Multi-agent call prep (July 2026)
Intro and 2 caveats
This is a set of prompts to use for call preparation.
Caveat 1: Good, but not great. This represents a decent general approach, though not as nuanced as I'm sure you could get when tying it heavily to the specific business logic of a company that fully understands its ICP and buying signals. It's meant more as a look into what a decent baseline looks like than an assertion that nothing better is out there.
Caveat 2: Optimized for a founder with 10+ calls/day. The output format matches what a very busy founder with two minutes to skim before a call asked for, but is much less thorough than I would use as an AE who's solely devoted to maximizing the value of each deal.
Structure and overall output prompt
The following output prompt takes as input the outputs of four child prompts. I run the system with a mix of Python and Claude because Claude Skills, as of July 2026, do not handle the "run X prompts first, then when they are done, run Y" workflow gracefully; other systems like Dust.tt do subagents more gracefully. I have found that more narrowly scoped information-gathering prompts produce better output for cheaper than attempting to do everything in one giant prompt with a massive context window. As with all things AI, it's likely this is adorably primitive within a few months.
Output prompt
Goal is prep {NAME}, CEO of {COMPANY}, for a first prospect meeting. Synthesize the attached text, generated by LLM using several queries (the “Q:” after each QUERY HEADER heading) into a format meeting the following parameters:
Most importantly:
- Total output should be ¾ of a page or less, to promote quick reading
- Be clear about areas of uncertainty and areas of certainty
Format:
- Opening goals section
- Exactly 1 clear goal for {NAME}, CEO of {COMPANY}, to improve his personal selling style on this particular call
- Up to 3 goals, most important first, for the outcome of the call. First goal is primary objective, but if that becomes unfeasible (example: trying to sell a product they don’t need), fall back to second goal
- People in call
- For the known prospect(s) in call, a short background. If there are no additional people mentioned aside from whomever the cold call was with, assume they are the only prospect
- Call agenda
- Intro: 1-2 possible areas to open discussion that will create a personal bond or show we did research. Prioritize areas that prospect is likely to respond well to, or that highlight background overlap between prospect and {NAME}
- What we know: Generate a 2-5 sentence summary of what we think we know about the company to share early in the call and solicit feedback to make sure our understanding is correct. Be very clear where we’re certain (clear public source, or direct statement from prospect in call transcript summary) vs. where we don’t know
- At least 3 insightful questions to ask during discovery to help uncover priorities related to strategic sourcing, ingredient quality or delivery reliability, or overall business direction
- Ideal next steps, linked to goals
- Optional, if space remaining: Any other relevant details to keep in mind not captured by the above
Input prompts to the above output
Input prompt 1: Ingredient sourcing
Q:
“For the following URL, please guess at each and provide rationale as to why:
1. What is the rough order of magnitude of ingredients they likely have to source (tens, hundreds, thousands)? To assess, look at the quantity of products they sell, ingredients in each (don't need to check all for an order of magnitude guess), and likelihood they are sourcing ingredients themselves vs. whitelabeling completed products?
2. Given the products they sell and estimated quantity of what's most popular, which ingredients would they likely be spending the most money on?Input prompt 2: Prior communications
Q:
What if anything is relevant to know to influence a future sales call given the existing set of communications and correspondence so far? Use connected call transcripts (Nooks, Granola) and emails (HubSpot), focusing on what the prospect said vs. what the {COMPANY} representative said. Prioritize any "Key themes" called out in the {COMPANY} representative internal notes (Drive).Input prompt 3: Import records
Q:
What can we learn from {COMPANY} public import records to help understand what's likely important to their supply chain and sourcing strategy?