RevOps playbook · AI Employee: Rev

Call Recording Lookup

Sellers find call moments 10× faster

Install this playbookBack to marketplace
The problem

Every sales team has hundreds of recorded calls in Gong or Chorus. Finding the moment a customer said "we care about SOC 2" or "we already use Snowflake" means listening to a full call — nobody does. So competitive intel, product feedback, objection patterns, and pricing signal all live in a corpus nobody queries. The recordings are an archive, not a knowledge base.

At a glance
Trigger
Chat
Approvals
None
What it does
Read-only
Systems
Gong · Chorus
How it feels in production

An hour-by-hour walkthrough.

"Rev, when did anyone last mention data-residency on a call with a bank?" Rev queries Gong across the last 90 days for calls where account.industry = "Banking" and the transcript contains data-residency terms. Returns 14 calls. Ranks by recency + deal-stage relevance. Presents the top five with: - Account name + deal stage + ACV - Rep + account exec on the call - Direct quote (verbatim from transcript, with speaker attribution) - 30-second time-jump link into the recording - What the rep answered and how the customer reacted The PMM building the banking sector deck now has five real customer quotes with context in 30 seconds instead of a week. "Rev, has any customer objected to our new pricing this month?" — same mechanism, different query. "Rev, show me the last three times we lost to Snowflake" — same mechanism, filtered on Closed Lost + competitor mention. Every question against the corpus of things customers actually said. Reps can ask the same way. "Rev, how have other reps handled the SSO objection on Enterprise deals?" — three examples, three responses, three outcomes.
How it works

Step by step.

  1. 01

    Sync transcripts + speaker attribution

    Continuous sync from Gong / Chorus / Fireflies. Each call carries: transcript with speaker labels, account + opportunity linkage, participants, call type, deal stage at the time.

    Gong · Chorus · Fireflies · Salesforce join
  2. 02

    Parse the natural-language query

    "Times we lost to Snowflake" → filter on stage=Closed Lost + competitor.mentioned=Snowflake. "Data-residency mentions in banking" → industry=Banking + phrase search. Rev interprets and confirms the interpretation.

    Query understanding · CRM filters
  3. 03

    Rank results by relevance + recency

    Not just keyword match — rank by deal-stage relevance to the asker's context (a PMM sees closed deals first, a rep sees open deals first), recency, quote clarity, and outcome.

    Ranking · Reasoning
  4. 04

    Return verbatim quotes + time-jump links

    Each result: account context, verbatim customer quote with speaker attribution, deep link to the exact timestamp in Gong / Chorus, the rep's response, subsequent deal outcome.

    Slack · Teams · Web UI
  5. 05

    Log the query + refine the corpus

    Repeat queries become saved views. Top-asked topics feed a weekly "what customers are talking about" digest for product + marketing. Queries with poor results flag missing tags.

    Query log · Analytics · Digest
Systems and wiring

What you connect to make this run.

Gong · Chorus · Fireflies

read

API sync of transcripts, speaker labels, call metadata, deal + account joins. Deep-link generation for time-stamped playback. No writes back — call intelligence stays canonical there.

Salesforce · HubSpot

read

Account, opportunity, industry, ACV enrichment. Enables filters like "banking" or "deals over $100K" without the asker specifying account IDs.

Slack · Teams

read+write

Ask questions in-channel or DM. Results returned in-thread with quotes and deep links. Follow-up questions chained conversationally.

Analytics · Query log

write

Every query logged. Top topics feed weekly customer-voice digest for PMM. Frequency-of-asked drives which topics get pre-tagged for faster future lookup.

What changes

Before and after, honestly.

Time to find a specific customer quote
Before
45-180 minutes (listen to full calls)
After
Under 30 seconds
% of recorded call value actually referenced later
Before
Under 5%
After
30-50%
Objection-handling training loops per rep per quarter
Before
0-2 (manager finds one example, shares it)
After
10-20 (rep pulls three examples before every enterprise call)
Time PMM spends on customer research
Before
3-5 days per sector deck
After
2-4 hours
Related

Playbooks that pair with this one.

RevOps

Deal-at-Risk Detection

Same corpus feeds the competitor-mention signal in at-risk scoring.

Frequently asked

Answers about this playbook.

What about privacy — every rep can see every call?

Rev respects your Gong / Chorus permission model. If a rep can't see a recording in Gong, Rev's results filter that recording out. Managers see their team; leaders see the region; nobody sees more than they'd see natively.

Can it summarise a single call?

Yes — one call in, structured summary out: participants, deal context, top three topics discussed, objections raised, next steps agreed, sentiment shifts. Faster than listening at 1.5x.

What if the transcript is wrong (bad audio, accents)?

Rev flags low-confidence transcript segments and surfaces them; verbatim quotes come with a confidence indicator. For critical evidence (renewal negotiations, exec quotes), the human can jump to the audio to verify.

Can it work across languages?

Yes — Gong / Chorus support multi-language transcripts. Rev queries in whichever language you ask; results returned in matching language with translations when needed.

How does it compare to Gong's own search?

Gong search is keyword-first and single-recording-scoped. Rev synthesises across recordings, joins to CRM context, ranks by relevance to the asker's role, and returns actionable answers rather than lists of hits.

See it run on your data.

Free plan, no credit card. Connect the systems this playbook needs and run it against a past event first.

Book a demoMore RevOps playbooks