Skip to content
GTM Agent Weekly
Subscribe

Roll out Nooks agents with CRM context and a custom signal layer

This is how one team moved its outbound reps onto a single platform where agents research accounts, write messages and support calling. The lesson: agents only help if the CRM context behind them is good.

Paste this into ChatGPT, Claude or your agent tool. See the prompt

  • Prospecting / Outbound
  • intermediate
  • SDR manager
  • RevOps
Category
Prospecting / Outbound
Difficulty
Intermediate
Est. cost
Not stated by source

Works with

Problem

BDRs were running list building, research, messaging and dialing across four separate platforms. Agents that research accounts are only as good as the context stored in the CRM.

Setup

A GTM engineer prepares Nooks before reps log in: rebuilds every back-end rule from the old sequencer, makes sure Salesforce holds context such as closed-lost reasons, and sets up the agents reps will use with context files on how the company sells. He picks out-of-the-box signals and writes a scheduled web-scan prompt for niche signals; a signal overnight triggers enrichment and research so a rep can email first thing in the morning.

Prompt

Starter prompt by The GTM Agent Weekly, written from the setup described on this page. It is not the source's own prompt; see the source link below for their write-up.

Starter prompt · 26 lines 
# ROLE
You run an overnight web scan for niche buying signals at our target accounts, so a rep can
email first thing in the morning.

# INPUTS
- {{ACCOUNTS}}: target accounts with domain and owner.
- {{NICHE_SIGNALS}}: the signals our standard tools miss, each with a definition and why it
  matters to us (e.g. a specific regulation filing, a new location, a tech migration post).
- {{HOW_WE_SELL}}: a short context note on our product, buyers and messaging.
- {{CRM_CONTEXT}}: closed-lost reasons and past conversations per account.

# TASK (nightly)
1. For each account, search public sources for each signal in {{NICHE_SIGNALS}} published
   since the last run.
2. Keep only signals with a quote, a date and a link.
3. For each hit, write a two-line research note: what happened, why it matters given
   {{HOW_WE_SELL}} and {{CRM_CONTEXT}}.
4. Suggest an opening line for the morning email.

# OUTPUT
Per hit: account, owner, signal, quote, date, link, note, opening line.

# GUARDRAILS
- No link, no signal. Ignore anything older than {{MAX_AGE_DAYS}} days.
- If the account was closed-lost, reference what changed since, not the loss itself.
- Never contact anyone; the rep sends.

Replace every {{PLACEHOLDER}} before running. Works with any capable LLM and runner.

Tools

Steps

  1. Audit every automation in the existing sequencer and rebuild each one before go-live.
  2. Store the context reps need for plays in the CRM, such as closed-lost reasons.
  3. Configure the agents with context files on positioning so research and copy are not generic.
  4. Select the third-party signals that matter and add a scheduled web-scan prompt for niche ones.
  5. Train reps on basics first, then on using the AI assistant to research and build plays.

Results

The author reports large meeting and call-volume increases at the champion's previous company; sponsored post, not independently measured.

What goes wrong

Gotchas

  • Sequencers hide rules nobody remembers, like suppressing existing customers or removing repliers; miss one and reps email the wrong people.
  • If a detail is not in the CRM, the agent will not know about it.
  • The post is sponsored by the vendor, so treat the outcome figures as self-reported (editor note).

Source and credit

thegtmengineer.substack.com ›

Noah Adelstein, The GTM Engineer (featuring Evan Inscoe) · newsletter

Source dated 2026-09-23 · last reviewed 2026-10-11. Summarized in our own words; see the source for details.

Related.