Who This Is For
This is for founders, consultants, and salespeople at B2B practices who want a more systematic, research-backed way to qualify inbound leads before a discovery call.
It is especially useful if you are:
Running 5–15 discovery calls a month and spending 20–60 minutes manually researching each lead the night before
Opening a dozen browser tabs per lead — LinkedIn, the company website, Crunchbase, Google News — and stitching the picture together yourself
Realizing after a call that the lead was obviously not a fit, and that you could have seen that in five minutes with the right information
Prepping leads inconsistently — doing deep research when you have time and shallow research when you don't
Booking calls through Calendly or a similar tool but having no system that connects the booking to any research or record
Managing a pipeline in Notion or Airtable but filling it manually, row by row, after calls have already happened
Working solo or with a very small team where there is no SDR or ops person to own this work
This is not for someone who only takes two or three sales calls a month and does not mind spending 30 minutes prepping each one manually.
This flow makes the most sense when you are running enough calls — roughly five or more per week — that the cumulative research overhead is visibly eating into your productive time.
How the Workflow Runs
Calendly pulls upcoming bookings → Tavily MCP researches each company → Claude scores against your ICP → Notion logs the full record → Slack DMs you the ranked briefing
Time to Set Up
Time to set up: 30–60 minutes for first setup, connector configuration, and ICP customization. Budget the higher end if you have not used MCPs before or if you need to restructure your Notion workspace to use a real database.
Daily time after that: 2 minutes each morning to read the briefing. The agent runs everything else on its own.
Tools Needed
Claude (with Routines / scheduled agents)
Calendly (or an alternative booking tool — see Step 2)
Tavily MCP (web search and page extraction — not a standard app connector; see Step 2)
Notion database (not a static table — see Step 2 for why this matters)
Slack
A Tavily API key (free tier available at tavily.com)
Step-by-Step Walkthrough
Step 1: Understand what this flow should and should not do
Set expectations first.
This flow helps you:
Research a company’s identity, size, industry, and AI maturity from live web sources (no stale data)
Surface recent signals: funding, hires, layoffs, lawsuits, revenue news (last 30 days)
Score each lead 0–100 against your ideal customer profile
Recommend one action per lead: full call, screen first, or decline politely
Write a permanent, searchable database record for every briefed lead
Send a ranked daily briefing to your Slack DMs (no channel noise, no manual forwarding)
Flag already-processed leads to prevent duplicate research
It cannot replace your judgment when signals are mixed or incomplete.
A lead with no public web presence is scored on intake-form answers alone (lower confidence).
It does not:
Pull data from gated sources (e.g., LinkedIn Sales Navigator, ZoomInfo)
Verify info not on the open web (e.g., private company’s internal revenue)
Send outbound messages or edit your calendar
Make the final call on meetings — it gives a recommendation, not a decision
Step 2: Set up the four connectors in Claude
This flow runs inside a Claude Routine — a scheduled agent that executes end-to-end on a daily schedule. You need four connectors attached to the Routine before it can do anything.
Calendly — connect your Calendly account through Claude's native connector. This is the only source the agent reads leads from. Do not also connect a spreadsheet or CSV export of your bookings — if you give the agent multiple lead sources it will pull from all of them and count duplicates.
Tavily MCP — Tavily is the research engine. It is not a one-click app connector; it runs as an MCP server and requires a separate install step. The right approach is to install it at the project level — not per session — so it persists across every Routine run. A per-session install writes to a temporary container that gets wiped on restart. Get your API key from tavily.com, add it to your project config, and keep it out of any shared prompt text or public repository. If a key has ever been pasted into a chat or committed to a repo, rotate it immediately.
Notion — connect your Notion account and point the Routine at a real Notion database, not a static table block inside a page. This distinction caused the build to fail in its first version: the Notion API only supports querying and appending rows against a proper database object. A table that visually looks like a database — one created with the
/tableslash command inside a page — will not work. Create a new database (inline or full-page) with the following typed columns: Name (Text), Company (Text), Email (Text), Domain (Text), Score (Number), Recommendation (Select), AI Maturity (Text), Budget Signals (Text), Contact Summary (Text), Sources (Text), Processed (Checkbox), Brief Posted At (Date).Slack — connect your Slack workspace. The agent resolves your own user ID at runtime and sends the briefing as a DM to you directly, not to a channel.
Note: A connector being linked to your Claude account is not the same as it being attached to a specific Routine. Every Routine has its own connector list. After creating the Routine, go into its settings and attach all four connectors explicitly — Calendly, Tavily, Notion, and Slack. If any connector is missing from the Routine, the agent will silently skip that step.
Step 3: Replace the default ICP block with your real context
The prompt in Step 5 includes an ICP block — a written definition of your ideal customer. The defaults in that block are placeholders based on a generic B2B SaaS profile. Before the agent runs a single live briefing, replace every line with your actual business context.
Update the ICP block with your:
Target industries (e.g., "B2B SaaS, professional services, fintech")
Company size range (e.g., "50–500 employees")
Funding or revenue signals that indicate a real budget (e.g., "Series A or B in the last 18 months, or demonstrably profitable")
AI maturity stage you serve best (e.g., "exploring AI applications or actively building a first AI product — not already mature")
Contact seniority and function (e.g., "VP or above in Product, Engineering, or Operations")
Ideal buying timeline (e.g., "1–6 months to a decision")
Hard disqualifiers — signals that should override everything else and produce a DECLINE recommendation (e.g., "recent layoffs, active litigation, B2C business model, no discernible budget signal")
Also update your Calendly intake form before the first real run. The agent's scoring improves significantly when intake answers are structured. Add these fields if they are not already there: company name, company size (dropdown), buying timeline (dropdown), and a short-answer field asking what they want to use AI for. Vague or empty intake answers reduce scoring confidence.
Want the full breakdown?
This is where you get real AI workflows, prompts, and systems you can use to automate your work. If you're serious about using tools like Claude to grow your business, this is for you.
