LangChain 
-
The problem
Inbound leads often sat in the CRM until someone researched the person and company and wrote an email. By then the lead might have gone cold, or another rep had already reached out, because there was no process to reconcile who was working what.
-
What they built
An agent that picks up every new inbound lead. It screens out leads that aren't worth contacting, such as people who are too junior, researches the person and company, and drafts four follow-up emails that land in the right rep's Slack to edit and send. When a rep edits a draft, the agent learns that rep's style and uses it on their next leads. Vishnu based it on how LangChain's best reps already researched and wrote outbound, using the same data sources they used.
-
What changed
Vishnu says it brought down the time from a lead coming in to the team reaching out. When an outage stopped the agent loading its learned rep preferences, reps' use of it to draft emails dropped.
-
So what
Every inbound lead is researched and has a first draft right away, and reps keep the final say. The drafts learn how each rep writes, which is why reps kept using it.
Under the hood
For GTM engineers
A new Salesforce lead triggers the agent. Leads from sources like contact-sales or the trust center are deflected, the person is looked up in the CRM or on LinkedIn for seniority and role, and the agent pulls past emails and meeting notes for warm accounts or researches the company's AI initiatives and launches for cold ones. Its starting examples came from the team's outbound-wins Slack channel. After a rep edits a draft in Slack, an LLM compares the original with the edit and writes any meaningful preference to an append-only log keyed by the rep's user ID, which is loaded before each run and compacted monthly. Traces, opens and replies are logged in LangSmith, and changes are checked against an eval set of deterministic and LLM-as-judge tests before deploy.
More examples
Other systems GTM engineers have built, or see all of them.
-
Profound AI agents check the company's Google Ads on a schedule and send suggested changes to Slack, where a person approves them
Nick Lafferty · Founding Marketing Engineer -
Sequel.io Turns a product brief into a full launch kit, from blog post to sales messaging and BDR emails, so a four-person team can launch weekly
Kathleen Booth · VP of Marketing - Profound
Shows which of the team's LinkedIn posts bring in sales leads, by lining up post views with leads in the CRM
Nick Lafferty · Founding Marketing Engineer
Hiring someone who can build like Vishnu?
We search for GTM engineers who have shipped systems like this one, and Joe speaks to every candidate before you do.