AI sales research and outbound agent for a 50-rep B2B team
An agent that researches accounts, drafts genuinely personalized outbound, and books meetings — reps approve instead of grinding.
Personalization that didn’t scale
Reps were told to personalize every touch, which meant 20 minutes of research per prospect — so most quietly stopped. Generic sequences followed, and reply rates sank.
The cost of leaving it alone
The team’s pipeline math required reply rates the generic sequences couldn’t produce. Leadership faced a choice: more reps doing bad outreach, or better outreach per rep.
Research, draft, approve, send
An agent researches each account across news, site, and CRM history, drafts a personalized sequence, and queues it for one-click rep approval in Slack.
- Account research across live web sources and CRM interaction history
- Drafts grounded in verifiable facts, with sources shown to the rep
- One-click approve/edit flow in Slack; nothing sends unreviewed
- HubSpot logging so every touch and outcome feeds reporting
Stack: Claude · HubSpot · Slack · Exa
How it was built
- Week 1–2: research pipeline and fact-grounding rules with sales leadership
- Week 3–4: drafting engine tuned on the team’s best historical emails
- Week 5: Slack approval flow and HubSpot write-back
- Week 6: pilot with eight reps, then staged rollout to fifty
What the numbers say
What happened next
The approval step stayed permanently — reps trust the system because they remain the byline. Drafting rules get retuned quarterly from the reply data the system itself collects.
This system is an example of CRM & Lead Management Systems work.
Need a similar system?
Let's talk through your version of this — same architecture thinking, scoped to your operations and tools.
30 minutes · no pitch deck · reply within 24h if you write instead