Growth-stageAI & Machine Learning

AI that updates the CRM, with a human approval gate in front of every write

Four plugins live in the customer's environment

ClaudeSlackn8nFireflies.aiSalesforce
Play leadSamuelSenior GTM Engineer · LeanScale
Read the case study ↗ Share

The problem

Leadership wanted to become an AI-first go-to-market organisation, not simply buy an AI tool. Reps were updating pipeline by hand and preparing for calls from scratch, and the company had no governed way to distribute internal AI workflows across the team, nor any answer to the obvious objection, which is that a model writing directly into the CRM is a data-quality risk rather than a data-quality fix.

The play

  1. 1
    Sales skills scoped to the repetitive parts of the weekClaude

    Shipped Claude-based plugins and sales skills covering morning brief, call prep, competitor intel, customer follow-up and post-call summary, the recurring work that reps skip when the week gets busy.

  2. 2
    A transcript-to-CRM enrichment pipeline

    Built a pipeline that takes call transcripts from the meeting recorder, routes them through an automation layer, and lands them on a dedicated meeting-transcript object in the CRM where the model proposes field enrichment against the opportunity.

  3. 3
    A human gate in front of every CRM writeSlack

    Put a Slack approval step ahead of every write, so the model proposes and a person approves. Nothing reaches the CRM unreviewed, which is what makes the pipeline deployable in a system the forecast depends on.

  4. 4
    Governed distribution rather than pasted prompts

    Governed the plugins through a version-controlled plugin marketplace with an enablement and governance plan, so internal AI workflows are versioned and distributed deliberately instead of copied between people.

What it moved

Four plugins live in the customer's environment

Marketing, revenue-operations, sales and company-context plugins went live in the customer's own instance.

Transcript-to-CRM pipeline built and tested. The pipeline was built and tested end to end on LeanScale's systems and was pending transfer into the customer's environment at the end of the window, built and proven, not yet running in their production.

From a LeanScale case studyRead the full case study →

More from Samuel

GTM AI agents +2 6 steps Putting AI agents inside the CRM, and fixing the routing data first so they aim at the right people The routing waterfall is live and backfilled GTM AI agents +3 7 steps Turning every sales call into proposed CRM updates, with the rep approving in chat, not in the CRM The pipeline works end to end and was demoed live