NOTE: Demo visuals use either blurred real data or synthetic placeholders to protect customer privacy.

Customer-facing teams spend hours every day in nonstop meetings. Between demos, technical deep dives, and renewal conversations, sales engineers (SEs) are constantly fielding questions.

And inevitably, some of those questions get postponed. But with calendars packed end-to-end, follow-ups slip through the cracks. The result? Delayed answers, slower deals, and unnecessary customer friction.

To solve this, Andy Chen expanded the AI Question Bot into Season 2: Meeting Question Bot: a system that listens to Gong calls, identifies unanswered customer questions, and drafts AI-generated responses for SEs automatically.

Even with the original AI Question Bot answering 30–40% of GTM Help questions in Slack, a fundamental gap remained. Customer questions asked in meetings were still going untracked.

Sales engineers face a few consistent challenges:

• Back-to-back meetings: No time to review call notes or recall everything said.

• High cognitive load: Dozens of details per account, across multiple products.

• No single source of truth: SEs juggle Slack, Salesforce, Gong, email threads, and internal docs.

• Follow-up risk: A forgotten answer can frustrate a prospect or slow a deal.

The solution was clear: Bring AI to the meeting itself, the place where questions originate.

With the core Question Bot already highly accurate, Andy extended it to analyze meeting transcripts directly through Gong.

Here’s what the new system does:

1. Scan every new Gong call for customer questions.

2. Detect unanswered or deferred questions using semantic patterns like:

   • “Let me get back to you.”

   • “I’m not 100% sure.”

   • “I’ll check with engineering.”

3. Auto-generate a concise answer to each question using Abnormal’s knowledge base.

4. Include sources so SEs can verify or expand on the answer.

5. Send a follow-up email to the SE, eliminating the mental bookkeeping of remembering open loops.
   

In Andy’s demo, the bot scanned a real customer call and sent an email to the SE containing:

• The original customer question

• A polished AI-generated answer

• Links to internal docs and validated sources
  

The SE didn’t need to re-listen to the call, take notes, or check with product. The bot handled everything.

Even in early testing, SEs immediately saw value:

• Reduced cognitive load: No need to mentally track unanswered questions.

• Faster follow-ups: Customers receive answers quickly, strengthening trust.

• Fewer internal pings: SEs no longer need to chase product or engineering for common questions.

• Increased adoption: SEs seeing the results asked for the bot to be enabled for all their calls.
  

One SE even asked for the bot draft the entire follow-up email. That’s already on the roadmap. This is just the beginning.

Once refined, the Meeting Question Bot can support account executives (AEs), customer success managers (CSMs), renewal teams, technical onboarding teams.

Anywhere customer questions appear, the bot can help answer them.

Season 1 of AI Question Bot transformed GTM Help by reducing reliance on product experts. Season 2 brings that intelligence into real customer conversations, enabling:

• Call-level semantic understanding

• Real-time identification of follow-up dependencies

• Automated knowledge retrieval

• Hands-free follow-up support
  

It’s an entirely new way of supporting GTM, where AI listens to calls and helps reps close every loop automatically.

Andy’s Meeting Follow-Up Bot is a standout example of creative AI application: taking an existing tool, expanding its scope, and solving a root-layer workflow problem that impacts deals every day.

It’s a perfect reflection of Abnormal’s innovation culture, where teams think bigger, build boldly, and use AI not just to answer questions, but to transform how customer interactions happen.

Season 2 is just the start.

As Andy continues iterating, we may soon enter a world where every customer question, no matter where it’s asked, gets answered automatically.