Agents
An agent is only as good as the measurement under it.
Quantiv's agents work from the event store, taxonomy, and attribution your brand owns. They propose what to do, and act only through paths that are guarded and recorded. You decide how much latitude they have.
Three agents
Each one covers a job your team does by hand today.
Instrumentation agent
Proposes the events and classifications your taxonomy is missing. You approve.
Read moreSignal health agent
Clusters the failure surface, names the cause, proposes the fix.
Read moreMedia buying agent
Budget proposals through the same guarded apply path, with the audit log attached.
Read moreThe approval model
Proposal, review, guarded apply, audit log.
No agent has a private way to change anything. Every action takes the path a person's change would take, and every step is recorded.
- 01
Proposal
The agent drafts a change and attaches the data behind it.
- 02
Review
A person approves, edits, or rejects it. Nothing is applied yet.
- 03
Guarded apply
Checked against spend caps and limits, then applied once, idempotently.
- 04
Audit log
Who proposed, who approved, what changed, and when.
Autonomy
Latitude is granted, not assumed.
Set per agent, per campaign, and per spend band. Extend it as confidence builds, and reduce or switch it off at any time.
- 1
Assistant
Proposes. Every change waits for a person to approve it.
Suggests moving $600/day from a fatigued ad set to a stable one.
- 2
Copilot
Applies small changes inside limits you set. Larger ones still wait for approval.
Applies budget moves under 10% on its own and queues anything larger for review.
- 3
Autopilot
Operates within explicit guardrails per campaign and spend band, and reports what it did.
Manages daily budgets on prospecting sets up to a $3,500/day cap, with a daily summary.
Boundaries
What agents can see, and what they can never do.
They see your spine
Agents read the events, taxonomy, attribution, and campaign data in your own workspace, scoped to what you assign them.
They cannot activate a definition
Taxonomy changes and sensitivity classifications always require a person.
They cannot exceed a limit
Spend caps and change limits are enforced on the apply path itself.
They cannot bypass the egress filter
Agents are subject to the same classification rules as every other part of Quantiv.
Questions
Common questions.
Do the agents send our data to a third party?
Agents operate on data in your own workspace. Data leaves only through the destinations you configure, and only through the egress filter.
Who is accountable for an agent's change?
The audit log records the proposal, the person or rule that approved it, and the change applied. Accountability is explicit at every step.
Can we turn an agent off?
Yes, immediately, per agent or per campaign. Turning an agent off does not affect your data or your other settings.
See the agents on your own data.
A two-week pilot runs your conversions through Quantiv alongside what you report today. The difference between the two is the business case.