Conversation intelligence is useful when it turns calls into actionable coaching rather than just recordings. The CRM connection matters because it lets managers relate talk tracks, objections, and call quality back to specific deals and reps.
Conversation intelligence (CI) platforms record, transcribe, and analyse sales and customer success calls, then surface patterns and coaching insights to managers and reps. Connected to your CRM, they transform calls from ephemeral conversations into structured data: who talked and for how long, which topics were discussed, which objections were raised, how the prospect responded, and what was agreed at the end. For sales managers, CI provides the ability to coach individual rep performance at scale without listening to every call manually. For CRM strategy, CI provides the richest source of deal intelligence and rep behaviour data available. This guide covers how CI works, the major platforms, and how to integrate them with CRM effectively.
That is what makes the feature operational instead of decorative.
What Conversation Intelligence Captures
| Data Point | How Captured | CRM Use |
|---|---|---|
| Call transcript | Automatic speech-to-text of the recording | Stored as note on deal/contact record; searchable across all calls |
| Talk/listen ratio | Speaker identification and time calculation | Coaching metric — reps talking >65% of the time are failing to listen |
| Topic detection | NLP identifies topics discussed: pricing, competitor, implementation, contract | Deal intelligence — did pricing get discussed? Was competitor mentioned? |
| Question ratio | Count of questions asked by rep vs prospect | Coaching metric — discovery quality indicator |
| Sentiment | Tone analysis of prospect’s responses | Deal risk indicator; prospect engagement signal |
| Next steps confirmed | Detection of next-step language at call end | Deal progression quality — calls without confirmed next steps stall 3× more often |
| Filler words / hesitation | Frequency of “um”, “like”, “you know” | Rep confidence and fluency coaching |
| Monologue detection | Flags uninterrupted talking segments longer than 4 minutes | Coaching signal — long monologues indicate feature-dumping rather than discovery |
Major Conversation Intelligence Platforms
Gong: the market leader in conversation intelligence. Integrates with Zoom, Teams, Meet, and telephony systems. Records and transcribes all sales calls. Analyses talk/listen ratio, topic coverage, risk signals, and deal momentum. Pushes call summaries, sentiment, and next steps to Salesforce and HubSpot. Also provides portfolio-level analytics: what do top reps do differently from average reps across thousands of calls? Enterprise pricing (typically $100-200K+/year for mid-sized teams). Best for organisations that want deep call analytics and can invest in making it a core part of the sales process.
Chorus (by ZoomInfo): Gong’s primary enterprise competitor. Similar feature set: recording, transcription, AI analysis, CRM integration. Advantage: bundled with ZoomInfo’s data enrichment for organisations already using ZoomInfo. Similar price point to Gong.
Fireflies.ai: lower-cost option with strong CRM integration. Records and transcribes meetings across all major video platforms. AI generates meeting summaries, action items, and topic highlights. Pushes to HubSpot and Salesforce. Free tier available; paid plans from ~$10-19/user/month. Less sophisticated analytics than Gong but sufficient for teams that primarily want call logging and summary generation rather than deep rep behaviour analysis.
Fathom: free for individual users; team plans available. Focuses on meeting summaries and note-taking rather than sales analytics. Good for teams that want automatic call notes without enterprise CI pricing. Limited CRM integration depth vs Gong/Chorus.
Salesloft Conversations / Outreach Kaia: built-in CI features within the Salesloft and Outreach sales engagement platforms. For teams already on these platforms, the built-in CI may be sufficient without a separate CI licence. Less sophisticated than Gong but integrated with the cadence/sequencing workflow.
CRM Integration: What Gets Synced
Configure the CI platform to push the right data to the CRM — not everything, just what’s actionable:
Push to CRM deal record:
- Call summary note (AI-generated, 3-5 bullet points: what was discussed, concerns raised, next step)
- Recording link (clickable from the deal timeline)
- Topics discussed (deal properties: “Pricing discussed: yes”, “Competitor mentioned: Competitor X”)
- Next step confirmed (property: yes/no — for automation triggering)
- Deal risk flags if CI platform generates them
Do not push to CRM: full transcript text (too long, not useful in CRM), detailed analytics per call (belongs in the CI platform’s analytics dashboard, not CRM records). Keep CRM records clean — the CI platform is where detailed call analysis lives; CRM is where the key signals and summaries surface.
Using Conversation Intelligence for Sales Coaching
The coaching value of CI comes from two capabilities: (1) managers can review calls without attending them, and (2) patterns across many calls reveal what separates high performers from average performers.
Deal review workflow: when a deal is flagged as at-risk (stalled, competitive, large and late-stage), the manager can review the last 2-3 calls directly from the Gong or Chorus interface — linked from the CRM deal record. Without CI, this review would require scheduling a joint call or relying on the rep’s self-report. With CI, the manager can listen to the specific moment where the prospect raised a concern and evaluate how the rep handled it.
Portfolio coaching: Gong’s “People Intelligence” and Chorus’s equivalent allow managers to compare rep behaviour metrics across their team — which reps have talk/listen ratios above 65%, which reps rarely confirm next steps, which reps use the most filler words. This comparison identifies the specific behaviours to coach rather than generic performance conversations.
Onboarding acceleration: new rep onboarding is accelerated by CI libraries. Rather than job shadowing a single senior rep, new reps can review a curated library of excellent discovery calls, objection handling examples, and demo techniques. Gong and Chorus both support call libraries with tagging and search.
Making CRM Conversation Intelligence Operational: Beyond Call Recording
Conversation intelligence platforms generate data that most organisations underuse. Call recordings, transcripts, keyword alerts, and talk-to-listen ratios accumulate in the conversation intelligence dashboard while managers continue to coach based on rep self-reporting and occasional live call shadowing. The gap is not in the data but in the process: conversation intelligence data that is not systematically reviewed, routed to the CRM, and incorporated into coaching conversations does not change outcomes.
The most useful version of the workflow is the one that keeps improving behavior over time. If the team cannot connect the insight to a concrete next step, the analytics are not doing enough work.
Common Problems and Fixes
“Reps are uncomfortable with call recording — they feel surveilled and their performance declines”
This is a change management problem more than a technology problem. Discomfort with recording is common and rational — reps are suddenly aware that every conversation is reviewable by their manager. Fix: (1) introduce CI with a coaching-forward framing, not a performance-management framing — “this is to help you improve, not to catch you making mistakes”; (2) make recorded calls available to the rep themselves first — many reps improve significantly from self-reviewing their calls; (3) establish a norm that managers use CI for coaching conversations, not disciplinary ones, and enforce this consistently. Trust is built over months; breaking it (using CI for disciplinary action without prior disclosure) destroys it permanently.
“We’re getting call summaries in CRM but nobody is reading them — they’re too long”
AI call summaries can be verbose if not configured properly. Fix: (1) configure the summary template in the CI platform to generate 3-5 bullets maximum, not a full paragraph: “Problem identified: [X]. Objections: [Y]. Next step agreed: [Z]. Risk flags: [W].” (2) Push the summary as a CRM note with specific properties in separate fields (not as a wall of text in a single note field) — this makes them scannable without reading. (3) Train reps to review and edit the CI summary before it syncs to CRM — they spend 60 seconds adding anything the AI missed, and the CRM record reflects the edited version.
Problem: Conversation Intelligence Data Is Reviewed Reactively Rather Than Systematically
Managers who review call recordings only when a deal is lost or when a rep asks for help are using conversation intelligence reactively. By the time a deal is lost, the call patterns that contributed to the loss have been repeated across many calls. Systematic review, where a defined set of calls is reviewed every week based on objective criteria rather than manager availability, identifies coaching patterns before they become performance problems.
Fix: Define a systematic call review protocol that specifies how many calls per rep per week are reviewed, and which calls are selected. A practical protocol for a team of six to ten reps is: review two calls per rep per week, selected as follows: the call with the lowest talk-to-listen ratio (the rep is doing most of the talking, which is a coaching risk), and the most recent discovery call on a new opportunity above a defined value threshold. Use the conversation intelligence platform’s filtering tools to identify these calls automatically rather than manually searching through recordings. In the CRM, log the call review as a coaching activity and record the feedback given. Track coaching activity volume alongside rep performance to build evidence that coaching produces improvement.
Problem: Keyword Alerts Are Configured but Not Acted Upon
Most conversation intelligence platforms allow keyword alerts to be configured: when a defined word or phrase is detected in a call (competitor name, budget question, contract term), an alert is generated. Organisations that configure keyword alerts but do not build a process for acting on them find that the alerts accumulate unread. The value of keyword alerts is entirely in the response they trigger, not in their generation.
Fix: Build CRM workflows that respond to keyword alerts automatically. When a competitor keyword alert is triggered, the CRM should write the competitor name to the Competition field on the associated deal record if it is not already populated, and create a task for the rep: competitor mention detected in most recent call, review positioning and update deal notes. When a buying signal keyword is detected (words like budget approved, need to decide by, we have chosen to shortlist), create an urgent task for the rep to follow up within 24 hours and advance the deal stage if appropriate. When a risk keyword is detected (not a priority, budget has been cut, we are pausing the evaluation), alert the manager and flag the deal for pipeline review. Automated CRM responses to keyword alerts transform alerts from passive notifications into operational triggers.
Problem: Talk-to-Listen Ratio Data Is Collected but Not Used in Coaching
Conversation intelligence platforms report talk-to-listen ratios at call and rep level, with research consistently showing that reps who listen more and talk less in discovery calls achieve higher close rates. Despite this data being available, most managers do not incorporate talk-to-listen ratios into their coaching conversations because they are not sure how to use the metric constructively.
Fix: Make talk-to-listen ratio a standard metric in rep performance reviews and coaching conversations, alongside pipeline and activity metrics. Establish a team benchmark: the average talk-to-listen ratio for closed-won deals versus closed-lost deals in your last 12 months of data. Show reps their individual ratio compared to the team benchmark and compared to their own historical trend. For reps with consistently high talk-to-listen ratios (talking more than 60% of the time in discovery calls), coach specifically on the questioning techniques that shift the conversation: SPIN situation and problem questions require the rep to listen; feature-benefit pitching does not. Record the coaching conversation in the CRM and schedule a follow-up review of the rep’s next five discovery calls to measure whether the ratio improves.
Frequently Asked Questions
What is a good talk-to-listen ratio for B2B sales calls?
Research from Gong and other conversation intelligence platforms consistently shows that the optimal talk-to-listen ratio for discovery calls is approximately 43% talking and 57% listening for the rep. For demo calls, the ratio shifts towards higher rep talk time (around 55-60%) because the rep is presenting. For closing conversations, a more balanced ratio (50-50) tends to produce better outcomes than highly rep-dominated conversations. These benchmarks apply to B2B sales with complex products. Simple transactional sales may have different optimal ratios. Use your own closed-won versus closed-lost call data to establish the benchmark that applies to your specific selling context, as vendor-provided benchmarks may not reflect your market.
How does conversation intelligence integrate with Salesforce and HubSpot?
Gong integrates with Salesforce and HubSpot through native connectors that sync call data to CRM deal and contact records. Gong can write deal insights, next steps, competitor mentions, and risk signals back to Salesforce Opportunity and HubSpot Deal fields automatically. Chorus (now Zoominfo Chorus) also integrates natively with both platforms. For Salesforce users, Salesforce’s native conversation intelligence tool (Einstein Conversation Insights) provides built-in call intelligence without a third-party integration. For HubSpot users, HubSpot Calls with AI transcription and the Gong integration are the two most common approaches. The key integration requirement is bidirectional sync: conversation data flows from the intelligence platform to the CRM, and CRM deal data (stage, value, contact details) flows from the CRM to the intelligence platform for call context.
What categories of call data are most useful for sales coaching?
The conversation intelligence data categories most useful for sales coaching are: questions asked per call (reps who ask more questions in discovery, particularly in the problem and implication categories, close at higher rates), topic coverage (was the business problem discussed before the product? was pricing discussed before value was established?), objection handling (how many objections were raised and how were they addressed?), next step commitment (did the call end with a specific agreed next step, or a vague will be in touch?), and competitor discussion (how often and in what context were competitors mentioned?). Of these, next step commitment rate is the highest-use coaching metric: calls that end with a clear, agreed next step advance to the next pipeline stage at significantly higher rates than calls that end without one.
Is conversation intelligence appropriate for all sales team sizes?
Conversation intelligence platforms are most cost-effective for sales teams with five or more reps who make a significant volume of calls (20 or more per week per rep). For smaller teams, the call volume may not justify the licence cost, and a manager can review calls manually using the native recording features of Zoom or Teams without a dedicated intelligence platform. For teams of 15 or more reps, a conversation intelligence platform becomes almost essential: manual call review at scale is not feasible, and the pattern detection that AI provides across hundreds of calls per week produces coaching insights that no manager could generate manually. The per-rep licence cost for tools like Gong ranges from approximately 1,400 to 1,800 GBP per rep per year, which needs to be justified by the coaching improvement and deal outcome improvement it enables.