Skip to main content
Participant Memory gives your AI a sense of continuity. When the same person comes back for a second-round interview, a follow-up coaching call, or another sales pitch, the AI picks up where things left off instead of starting from scratch.

​
How it works

When a meeting ends, Waterr automatically distills the transcript into a short, factual memory about the participant — what they said, what they decided, what they cared about, what was left unfinished. Nothing about the persona’s own instructions, no small talk, no PII the participant didn’t volunteer themselves. The next time that same signed-in participant joins the same scenario, those memories are injected into the AI’s system prompt before the call starts. The AI uses them as ambient context — it doesn’t recite them back, it just acts as if it remembers. Memory is scoped per (participant, scenario). The same person joining a different scenario starts fresh — there’s no cross-scenario leakage.
Only signed-in participants are remembered in v1. Guests joining through public links or embeds aren’t remembered — the AI treats every embed session as a clean slate.

​
Use cases

Multi-round Interviews

Round 2 picks up the candidate’s experience, projects, and unresolved questions from Round 1 without re-asking.

Coaching & Therapy

The AI coach remembers goals, blockers, and what the participant tried since the last session — real continuity, no notes to upload.

Sales Discovery

A returning buyer’s role, objections, and the deal stage carry forward, so the AI doesn’t re-qualify the same lead twice.

Onboarding Journeys

A multi-week training program remembers what each learner has covered, where they got stuck, and what to focus on next.

​
Enabling Participant Memory

1

Open your scenario

Navigate to the scenario you want participants to be remembered in.
2

Go to Settings

Click the Settings tab.
3

Turn on Remember Participants

Under Participant Memory, toggle on Remember Participants.
That’s it. The next time a signed-in participant has a session in this scenario, Waterr writes a memory at the end. From their second session onward, the AI starts the call already knowing them.

​
What gets remembered

The post-meeting distillation captures durable, useful context:
  • Facts the participant shared (role, company, project they’re working on)
  • Stated preferences and decisions
  • Names and relationships they mentioned
  • Loose threads — questions they didn’t finish answering, things they said they’d think about
It deliberately skips:
  • Small talk and pleasantries
  • Anything already in the persona’s own instructions
  • PII the participant didn’t volunteer themselves
Each memory is capped at roughly 500 tokens. On read, Waterr injects up to ~4,000 tokens of memory — newest first — into the system prompt. Older memories beyond that ceiling are dropped from injection but remain stored.

​
Guiding how the AI uses memory

You don’t need to write special prompts to make memory work — that’s automatic. But you can steer how the AI uses it in your meeting script:
Memory works best in scenarios designed for repeat sessions — coaching programs, multi-round interviews, ongoing customer conversations. For one-shot scenarios, leaving it off keeps things simpler.

​
Privacy & data handling

  • Encryption at rest — memory text is AES-256 encrypted in the database, same as transcripts.
  • Opt-in per scenario — off by default. Nothing is written unless you toggle it on.
  • Scenario-scoped — memory from Scenario A is never visible to Scenario B, even for the same participant.
  • Authenticated only — no anonymous participant fingerprinting. If there’s no signed-in identity, nothing is written.
For broader privacy controls, see Privacy and Data Encryption.

​
Limitations (v1)

  • No creator-facing memory editor yet. You can turn the feature on/off per scenario, but you can’t review or hand-edit individual memory entries from the dashboard. Coming in v2.
  • Embed participants aren’t remembered. Public-link and embed joins are unauthenticated; memory only applies to signed-in users.
  • One scenario at a time. A participant’s memory in Sales Discovery v2 is independent from their memory in Sales Discovery v1 — even if the scenarios are nearly identical.