What is Pér

Pér is Amperity’s AI agent for customer data. Ask it about your customers in plain language, and it answers from the customer data your organization already keeps in Amperity. When the answer implies work — an audience to build, a campaign to set up, a model to train — Pér proposes that work as a plan and carries it out in Amperity once you approve it.

That cycle is the customer decision loop: Understand → Recommend → Approve → Act → Learn. Every part of Pér serves one of its stages.

Who Pér is for

Pér is for the people who already work in Amperity: the marketers and analysts who build audiences, segments, campaigns and predictive models, and the administrators who look after the tenant they work in.

Pér makes it simple to ask a question about customers, get answers grounded in data, and do the work in Amperity that acts on the answers. The same question that starts a conversation can end in configured, approved, running work, without leaving the conversation to do it.

What that looks like depends on the work you do:

  • If you plan and run marketing programs, ask Pér what is happening with a group of customers, then have it set up the audience, campaign or predictive model that acts on the answer.

  • If you analyze customer data, ask Pér to find and explain something in your tenant’s data, and keep what it produces as a report you can share with your team.

  • If you administer Amperity, control who can reach Pér and what they are able to do there.

Trusted customer context

Everything Pér says and proposes is grounded in your Amperity tenant’s own customer data, together with the standing instructions you have given it.

Four things make up that context:

  • Your identity-resolved customer data. Pér queries your tenant’s own customer tables, where records from different systems have already been resolved into one view of a person. It reads that data through Amperity, and never beyond the access it has been given.

  • The history in that data. Not only who your customers are, but what they have done.

  • The rules you have given Pér to work inside. Three things make these up:

    • Company context — the business priorities, definitions and KPIs you want Pér to work from, carried into every session.

    • Must-follow memories — standing rules you have told Pér to observe. Pér reads must-follow memories first and treats them as rules it must not break.

    • The approval boundary — nothing Pér proposes reaches Amperity until a person approves it.

  • The predictive intelligence available in your tenant. Where your tenant already has predictive models, Pér looks at the audiences they identify and proposes work that acts on them.

Alongside your Pér company context, Pér also reads the context documents and the AI Assistant system prompt your tenant has set up in Amperity.

Note

All of that material — company context, memories, your Amperity context documents, the AI Assistant system prompt, and anything Pér finds on the web — is treated as information to work from, but it cannot change Pér’s operating rules, grant it a permission, or move it to another tenant.

What Pér will and won’t do on its own

Pér reads on its own. It writes only with your approval.

This is a boundary, not an adjustable setting, and is why you can let an agent work directly in a production tenant. You can hand Pér a broad question without first deciding how much of your tenant you are willing to let it change.

How the boundary works:

  • Reading is unrestricted within your access. Pér can look at anything in your tenant that you could look at yourself, and it reads before it proposes.

  • Every write becomes something you approve. When Pér wants to change something in Amperity, it does not just do it. The change becomes a write confirmation or a step in a plan, and nothing is sent to Amperity until a person approves it.

  • One approval can cover a whole plan. When you approve and run a whole plan, you approve the plan — not each write inside it one at a time. Pér then re-checks at every step whether it may still go on, stops at any step that needs a person, and records which steps it approved on your behalf.

  • Some tools are withheld from Pér entirely. Whatever else is permitted, Pér cannot read credentials, create or delete people, grant or revoke access, change the shape of your tenant, or relax the confirmation gate itself. These are refused before any other rule is considered, so no setting and no instruction can re-enable them. The full list is in What Pér can’t do, whatever you approve.

Important

Approving a plan is not the same as watching each write go by. One approval can set a sequence of Amperity writes running, and a write that has run cannot be undone from Pér. Read a plan’s steps before you approve it.

Two things Pér does not do:

  • Pér does not watch your tenant between sessions. It does not act on its own, and it does not start a new round of the loop by itself. Each turn begins when a person begins it.

  • Pér does not tell you what your marketing achieved. It keeps a record of what it did and what it produced. Judging the business result of that work is still yours.

Where to use Pér

You can access Pér via:

  • The Pér web app, where the full experience lives — chat, recommendations, plans, approvals and the reports Pér produces.

  • Pér in Slack and Pér in Teams, where Pér answers in the channel where the work is already being discussed. Both answer only; a change is made in the Pér web app.

For the address, signing in and choosing a tenant, see Accessing Pér.