Replayfy AI uses leading language models under the hood to plan and
narrate. The reasoning is grounded in your workspace’s own data — the model
never sees another customer’s sessions, and it only reports numbers it
actually pulled.
Where to find it
Replayfy AI is reachable from every page of the dashboard at app.replayfy.app:- The Ask AI panel — a floating Ask AI button in the bottom-right corner. Click it and a chat panel springs open. It stays with you as you move between Recordings, Funnels, Users, and Settings, so you can ask about whatever you’re looking at.
- The Ask bar on the Overview — a command box at the top of your Overview with rotating example questions. Start typing and press ⏎ to ask.
- “Investigate” actions — buttons scattered through the product (on a signal, a crash, a drop-off) open the panel pre-loaded with the right question.
How to ask
1
Open Ask AI
Click the Ask AI button in the bottom-right of
any page, or use the Ask bar at the top of your Overview.
2
Ask in plain language
Type a real question — no query syntax, no filters to configure. Press
⏎ to send. You can pick one of the suggested prompts to start.
3
Watch it investigate
Replayfy AI shows its plan and streams each step as it runs —
“Investigating conversion…”, “Reading release 1.6.2…” — then writes the
answer token by token. Expand the thinking row to see exactly which
sources it read.
4
Follow the evidence
Answers come with citation chips. Click a session chip to jump straight into
that recording, or a funnel chip to open the funnel. Use the suggested
follow-up buttons to keep digging.
Grounded answers, not guesses
Replayfy AI is deliberately honest about what it knows:Answers are grounded in your structured product data — tracked events,
funnel steps, releases, crash groups, journeys and session metadata — not in
the pixels of a replay. The richer your event tracking, the sharper the
answers. See Track events.
Actions it can take for you
Beyond answering, Replayfy AI can do the work. Ask it to build something and it assembles the definition, shows you a preview, and creates it — stamped “Created with Replayfy AI” so your team can see where it came from.Build a Funnel
“Build a funnel from Landing → Signup → Purchase” — it defines the steps and
creates the funnel for you.
Create a Cohort
“Make a cohort of users who rage-tapped checkout on Android” — it translates
the description into a real, saved cohort.
Watch an issue
“Alert me if crash-free rate drops below 99%” — it drafts the alert and can
watch a specific issue for recurrence.
File it out
With an integration connected, it can draft a Linear or GitHub issue, or a
Slack message, from what it found.
The confirmation gate
Anything that writes outside Replayfy — filing an issue, posting to a channel — pauses for your approval first. Replayfy AI shows you the full outgoing payload (the issue body, the exact Slack fields) and nothing is sent until you click Confirm.It learns your workspace
When a question is ambiguous — “which checkout event do you mean?” — Replayfy AI asks, and remembers your answer so it never has to ask again. Admins can review, correct, or forget everything the AI has learned about the workspace, and every action it takes is written to an audit trail you can inspect.Proactive intelligence on the Overview
You don’t have to ask first. When AI is on, your Overview leads with insights the AI generated by reading the same data you’d query:Storyline
Storyline
An analyst-written lead at the top of the Overview — the single most
important thing happening in your product right now, in prose, with a
confidence chip and the supporting evidence. It’s regenerated as your data
moves, never templated.
Signals
Signals
The ledger of what moved — conversion, stability, engagement and segment
shifts, ranked by impact. With AI on, each signal gains a short analyst read
on top of the raw numbers.
Worth watching & experience health
Worth watching & experience health
A rolled-up health read of the current window and a short watchlist of the
things most likely to need your attention next — surfaced before they become
incidents.
AI credits & metering
Replayfy AI runs on credits. Every request — an Ask turn, a Storyline refresh, a set of Signals — consumes credits in proportion to the work it did. Credits are generous, round numbers rather than fractions of a cent:- ≈ 100,000 credits = $1. Every request is metered against this rate.
- A monthly allowance is bundled with your plan — from a taster on Free up to millions of credits on higher tiers — sized so normal usage stays inside it.
- Optional top-ups. If you run through the monthly allowance, buy more with a one-time Buy credits purchase; they stack on top of your bundled allowance.
When a workspace runs out of credits, Replayfy AI pauses rather than billing
you by surprise — it resumes when the allowance resets or you top up. Full
plan allowances and pricing live in
Workspaces & billing.
Who can use it
Viewers are invited to watch recordings, so the Ask panel isn’t offered to them.
What feeds better answers
Replayfy AI reasons over whatever your SDK sends. The more you identify users and track meaningful events, the more precise your answers about conversion, segments, and journeys become.Once events flow in, ask “How is checkout_started converting this week?” —
the AI will pull the numbers, state the window, and cite the sessions behind
them.
Related
Product analytics
Where Storyline, Signals, and the health read live.
Track events
Send the events that make AI answers precise.
Workspaces & billing
Plan allowances, credits, and top-ups.