A sample Advanced Brief

Turn one call recording into a scored, next-step deal review

You have the call. Most people summarize it and stop. Run it through this and you get a MEDDPICC score, the real objection, and a follow-up that lands.

Capture. Pull the raw transcript from Gong or Fathom. Paste the full text, not the auto-summary. The auto-summary strips the exact phrasing you need for scoring and objection work.

Score.

Diagnose.

Build.

Tighten. If the output is generic, add: 'You are scoring this at a 15-person Series B whose CFO just froze new spend. Weight budget and paper process harder, and assume the champion has no real authority until proven.' Context about company stage and budget climate forces sharper scores.

Ten minutes turns a call you half-remember into a scored deal with a written next move. Do it the same day, while the language is still exact.

A discovery-prep chain you run before every first call

Save this as a Claude Project called 'Discovery Prep' with your ICP, your product one-pager, and three of your best past discovery transcripts loaded as context. Then run these four prompts in order.

Research the account.

Map the buying group.

Write the questions.

Pre-mortem the call.

The project holds your ICP and your best past calls, so every run is grounded in how discovery actually works for your product, not a generic template.

Wire Gong to your CRM so scored calls land in the deal automatically

Stop hand-copying call notes into HubSpot. Chain Gong, Zapier, and Claude so every call ends as a scored note on the right deal.

Connect. In Zapier, set the trigger to Gong's 'New Call Transcript Ready' event. Authenticate Gong with an admin key so it can pull the full transcript, not just metadata. Add HubSpot as the action app in the same Zap.

Trigger. Filter the Zap so it only fires on calls tagged as a specific deal stage (discovery, demo, negotiation). You do not want internal syncs and no-shows running the model and burning tokens. Use Gong's call tags or the associated deal stage as the filter.

Hand-off. Add a Claude step (via the Anthropic API action or a webhook) that takes the transcript and returns the MEDDPICC table plus a two-line next step. Then map that output to a HubSpot 'Create Note' action logged against the matching deal, keyed on the prospect email.

The power move. Add a second HubSpot action that updates a custom 'Deal Health' property from the score, and route any deal scoring under 6 to a Slack channel for the manager. Now weak deals surface the day they weaken, not at the forecast call.

The rep does nothing after hanging up. The scored note, the health flag, and the manager alert all fire on their own.

Clay, run as a signal engine and not a scraper

Most teams use Clay to find emails. That wastes it. Clay's real value is enriching a list with buying signals and letting AI write the account-specific angle at scale.

Wire it in. Start a Clay table from a saved list (LinkedIn Sales Navigator export or a HubSpot view). Add enrichment columns from the built-in providers: headcount growth, recent funding, job postings, and tech stack. These are your signals, not the email address.

The setting that matters. Use Clay's 'Claygent' AI column with a specific research question, not a generic prompt. Set it to read each company's careers page and return only whether they are hiring for the role your product serves. Cap it to run on rows that pass a signal filter first, or your credits vanish.

The gotcha. Claygent will confidently hallucinate a funding round or a headcount number if the source page is thin. Always add a column that returns the source URL alongside the claim, and filter out rows where the source is blank. Unsourced signals are worse than no signals because reps trust them.

The power move. Add a final AI column that writes the first line of the outreach using the enriched signals: 'Saw you posted 4 SDR roles this month.' Push the whole enriched row, first line included, into your sequencer via the HubSpot or Outreach integration. Reps send, they do not research.

Clay stops being a contact database and becomes the thing that tells you which 30 accounts are in-market this week and what to say to each.

A territory-planning GPT your whole team reuses

Every quarter your reps stare at a list of 200 accounts and guess where to start. Build this once as a custom GPT and the prioritization becomes a two-minute paste, not a two-day project.

Load the context. In the custom GPT builder, upload your ICP definition, your closed-won and closed-lost history as a CSV, and your product's three strongest use cases. The history is what lets it learn which account traits actually convert for you.

Set the output shape. Tell it to always return a ranked table: account, tier (A/B/C), the one signal that placed it there, and the opening play. Consistent shape means reps can act without interpreting a wall of text.

Paste this as the instructions.

Ship it. Publish it to your team workspace so every rep uses the same logic. When win/loss data changes next quarter, re-upload the CSV and the whole team's prioritization updates at once.

Prioritization stops being one smart rep's gut feel and becomes a repeatable asset the whole team runs the same way.

Agentic CRM actions are landing now

The tools you already pay for are shipping agents that take actions, not just summarize. Two are worth a move this week.

HubSpot Breeze agents. HubSpot's Breeze can now draft and queue outreach, enrich records, and prep meeting briefs on its own. What's here: an agent that reads the deal and writes the next touch. Why it matters: the busywork between calls shrinks to a review-and-send. Move this week: turn on the prospecting agent for one rep, have them approve or reject every action for five days, and log which drafts they kept.

Claude and ChatGPT connectors. Both now connect directly to Gong, HubSpot, and Salesforce, so you can ask 'which of my open deals went quiet in the last 14 days' and get an answer from live data. Why it matters: forecast prep moves from a spreadsheet crawl to a question. Move this week: connect your CRM and run your Monday pipeline review as five plain-language questions instead of a report.

The early-mover advantage is not the tech, it is the two weeks you spend learning which agent actions you actually trust before your team scales them.

Make deal intelligence a system, not a habit

The reps using AI ad hoc get a good day here and there. The quarter-level play is to make the scoring and prep automatic so the whole team operates a level up by default.

The shift. Stop treating AI as a personal productivity trick each rep discovers alone. Treat it as team infrastructure: shared prompts, a scoring pipeline, and a prioritization GPT everyone runs the same way. The value is in the consistency, not the cleverness.

Start this month. Pick one pipeline: the Gong-to-CRM scoring loop. Build it for one team, measure whether flagged-weak deals actually close worse, and tune the score. One working pipeline earns you the right to build the next.

What it compounds into. Every scored call and every win/loss upload makes the next quarter's prioritization sharper. In a year you have a system that tells you which deals are real and which accounts are in-market, built on your own data. That is a forecast your board trusts and a moat a competitor cannot copy by buying the same tools.

The teams that win the next two years are not the ones with the best prompts. They are the ones who turned the best prompts into a system nobody has to remember to run.