Stop Winging Discovery Calls. Read the Room Before You're In It.
The Departure
You're at gate B14. Your call is in 22 minutes. You haven't looked at the prospect's website since the SDR booked the meeting two weeks ago.
You'll wing it. We all wing it. And then we wonder why the call meanders into "tell me about your company" and never lands on a next step.
I've been there, I've literally pulled up a LinkedIn profile in the boarding line. Last week Anthropic quietly handed every Pro subscriber a fix for that exact moment. Most reps still don't know they have it.
The Co-Pilot
Tool: Claude Opus 4.7 with vision (shipped April 16 -- 3x resolution upgrade, included in Claude Pro)
The Use Case: Turn a screenshot of a prospect's homepage, pricing page, LinkedIn, or recent press release into a structured pre-call brief in under 4 minutes; on your phone, at the gate, between meetings.
The Prompt: "You're my pre-call analyst. I've attached a screenshot of [prospect's homepage / pricing page / LinkedIn / press release].
In under 200 words, give me:
What they're actually selling -- in plain language, stripped of marketing copy.
Three things they emphasize that I should mirror or address on the call.
One visible gap or weakness -- something they're not saying, or saying poorly.
Two questions to ask that show I read their site, not just their LinkedIn headline.
Be direct. No hedging. If something is unclear from the image, say so."
That's the whole play. Screenshot. Paste. Read on the jet bridge.
The trick most people miss: don't stop at one image. Take a second screenshot of their pricing or careers page and run it again. The careers page is where companies tell you what they're actually struggling with, they just dress it up as a job description.
The Upgrade
Topic: Why this lands now -- and which model to use for what
The freshness window is tight. Opus 4.7 vision launched April 16. By mid-May, every sales enablement vendor will be charging $40/seat/month for "AI prospect intelligence" that does exactly this. You already pay for it.
A quick framework so you stop guessing which model to open:
Use Claude Opus 4.7 for vision tasks. Screenshots, pricing pages, contracts you photographed, whiteboard pictures from a workshop. The 3x resolution bump means it actually reads small text on a pricing tile, the older models guessed.
Use GPT-5.5 for long-form drafting and research synthesis. Multi-source reports, proposal first drafts, anything where you're feeding it 10+ pages of context.
Use either for the briefing file we built in #012-#014. Both handle it well. Pick the one already open.
One housekeeping note for the calendar: ChatGPT Workspace Agents free preview ends May 6. If you've been meaning to test it on a recurring task, weekly pipeline review, expense categorization, inbox triage; this is the week. After May 6 it moves behind the Team plan.
That's the whole news cycle worth your attention. The Musk-OpenAI trial, the Adobe integration, the DeepSeek pricing war, skip them. None of it changes what you do Monday morning.
The Landing
Your task today: Open your calendar. Find the next external call you've got booked; discovery, renewal, intro, doesn't matter.
Pull up the prospect's homepage. Take one screenshot. Paste it into Claude with the prompt above.
Read the brief before you join the call.
If it takes more than 4 minutes, you did it wrong. If it changes one question you ask on that call, you did it right.
That's it. No new tool to install. No subscription to add. The pipeline you already have, worked harder.
Safe travels,
Marcellus