The hidden gold in sales calls

Every sales call contains information that no other channel provides. The prospect tells you exactly what concerns them, which competitors they are evaluating, what budget they have, which objections hold them back and which benefits matter to them. But that information is lost because nobody records it systematically.

An average seller makes 15-20 calls per week. If each call lasts 30 minutes, that is 7-10 hours of conversation per week. Of those hours, the seller remembers (and logs in the CRM) perhaps 20%. The other 80% is lost: nuances, exact phrases, subtle buying signals, competitor names mentioned in passing.

Conversational AI captures 100%. It transcribes every word, analyzes patterns, identifies buying signals and objections, and generates actionable insights for the seller and their manager.

The main use cases:

• Coaching: the manager can review calls and give specific feedback based on data, not impressions.
• Onboarding: new sellers can study top performers' calls to learn what works.
• Competitive intelligence: every mention of a competitor is recorded and aggregated.
• Product improvement: complaints and feature requests are captured automatically.
• Forecast: analyzing the prospect's language ("this is exactly what we need" vs "we will evaluate it internally") predicts close probability.

Automatic call transcription

The foundation of conversational analysis is automatic transcription. Modern tools transcribe with over 95% accuracy, identify each participant (speaker diarization) and generate automatic summaries.

Types of transcription

• Real-time: the transcription appears while the call is in progress. Useful for the seller to take notes and search for information during the conversation.
• Post-call: the transcription is generated after the call ends. It is the most common and sufficient for analysis.
• With integrated analysis: besides transcribing, the tool automatically analyzes sentiment, topics discussed, questions asked and action items.

Privacy considerations

Recording calls has legal implications that vary by jurisdiction:

• EU (GDPR): you need explicit consent to record. Informing at the start of the call is mandatory.
• One-party vs two-party consent: in some countries it is enough that one party (you) knows it is being recorded. In others, both parties must know.
• Storage: recordings contain personal data. GDPR rules on retention and access apply.

Best practice: at the start of each call, inform: "We are going to record this meeting to generate an automatic summary. I will share it with you afterwards. Is that OK?". Most prospects accept without issue.

Gong: the revenue intelligence platform

Gong is the market leader in conversation intelligence. It records calls (Zoom, Meet, Teams, phone), transcribes them, analyzes them and generates insights at the deal, seller and team level.

Main Gong features

• AI transcription: identifies participants, topics, questions and answers.
• Deal Intelligence: analyzes all calls for a deal and predicts risk. "This deal has a 35% probability of being lost because the prospect mentioned 'we need board approval' and no follow-up has been scheduled."
• Coaching cards: generates coaching cards for each call with metrics (talk ratio, questions asked, objections handled) and key moments to review.
• Competitive intelligence: aggregates all competitor mentions across all team calls. "In the last month, Competitor X was mentioned in 23 calls. The main arguments used against us are..."
• Trackers: configure keywords and the AI alerts when they are mentioned in a call. Example: "budget", "timeline", "Competitor X", "discount".
• Ask Gong: ask questions in natural language about your calls: "in which calls this week did the prospect mention problems with their current CRM?".

Pricing: Gong does not publish prices. Typically from 1,000-1,500 USD/user/year for teams of 10+. It is a premium tool for teams whose revenue justifies the investment.

Fathom: the affordable alternative

Fathom is an alternative to Gong with a different approach: free for individual use, with AI features that cover 70-80% of what you need.

Fathom features

• Recording and transcription: compatible with Zoom, Meet and Teams. Automatic transcription with participant identification.
• AI summaries: generates a structured summary of each call with key points, decisions made and action items.
• Highlights: you can mark key moments during the call with a click. Useful for going back to important sections without reviewing the entire recording.
• CRM integration: syncs summaries and notes with HubSpot, Salesforce and other CRMs. The summary automatically appears in the deal record.
• Search: search all your calls by keyword or topic. "Find all calls where 'budget 2026' was mentioned".

Pricing: free for individual use (unlimited). Team plan from 32 USD/user/month (with advanced analytics and CRM sync).

Gong vs Fathom

• Fathom if you are a team of 1-5 sellers, need recording + summary + CRM sync, and want to start for free.
• Gong if you are a team of 10+ sellers, need advanced analytics, systematic coaching and aggregated competitive intelligence.

AI sales coaching

Traditional sales coaching is based on the manager listening to a call (or part of it), giving subjective feedback and hoping the seller improves. The problem: it is inconsistent, depends on the manager's experience and does not scale.

AI coaching is systematic, data-driven and scalable.

Coaching metrics that AI can measure

• Questions asked: how many open questions did the seller ask? Open questions generate more information than closed ones.
• Active listening: does the seller interrupt the prospect? Does the seller allow pauses for the prospect to elaborate?
• Objection handling: when the prospect raises an objection, does the seller acknowledge it, reframe it and respond? Or ignore it?
• Clear next step: does the call end with a concrete next step (meeting date, deliverable) or a vague "let's talk soon"?
• Discovery vs pitch: in the first call, does the seller spend more time asking than presenting? The ideal ratio is 60% listening, 40% talking.

AI coaching using transcripts

If you use Fathom or another transcription tool, you can do AI coaching by passing the transcript to Claude or ChatGPT:

"Here is the transcript of a sales call [paste]. Analyze the seller's performance (Speaker 1) on these criteria: 1) Question quality (open vs closed, relevance). 2) Objection handling (did they identify them? acknowledge them? resolve them?). 3) Talk-to-listen ratio (percentage of time talking vs listening). 4) Next steps (was a concrete next step agreed?). 5) Buying signals (were there any? did the seller identify them?). For each criterion: score 1-5, concrete evidence from the transcript, improvement suggestion."

This analysis, which a manager would take 30 minutes to do manually per call, AI completes in 2 minutes with greater granularity.

Objection pattern analysis

Objections are not random. There are patterns. The same 5-10 objections appear in 80% of calls. If you identify them and prepare responses, your conversion rate improves.

Common B2B objections

• Price: "It's expensive", "We don't have the budget", "Competitor X is cheaper".
• Timing: "Not the right time", "We're focused on another priority", "We'll revisit in Q3".
• Authority: "I need to consult with my boss/board", "I'm not the decision maker".
• Need: "We already have a solution", "We're not sure we need this".
• Trust: "I don't know your company", "You're too small", "You don't have references in our sector".

AI objection analysis

If you have transcripts from your last 50 calls, AI can do a pattern analysis:

"Here are transcripts from 50 sales calls. Identify all objections raised by prospects. Group by category (price, timing, authority, need, trust, other). For each category: frequency (how many calls it appears in), most common exact phrases, how the seller typically responds, success rate of each response type (is the objection resolved or is the deal lost?). Generate an 'objection playbook' with the best response for each."

Objection playbook

An objection playbook is a document the entire sales team uses. For each common objection, it includes:

1. The exact objection (as the prospect says it).
2. What it really means ("we don't have budget" sometimes means "I don't see the value").
3. The recommended response (tested and effective).
4. An example of a successful conversation (extract from a real call).

AI can generate the complete playbook from your transcripts. The manager reviews it, adjusts it and shares it with the team.

Talk-to-listen ratio and conversational metrics

The talk-to-listen ratio measures how much the seller talks versus how much they listen. It is the most important conversational metric and the easiest to improve.

Benchmarks

• Discovery call: the seller should talk 30-40% of the time. The rest, listen. If they talk more than 50%, they are doing a pitch, not a discovery.
• Demo: the seller talks 50-60%. They are presenting, but must leave room for questions.
• Negotiation: 50-50. Listening is critical to understanding the prospect's position.
• Monologues: no monologue should last more than 2 minutes without a pause or question. Long monologues lose the prospect's attention.

Other conversational metrics

• Number of questions: top performers ask 11-14 questions per discovery call. Fewer than 7 indicates premature pitching.
• Speaking speed: speaking too fast generates anxiety and hinders comprehension. The ideal range is 130-170 words per minute.
• Filler words: "uhh", "umm", "basically", "you know". AI counts them. Reducing them improves the perception of confidence.
• Prospect sentiment: AI analyzes whether the prospect's tone is positive, neutral or negative throughout the call. A shift from positive to negative indicates a problem the seller should address.
• Next steps: percentage of calls that end with a concrete next step (date and time). Calls without a next step have 80% less probability of advancing.

Competitive intelligence from calls

Prospects mention competitors in sales calls. Sometimes directly ("we are also evaluating X"), sometimes indirectly ("a tool we use does something similar"). This information is gold for marketing, product and strategy.

What to capture

• Competitors mentioned: which ones and how often.
• Arguments in favor of the competitor: "X is cheaper", "X has better integration with Y".
• Arguments against the competitor: "X is hard to use", "X doesn't have support in our language".
• Win/loss by competitor: when we compete against X, we win 40% of the time. When we compete against Y, we win 70%.
• Features requested: "what we really need is for it to do Z". If Z does not exist in your product, it is input for the roadmap.

Competitive analysis with AI

"Analyze these 30 sales call transcripts. Identify all competitor mentions. For each competitor: number of mentions, context (active evaluation vs past reference), arguments the prospect uses in their favor, arguments against, deal outcome (won/lost/in progress). Generate a battle card for the 3 most mentioned competitors with: their strengths according to prospects, their weaknesses according to prospects and the arguments our team should use against each."

Battle cards generated with real call data are 10 times more useful than those generated by marketing without prospect contact.

Practical implementation

Step 1: Choose a tool

To start: Fathom (free). For teams of 10+: evaluate Gong or Chorus. For ad hoc analysis: transcribe with any tool and analyze with Claude/ChatGPT.

Step 2: Configure

• Connect with your video call tool (Zoom, Meet, Teams).
• Connect with your CRM to sync summaries.
• Configure keyword trackers: competitor names, "budget", "timeline", "decision".

Step 3: Team adoption

The main resistance: "I don't want to be recorded". The solution: show the value. Share an example of an automatic summary, data-driven coaching and time saved on notes. When a seller sees they no longer have to write notes after each call, they adopt the tool out of convenience.

Step 4: Weekly review

Dedicate 30 minutes a week to reviewing aggregated insights: this week's objections, competitors mentioned, deals at risk based on call analysis. Share findings with the team.

Ejercicio practico

Puntos clave