Sales Discovery Call Prep
Before any prospect call: research their company, recent news, key people, and likely pain points — pulled into a 1-page pre-call brief.
Specializedmedium~45m setup
- Tools
web_fetchmemorymemory_searchdelegate- Channels
telegramdesktop- Uses
subagents
Sales discovery calls live or die in the first 5 minutes — when the prospect realises whether you've actually thought about their business. Showing up with research-backed context (recent news, pain points relevant to your product, who's likely on the call) shifts the call from sales pitch to consultation.
What it does
- Given a prospect company name + a meeting time
- Spawns 3 parallel subagents to research:
- Company basics (what they do, size, stage, recent news)
- Key people on the call (LinkedIn, their published work)
- Likely pain points relevant to YOUR product (configured once)
- Produces a 1-page brief 30 minutes before the call
- Logs research in memory so follow-ups can build on it
What you'll need
- Web fetch for company sites, press releases, LinkedIn (public)
- Delegate for parallel research subagents
- Memory for prospect tracking
- A configured "your product profile" — what you sell, what pain points you address
Setup
1. Define your product profile
Send to Flowly
Remember (tag "product-profile"):
What we sell: Flowly — personal AI agent platform for solo
operators and small teams.
Pain we address:
- Switching cost between AI tools (memory locked in vendor X)
- Privacy/sovereignty (sensitive data going to cloud)
- Scaling personal workflows beyond a single founder
Our typical customer: 5-50 person teams, technical founder, has
specific data they don't want in OpenAI's cloud.
Disqualifiers: pure consumer plays, regulated finance/health
(separate compliance product), enterprise > 500 (we're not there
yet).
2. Schedule the prep workflow
Send to Flowly
When I say "prep <company> call <date> <time>":
(or, automatically: if a calendar event tagged "discovery" is < 24h
out and we haven't prepped yet)
Spawn 3 subagents in parallel:
1. Company researcher:
"Research <company>. Return:
- Bio (2 paragraphs: what they do, customer, scale)
- Founded year, team size, funding stage
- Recent news (last 90 days): launches, hires, funding, strategic shifts
- Tech stack signals (job postings, GitHub, public stack pages)
Cite sources."
2. People researcher:
"Identify who's likely on the call. From the calendar invite, the attendees
are <list>. For each, research:
- Title and role
- LinkedIn highlights (companies, key projects)
- Public writing or talks (blog, X/Twitter, conference talks)
- One-sentence point of view they're known for
Use web_fetch for LinkedIn public profiles, X bio pages."
3. Pain-point matcher:
"Given my product-profile and what we know about <company>, hypothesise:
- 3 ways our product could help them based on their public stance
- 2 reasons they might NOT be a fit (be honest — disqualifying questions
to ask early)
- Likely objection from someone in their position
Cite product-profile + the company research as your evidence."
After all 3 finish, spawn a 4th synthesiser:
4. Brief writer:
"Write a 1-page pre-call brief in this exact structure:
## Company in 30 seconds
2 paragraphs.
## Recent moves (90 days)
3-5 bullets.
## Who's on the call
Per attendee: name, title, key insight.
## Why this might be a fit
3 bullets connecting their context to our product-profile.
## What might not work
2 bullets — honest disqualifiers.
## Three questions to ask
- Open-ended question about their context
- Specific question rooted in recent news
- Disqualifying question that surfaces the not-fit risk early
## One-line opener
How to start the call to demonstrate research without showing off."
5. Save full brief to memory "prospect:<company-slug>:<date>"
6. Send to Telegram 30 min before call (or now if I asked manually)
3. Post-call note capture
After the call, a quick note for next time:
"Note for OpenAI prospect: timing not right, evaluating Q3. Decision-makers Sarah and Mike. Concern: data residency. Follow up Sept 1."
Tagged prospect:openai:notes. Next time you prep for them (or a
similar prospect), this surfaces.
4. Aggregate prospect view
"Show me all prospects in 'evaluating' status with follow-up due in next 14 days"
Memory search returns the list with last interaction context.
Tips
- Honesty in the brief. The "what might not work" section is what differentiates a good brief from a wishful one. If they're obviously not a fit, the brief should say so — saving everyone the call.
- Don't recite the brief on the call. Read it before, internalise, let it shape questions. Reciting research is creepy.
- One specific question is worth 5 generic ones. "I saw your CTO posted about X last month — how's that progressing?" beats "tell me about your roadmap."
- Update product-profile quarterly. As your product evolves, the matcher needs current context. Stale profile = stale matches.
- Time-box research. 30-minute brief generation is right. If a prospect needs 2 hours of research, they're either huge enterprise (different process) or you're over-investing.
- Privacy: prospect data is sensitive. Don't share briefs with others on your team unless they're in the deal. Memory stays per-user by default.