govern. Regulated steps are never auto-applied.
“What won over our compliance team wasn’t the speed. It was that the agent shows its work, stops at Draft, and tells you when it doesn’t know.”
Head of Marketing OperationsOne of the top five US biopharma companies
What Govern does not do
Govern enforces the customer’s rules; it does not write them.
It does not own consent, content or safety records.
It does not judge medical or legal questions.
Ontelis does not score employees or publish leaderboards.
Ontelis Govern is the AI agent governance layer of the Ontelis platform: it checks each agent action against your rules before the action runs. Before an agent emails a doctor or raises a purchase order, Govern reads your systems of record, such as OneTrust for consent and Vault PromoMats for approved content. Regulated decisions go to a named approver, and every step is logged so you can replay it later. Govern covers what passes through Ontelis, and it enforces your rules without writing them.
No. Ontelis has no admin mode that skips Govern, and no setting, contract term or admin role creates one. When a compliance or safety check stops an action, only the role that owns that check can clear it. A business approver can't. Ontelis staff never approve anything in your production systems.
Not by default. A sent email can't be undone, so Ontelis holds it for a named approver, such as the brand manager on a campaign. Before that, Govern reads consent live from OneTrust and approval status live from Vault PromoMats, removes doctors who haven't consented and blocks a piece whose approval has expired. If a live read fails, the send stops. Loosening this takes the step owner and compliance together, and only for one narrow, templated message, such as a logistics confirmation to a contracted venue.
You can, but Ontelis doesn't rely on it. Govern runs outside the model: an agent never holds your Salesforce or SAP credentials, and a change gets a short-lived credential only after it passes Govern's checks and any approval the step needs. Emails, invoices and web pages from outside your company are treated as data, never as instructions.
Accountability stays with named people at your company. Every Ontelis agent is registered to an owner who answers for it, usually the person who owns the process steps it runs. Each step an agent can't take alone has a named decider, a backup and a deadline, resolved from your org chart and delegation-of-authority matrix. Your compliance team owns the checks. Ontelis supplies the control and the record of who approved what.
Govern covers only what passes through Ontelis. Copilot, Agentforce or an agent your team built can act through the Ontelis Agent Gateway, and then gets the same checks, approvals and record as Ontelis's own agents. An agent that keeps its own Salesforce or SAP login gets no Ontelis checks. Govern detects its writes after the fact from the systems' audit fields and lists it as observed, or as unknown when a write can't be tied to any agent. Ontelis makes no governance claim for those.
Yes. Ontelis writes every action that passes through it, by an agent or a person, to an audit trail that can't be edited afterwards: who acted and for whom, which rule versions applied, what each check found and who approved. Replay rebuilds what the agent saw at that moment and shows the model output stored at the time, not a fresh one. An auditor gets a read-only export for a time window and process, and can follow it without Ontelis staff.
Govern produces the records, and your own AI governance team makes the compliance call. Its registry lists every agent, rule, policy, connector and model in Ontelis, each with a version. From that registry Govern produces the EU AI Act inventory and the GDPR record of processing, and it flags an impact assessment whenever an agent's data classes or purpose change. Ontelis doesn't certify your compliance or decide which obligations apply to you.
Your compliance team does. Ontelis ships starter checks for things like HCP communication, speaker programs and payments, and your team owns and edits them from there, including the rules for each country. Govern enforces them before an agent acts. Two things can't be edited out: the safety screen, and the rule that prescribing data never decides what a doctor receives. Govern doesn't judge medical or legal questions.