How We Used ChatGPT Voice Mode to Fix Sales Retros (and Actually Get Insights)

TL;DR

• Sales retros are usually a slog—either endless meetings or writing nobody wants to do.
• We ditched both: reps had 15-minute voice convos with ChatGPT, guided by a structured prompt.
• The result: richer insights, zero meetings, and a process sellers didn't hate.
• Full sanitized prompt included below.

---

Why Sales Retros Suck (and Why We Needed a Change)

Personally, I'm a big fan of doing retros. They often help us surface insights that I wouldn't necessarily have uncovered otherwise. However, I've also witnessed over the years how the most vocal people in a org can take over these sessions. I tend to do retros in my 1 on 1s and quarterly reviews but I find that people can be distracted and not really engaged in the retro.

So as Q1 2025 wrapped, our sales team braced for the usual "retro" routine: long meetings with fuzzy takeaways, some complaining, and endless forms nobody wanted to fill out. Sound familiar?

Retros are supposed to make us better at GTM execution. But in reality:

• Synchronous retros = time sink, unclear action items
• Async retros = writing fatigue, low engagement

We needed a way to get real insights—without the pain.

---

Operationalizing AI: The Voice-Mode Retro

This quarter, we tried something new. Instead of asking to populate docs or join meetings, I asked every rep to have a 15-minute conversation with ChatGPT using the Advanced Voice Mode.

• No forms, no typing -> just talk.
• The AI asked follow-ups, probed for specifics, and summarized key takeaways.
• The whole conversation was guided by a prompt I designed (see below).

The result? Sellers actually engaged. I got richer, more honest feedback—without dragging everyone into a 90-minute Zoom.

---

Workflow Redesign: From Meetings to AI Conversations

Here's how it worked:

1. Each AE got a calendar invite with a link to ChatGPT and the prompt.
2. They hit "record" and talked through the retro, one question at a time.
3. ChatGPT handled the rest—asking for details, clarifying, and summarizing.

What changed:

• No more groupthink or wasted time.
• Reps could be candid (AI doesn't judge).
• We got structured, actionable data—fast.

---

Prompt Engineering: The Secret Sauce

The real unlock was the prompt. Here's a sanitized version you can steal:

md
# 📊 Quarterly Sales Retro – ChatGPT Prompt
You are a Sales Strategist and Deal Coach helping me run a Q1 2025 retro for my company's sales motion.

## 𝗖𝗼𝗻𝘁𝗲𝘅𝘁: 
We're a X company selling to X companies. Our primary buyer is X, and our product is a X that helps X do X.

### 𝗜𝗻𝘀𝘁𝗿𝘂𝗰𝘁𝗶𝗼𝗻𝘀:
Please ask the questions 1 by 1. 
Ask follow-up questions if you determine the answer doesn't provide enough context 
Ask for specifics, examples... The goal is to really dive deep in this retro. 
Do not offer suggestions or interpretations, your goal is solely to ask questions and to gather information.

### More Information
𝗜𝗻 𝗤𝟭:
• We had XX AEs, mostly working inbound, with outbound and partner motions contributing XX% to their pipeline.
• X opportunities were created in Q1.
• X deal closed-won from Q1 pipeline, and X additional deals closed from Q4 pipeline.
• X deals closed-lost, with common reasons including low urgency, timing pushed to H2, internal bandwidth shifts, and weak champion alignment.
• Average sales cycle was XX days.
• We started Q1 with good pipeline coverage (X weighted pipeline to target ARR)

## Expected Output
𝗢𝘂𝗿 𝘁𝘄𝗼 𝗺𝗮𝗶𝗻 𝗼𝗯𝗷𝗲𝗰𝘁𝗶𝘃𝗲𝘀 𝗳𝗼𝗿 𝘁𝗵𝗶𝘀 𝗿𝗲𝘁𝗿𝗼 𝗮𝗿𝗲:
1. Identify ways to improve execution across discovery, deal strategy, and closing.
2. Uncover how we can better position our value and urgency to close more deals.

## Guidance
𝗣𝗹𝗲𝗮𝘀𝗲 𝘄𝗮𝗹𝗸 𝗺𝗲 𝘁𝗵𝗿𝗼𝘂𝗴𝗵 𝗮 𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲𝗱 𝗤𝟭 𝗿𝗲𝘁𝗿𝗼, 𝘀𝘁𝗮𝗿𝘁𝗶𝗻𝗴 𝘄𝗶𝘁𝗵 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻𝘀 𝗹𝗶𝗸𝗲:
• What worked well in Q1?
• Where did deals stall or fall through?
• Were we selling to the right people? Was there clear pain?
• Did we generate enough urgency?
• Were our proposals and pricing clear and compelling?
• What can we change to improve win rate in Q2?

At the end, summarize the key themes, wins, gaps, and 2–3 priority changes we should implement in Q2.
Include the full Q&A.

---

Case Study: What Actually Changed at MadKudu

• Richer insights: The AI surfaced patterns we'd missed in group settings.
• No meetings: Sellers didn't lose hours to Zoom.
• Frictionless: The process felt natural—just talk, and you're done.

What we'll tweak for Q2:

• Add a step for reps to review and edit the AI's summary before submission.
• Experiment with different prompt phrasings to get even more specific feedback.

---

Takeaways

• AI voice mode retros are a game-changer for sales teams who hate meetings and paperwork.
• The right prompt = better data, less friction, and happier sellers.
• If you want to try it, steal the prompt above and let me know how it goes.