Technical Sales Enablement Studio

Field engineering at an AI company has a peculiar economics problem: every discovery call is a unique research project. You're walking into a Fortune 500 with a specific stack, specific objections, specific competitive context, and you have one hour to be the most prepared person in the room. The traditional way to prep takes four hours. You do four calls a week. You do the math.

This demo is the workflow that solved it for one team — built in a weekend, deployed across four AEs, now generates the prep brief for every call automatically.

Problem

• Account executive forwards a discovery call invite to the field engineer 24 hours before the call.
• Field engineer needs: industry context, the customer's known architecture (from public sources + prior call notes), likely objections, competitive landscape, and three concrete demo paths tailored to the suspected use case.
• Average prep time: 4 hours, mostly Googling and re-reading old call transcripts.
• Recurring problem: the most senior FEs were burning out on prep; the most junior ones were under-prepping and learning the hard way.

Traditional workflow

1. FE reads the AE's notes from the prospecting calls. Half are missing.
2. FE searches the customer's engineering blog, recent press, and LinkedIn for the technical decision-makers.
3. FE searches internal Slack for "who's talked to <Customer> before?" and DMs three people.
4. FE digs through the competitive battle cards in Confluence. The page was last updated 8 months ago.
5. FE writes a one-page brief. Most of it gets thrown out within five minutes of the actual call as new info emerges.

The brief is good when it's done. The path to get there is unreasonable.

AI-native workflow

The structural change: a single workflow that ingests the same five sources and produces a living brief — short enough to be readable, structured enough to update in real time during the call.

1. Step 1InputsAE notes + customer + industry
2. Step 2ResearchAgent pulls public signals
3. Step 3SynthesizeBrief in our team's format
4. Step 4PersonalizeFE adds 2-3 known facts
5. Step 5Run callBrief stays open, updates live

The brief always has the same five sections, in the same order:

1. Who's in the room — names, roles, what they shipped recently, what they probably care about.
2. Their stack as we know it — from job postings, engineering blog, GitHub presence, prior calls.
3. Likely objections — ranked by probability for this specific company shape.
4. Three demo paths — short, ranked by likelihood of fit. Each one has a "if they say X, pivot to Y" line.
5. What we don't know — explicit gaps the FE should ask about in the first ten minutes.

The fifth section is the one that changed outcomes most. Naming the unknowns up front forces the conversation toward signal-rich questions instead of demo theater.

Technical breakdown

• A Vercel-hosted internal tool with a single form: customer name, AE notes, industry, optional URLs to dig into.
• Behind it, an agent run with three tools: web fetch, the company's internal call-notes search, and the competitive intel database.
• Output is structured: a Zod schema with the five sections. The agent runs once for the bulk content, then a Haiku pass tightens each section for skim-readability.
• Result is rendered as a one-page Markdown brief, kept open in a side panel during the call.

Operational impact

• Prep time: 4 hours → 30 minutes (the 30 includes 10 of human polish + 20 of "let me actually read this thoroughly before the call.")
• AE-to-FE handoff drop-off (calls where the FE was visibly under-prepared): roughly 60% reduction.
• Senior FE retention conversation: one less complaint.
• Junior FE ramp time: an order of magnitude shorter, because the brief format itself teaches the discovery-call mental model.

Lessons learned

• Structure the output, not the input. AE notes are messy and inconsistent. Don't try to make them clean. Let the agent normalize. Force the output into a schema instead.
• Optimize for the format that gets used in the room. A brilliant 10-page brief that nobody reads during the call is worth less than an OK 1-page brief that stays open in a side panel.
• Name the unknowns. Putting "what we don't know" into the brief shifted call quality more than any other change. It turns the first ten minutes from demo-theater into real discovery.
• The senior FEs are the prompt engineers. Their version of the brief is what the agent learns from. Their brain is the moat; the tool is the leverage.