Lab HoursGet new playbooks

Agents · Library

Agent library

Nine agents, each doing one job well. For every one: what you give it, what you get back, what it reads and what needs your approval.

Which agent should I start with?

Four core agents cover a deal from start to finish: Account context (what you already know), Account intelligence (what is happening now), First call prep (before the call) and Call follow-up (after it). Start where your team loses the most time. Each works on its own, and they get better together.

01 · Core agent

Account context

The full account history on one page, in about a minute.

You give it
A company name, or tomorrow’s meeting
You get
Every contact and their role, all past touchpoints, past deals and why they closed, and where the relationship stands today.
When to use it
Before any call, and whenever an account changes owner.
Your approval
Read-only. It summarises, it changes nothing.

What it reads

  • Your CRM: contacts, emails, meetings, deals and notes

How it works

  1. Confirms the account by website domain, and flags possible duplicates instead of merging.
  2. Reads only your CRM: contacts, meetings, replies, downloads, deals, signals and owners.
  3. Writes one note under the heading Account Context and shows it to you.
  4. After your approval, pins it to the company. Re-running replaces it, never stacks a second copy.

What you get, section by section

  1. Buying group by role: budget holder, champion, evaluators, daily users
  2. Coverage: how many people, which teams, which roles are missing
  3. Engagement, newest first, and the last real two-way conversation
  4. Deals: open, won and lost, with the lost reason in plain words
  5. The strongest signal on the account, with its date
  6. One suggested next step, and review flags such as a stale title or a duplicate record
  • CRM only, never the web
  • Empty field means the line is left out, never guessed
  • One pinned note per account

Playbook: Win more conversationsOpen skill file: coming to GitHub soon

Example output (made-up company)
  • Arden Retail · customer since 2024 · owner changed last month
  • 7 contacts · 3 active · champion: Lena Vos (finance chief)
  • Pilot won 2024 · expansion lost 2025: “timing, budget frozen”
  • 41 touchpoints · last two-way reply 6 weeks ago
  • Suggested next step: re-open expansion with the new budget year

02 · Core agent

Account intelligence

A cited brief on a top account: why now, who owns the budget and what to say.

You give it
A priority account, or your list of top accounts
You get
Fit and timing, company priorities, who owns the budget, dated triggers and an opening line per person, every claim sourced.
When to use it
Before a planned play on a top account, then monthly or quarterly.
Your approval
It shows the brief first. It is added to the pinned note only after you approve.

What it reads

  • News and press releases
  • Annual reports and filings
  • Leadership changes
  • Job ads
  • Public interviews and posts
  • Your Account Context note

How it works

  1. Reads the Account Context note first, so it knows who is already mapped.
  2. Runs a few targeted searches on company priorities, your four value themes and recent moves.
  3. Looks for public quotes from budget holders only. Everyone else: "No public statement found".
  4. Writes about 600 words, every claim sourced and dated or marked as an inference.
  5. After your approval, adds it under the Account Context in the same pinned note, replacing last time’s brief.

What you get, section by section

  1. Header: fit, timing, priority person and why now in one line
  2. Snapshot, then company priorities
  3. Four sections on the themes where your product matters (you choose them)
  4. Who owns the budget, and what the budget holders have said publicly
  5. Why now: dated, sourced triggers
  6. Your angle, each hook tied to a numbered finding and a person
  7. An opening line per person, and review flags
  • Cited and dated, or marked inference
  • “No public signal found” instead of padding
  • Runs in batches of five or six, highest priority first

Playbook: Win more conversationsOpen skill file: coming to GitHub soon

Example output (made-up company)
  • Harbor Foods · Fit: High · Timing: Hot · Priority: operations chief
  • Why now: agreed to buy a rival, closing in Q1 (press release, 24 Sep) [1]
  • Budget holder, on record: “We will cut our supplier list in half” (earnings call, Aug) [2]
  • Evaluators: no public statement found
  • Hook for the operations chief, tied to finding 9: one view across both companies
  • Review flag: website changed after the rebrand

03 · Core agent

First call prep

Walk into every first call knowing what matters to them.

You give it
Anything: a name, a LinkedIn link, a calendar invite or a CRM contact
You get
A one-screen brief: recent moves, who they are, what they posted in the last 60 days, your history with them, three hypotheses and five questions.
When to use it
The evening before, or an hour before the call.
Your approval
Nothing is sent or saved. It is your prep sheet.

What it reads

  • Your calendar
  • Your CRM and past emails
  • Their LinkedIn
  • Company reports and news

How it works

  1. Works from anything you give it: a name, a LinkedIn link, a calendar invite or a CRM contact.
  2. Confirms the right person and company first. If a colleague owns the record, it tells you.
  3. Pulls your history from the CRM. Only real replies count, not opens or clicks.
  4. Checks what the person posted or commented on LinkedIn in the last 60 days.
  5. Researches the company and writes a one-screen brief with numbered sources.

What you get, section by section

  1. What the company is and what it sells
  2. Recent moves that matter, dated
  3. Your contact: role, background, what they care about
  4. Digital maturity and pressures on their business
  5. What to know going in: three hypotheses, five questions, what could not be verified
  6. Where you stand with them: owner, replies, meetings, past deals
  7. Their public activity in the last 60 days
  • Read-only: never contacts anyone, never writes to the CRM
  • Specific to this company, not the industry
  • Every claim numbered to a source

Playbook: Win more conversationsOpen skill file: coming to GitHub soon

Example output (made-up company)
  • Lena Vos · finance chief, Arden Retail · in role 6 months
  • Came from logistics. Posts about delivery costs.
  • Also involved: head of operations, IT lead
  • Guess 1: late deliveries are hurting key customers
  • Ask: “What would make this a priority this quarter?”

04 · Core agent

Call follow-up

Turns a call into a summary, next steps and CRM updates.

You give it
The call transcript, your notes or the meeting in your note-taker
You get
Buyer questions grouped by theme, a SPICED summary, and CRM values checked against your allowed options, in deal-screen order.
When to use it
Right after the call.
Your approval
It shows current and proposed CRM values side by side. Nothing is saved until you approve.

What it reads

  • The transcript from your note-taker
  • The deal in your CRM

How it works

  1. Asks which outputs you want first, so it never does more than needed.
  2. Pulls the transcript from your note-taker once.
  3. Writes only from what was said. Anything not discussed is marked “not surfaced, ask next time”.
  4. Checks proposed CRM values against your allowed options and date formats before showing them.
  5. Shows them in the same order as your deal screen. You approve, optionally a manager too, then it writes.

What you get, section by section

  1. Buyer questions grouped by theme, with who asked and what dominated
  2. A SPICED summary: situation, pain, impact, critical event, decision, next step
  3. Proposed CRM values with the evidence from the call for each one
  • Nothing saved without your approval
  • Never overwrites filled fields without asking
  • Blank and flagged beats guessed

Playbook: Win more conversationsOpen skill file: coming to GitHub soon

Example output (made-up company)
  • Problem: late deliveries are costing key customers
  • Competition: talking to two other vendors
  • Deadline: new contracts from January
  • Next step: demo for the ops team on 14 October
  • 4 CRM updates proposed · approve to save

05 · Agent

Account brief

A one-page brief on any company, before you reach out.

You give it
A company website or name
You get
What they do, what changed recently, who decides, which buying signals are live and a suggested first line.
When to use it
Before the first email or call to a new company.
Your approval
Nothing is sent. The brief is a draft for you to read.

What it reads

  • The company website
  • News and press releases
  • Job ads
  • LinkedIn
  • Your CRM history

How it works

  • Confirms it has the right company before researching.
  • Looks for buying signals and dates each one.
  • Marks anything it could not verify instead of guessing.
  • Keeps the brief to one page.

Playbook: Win your top accountsAgent file on GitHub: coming soon

Example output (made-up company)
  • Northwind Freight · logistics · 400 people
  • New operations chief started 3 weeks ago (company news, 2 Sep)
  • Hiring two sales operations roles (careers page)
  • Likely priority: faster quoting for new customers
  • Suggested opener: how other freight firms cut quote time
  • Not verified: size of current sales team

06 · Agent

Right-person check

Sorts job titles into your personas, or “not a fit”, with a reason.

You give it
A list of names and job titles
You get
A persona for every title (or “not a fit”), the reason, which team they belong to and their likely role in a decision.
When to use it
Before any outreach list goes live, or when cleaning the CRM.
Your approval
Nothing is sent or saved. You get a sorted sheet.

What it reads

  • Your list
  • Your written customer profile

How it works

  • Uses fixed rules in a set order, so the same title always gets the same answer.
  • Excludes roles that borrow your words but do a different job.
  • Logs which rule fired for every verdict.

Playbook: Pick the right companiesAgent file on GitHub: coming soon

Example output (made-up company)
  • VP Revenue Operations → persona: revenue leader · decision-maker
  • Revenue Accounting Lead → skip · finance role, not your buyer
  • Managing Director → persona: founder or CEO · senior sponsor
  • Director, Event Operations → skip · different job

07 · Agent

Email check

Scores your cold email and tells you what a busy exec will skip.

You give it
Your draft email
You get
A score, the lines that sound automated or salesy, and a tighter version.
When to use it
Before you send a new sequence.
Your approval
Nothing is sent. You choose what to keep.

What it reads

  • Your draft
  • Your written customer profile, if available

How it works

  • Checks length, subject line, one question per email and no calendar links.
  • Flags openers like “just following up” and flattery like “congrats on the new role”.
  • Checks every claim can be backed up.

Playbook: Reach out at the right timeAgent file on GitHub: coming soon

Example output (made-up company)
  • Score: 58 / 100
  • Cut: “Hope you are well” and “quick question”
  • Too long: 142 words, aim for under 90
  • Two questions: keep one
  • Rewrite ready (78 words)

08 · Agent

AI visibility check

Tells you whether AI assistants recommend you, and who they name instead.

You give it
Your category, competitors and a few buyer questions
You get
Whether ChatGPT, Claude, Gemini and Perplexity mention you, how prominently, who they name instead and which sources they cite.
When to use it
Monthly, and after big content or PR pushes.
Your approval
Read-only. It reports, it changes nothing.

What it reads

  • AI assistant answers with web search on
  • The sources they cite

How it works

  • Runs each question several times, because answers change.
  • Scores every answer with the same rubric.
  • Checks that what the AI says about you is correct.

Playbook: Google & AI Search Visibility CheckAgent file on GitHub: coming soon

Example output (made-up company)
  • 24 buyer questions · 4 assistants · 3 runs each
  • You are mentioned in 21% of answers that do not name you
  • Most often named instead: Competitor A (46%)
  • Top cited source you are missing from: a “best tools” list

09 · Agent

Meeting sorter

Sorts every meeting transcript and finds patterns across calls.

You give it
Access to your meeting transcripts
You get
Each meeting sorted as sales, customer, product feedback or internal, linked to its account, and themed reports on objections, requests and competitors.
When to use it
Weekly, or before planning and product reviews.
Your approval
Read-only. Reports only.

What it reads

  • Your note-taker’s transcripts
  • Your CRM, to match accounts

How it works

  • Uses clear rules first and AI only for unclear meetings.
  • Logs which rule sorted each meeting.
  • Quotes buyers word for word, with who said it.

Playbook: Win more conversationsAgent file on GitHub: coming soon

Example output (made-up company)
  • 38 meetings this week · 14 sales · 9 customer · 6 feedback · 9 internal
  • Top objection: “We already use spreadsheets for this” (5 calls)
  • Most requested: an export to PowerPoint (3 calls)

Want one of these running on your own CRM and inbox? Share your challenge in Lab Hours and compare notes.