Sales calls contain valuable information about customer needs, objections, budget signals, decision criteria, and buying intent. However, much of this information can remain inside long call recordings that sales teams do not have time to review in full.
This use case explores how a connected AI automation workflow can capture completed sales calls, analyse the transcript, extract key insights, generate a structured report, and route the information into the next stage of the sales process.
The Problem
Sales calls are packed with signals, including pain points, objections, budget hints, and buying intent, but it all lives inside an hour-long recording that no one has time to re-listen to.
By the time someone gets around to writing up notes, half the nuance can be lost. There is also no consistent way to compare deal quality across sales representatives, identify coaching opportunities, or understand what happened during the conversation.
The Solution
The Call Insights Agent (Clark) operates immediately after the sales call itself. It receives completed calls from connected transcription platforms, normalises the transcript, and runs it through a structured AI analysis.
The workflow extracts pain points, objections, decision criteria, BANT information, competitive intelligence, customer sentiment, next steps, and sales-coaching feedback. It then sends a formatted report to the sales team, logs metrics for each call, and passes the analysis into a downstream interest-routing workflow that supports proposal generation and follow-up.
How the Agent Works
- Receive completed sales calls. The workflow receives call data from connected transcription platforms after a meeting has ended.
- Prepare the transcript for analysis. Call transcripts from different sources are standardised into one consistent format containing the meeting title, participants, duration, and transcript content.
- Extract sales intelligence. AI reviews the conversation to identify pain points, objections, decision criteria, buying signals, BANT details, competitor information, customer sentiment, and next steps.
- Create a structured sales report. The workflow turns the analysis into a formatted report that gives sales teams a clear view of the opportunity and possible follow-up actions.
- Support coaching and pipeline movement. Insights are logged for reporting, shared with the sales team, and routed into the next workflow for interest routing, proposal generation, or follow-up.
Technical Workflow
1. Multi-platform intake
A single webhook (/call-insights) receives events from two different call-recording tools. A code node detects which platform sent the payload, Fireflies.ai or Read.ai, and tags the payload accordingly. A Switch node then routes down the matching path.
2. Fetch and normalize the transcript
For Fireflies, the workflow calls the Fireflies API directly to pull the full transcript by meeting ID. Read.ai already sends the full transcript in the webhook body. Either way, a normalization code node reconciles the two very different data shapes into one consistent format: meeting title, participant list, duration, and a clean transcript string.
3. Extract the client’s email
A small code node filters the participant list against a known internal email address, isolating the external/client email(s) for later use.
4. Deep AI extraction (“Clark”)
The full transcript is sent to an LLM (GPT-5.2) running a comprehensive analyst prompt, persona-named Clark. It extracts, with mandatory evidence/quotes and no fabrication: pain points (severity-rated), objections (strength-rated, each with a suggested response), decision criteria, budget and timeline signals, next steps and buying intent, a full BANT breakdown, sales-coaching opportunities, competitor categories/mentions/comparisons, customer sentiment, and knowledge gaps, all as one strict JSON object.
5. Parse and enrich
A code node safely parses the AI’s JSON, then computes summary stats: total pain points/objections/criteria, high-severity/strong/critical counts, a BANT completeness score (0–4), competitor and coaching signal counts, and boolean flags like hasHighPriorityItems and hasBudgetInfo.
6. Formatted email report
Another code node builds a full, styled HTML email, Apple-style design, color-coded severity badges, a gradient executive-summary card, a BANT grid, and dedicated sections for pain points, objections, decision criteria, coaching opportunities, and competitive/customer intelligence, sent via Gmail to the sales team.
7. Metrics logging per call
In parallel with the email, a code node computes a compact KPI row (date, call ID, platform, total insights extracted, high-priority signal flag, BANT score, competitive intel detected, coaching opportunities identified) and appends it to a dedicated Google Sheet.
8. Handoff to the interest router
After the email sends, a code node POSTs a markdown-formatted analysis, along with the extracted client email, to a downstream webhook. This is the exact input the “Call Insight Interest Router” workflow (feeding the Proposal Generator) expects.
Technology and Integrations
Built with: n8n, Fireflies.ai, Read.ai, OpenAI, Gmail, Google Sheets, webhooks, and connected sales workflows.
Outcome
Every sales call gets automatically turned into a structured, evidence-backed intelligence report, with coaching feedback for the sales representative, competitive context, and a clean handoff into the proposal pipeline, without anyone manually writing up notes.
Explore Custom AI Automation for Your Business
The Call Insights Agent is one example of how AI automation can support a more structured sales-analysis and follow-up process.
We design and build custom AI agents, automation workflows, internal tools, and connected systems for a wide range of business processes.
From sales, marketing, recruitment, customer support, and reporting to finance, operations, and internal team workflows, each solution is designed around the way your business works.
Request a Quote for a Custom AI Automation Solution to get started.