Skip to main content

2. AI GTM Overview

Note: The video covers material not in the guide below — please watch in full.

Comment in Slack

Post your answer in your onboarding channel.

You just saw AI agents handle a tough sales call, coach an underperforming rep, analyze messaging, save an at-risk account, and audit a forecast — all autonomously. Which one surprised you the most, and why?

Training Guide

AI GTM Overview

There's a lot of hype around AI in go-to-market. In this training, you're going to see AI agents actually doing work across every GTM function -- sales, sales management, marketing, customer success, and RevOps -- live.

What You Just Saw​

In the video, five agents were launched in parallel, each tackling a real scenario:

  • Sales rep -- a tough call with Notion went sideways. The agent pulled in the call transcript, the sales playbook, and objection handling docs, then produced a call analysis, a recovery email, and CRM notes ready for Salesforce.
  • Sales manager -- a rep named Michael has been underperforming for three months. The agent analyzed his last five call transcripts against the company's sales framework and delivered a performance assessment with a coachability verdict and a 30-day plan.
  • Marketing -- six prospect calls (wins, losses, stalled deals) plus current messaging docs went in. The agent produced a messaging analysis showing what's landing and what's not, with direct citations from the transcripts. Then it wrote a blog post in brand voice and created wireframes for a slide deck.
  • Customer success -- a champion at Amplitude left with no handoff, six weeks before renewal. The agent built a stakeholder map, researched the company's current leadership, and gave an honest assessment on whether the account is salvageable.
  • RevOps -- a $2.3 million Q1 forecast commit heading into a board meeting. The agent cross-referenced pipeline data, flagged dead deals being carried as committed, and came back with a realistic number between $1.2M and $1.75M.

Each of these ran autonomously for 5 to 20 minutes. You kicked them off and kept working.

The Three Products in AI​

There are three big players right now -- OpenAI, Google, and Anthropic. Each has three distinct products:

  1. The model -- the baseline intelligence. GPT, Gemini, Claude Opus. This is the engine.
  2. The consumer app -- the browser-based chat you already know. ChatGPT, Claude.ai, Gemini. Great for general Q&A, but you're still doing the work -- typing, copying, pasting.
  3. The agent platform -- what you saw in the video. It doesn't just answer questions. It does the work. It reads your files, writes documents, launches sub-agents, connects to your tools, and runs autonomously.

The consumer app is for the general population. The agent platform is for professionals.

What Agent Platforms Can Do That Consumer Apps Can't​

A few features showed up over and over in those demos:

  • To-do lists -- the agent creates an internal checklist for itself so it can maintain intent over a long task. Some agents have run for 4 hours asynchronously on a single task.
  • Sub-agents -- the main agent spins up smaller agents to handle parallel work (like researching a company's recent news while simultaneously analyzing a call transcript) and then pulls the results back together.
  • File creation -- the agent writes files directly on your computer. Call analyses, recovery emails, performance assessments, stakeholder maps -- all saved as real documents you can open and edit.
  • Permission modes -- you control how the agent operates. Plan mode makes it think first and ask for approval. Bypass mode lets it execute autonomously.
  • Token visibility -- the platform shows you exactly how much of the context window you've used, so you know when the agent is running low on working memory.
  • Compacting -- when the agent runs low on tokens, it creates a summary and hands off to a fresh agent. Like an employee ending their shift and passing notes to the next one coming in.

Skills: Teaching Your Agent New Tricks​

Skills are pre-built instruction sets that teach the agent how to do a specific task -- make a PowerPoint, write an email in your brand voice, create wireframes.

Think of it like Neo in the Matrix. "I don't know kung fu." Boom -- now you know kung fu. You download a skill, feed it to your agent, and it knows how to do that thing. You can find and share skills across your team.

MCPs: Connecting to Everything​

Everything in the video used internal files. But agent platforms also connect to your external tools through MCPs (Model Context Protocol). From one agent platform, you can connect to Salesforce, HubSpot, your project management software -- and tell the agent to pull data, push updates, or run tasks without ever opening another tab.

The Setup Is Three Steps​

  1. Download VS Code
  2. Install the Claude Code extension
  3. Log in with your Claude subscription ($20/month)

That's it. Everything you saw in the video runs on that setup.