Agentic Growth Accelerator
Agentic Growth Accelerator: Full Curriculum
Build 4 agents, and learn how to build any agent for any problem.
24 February – 21 March 2026
Works with your existing AI setup
Strategy Coach
Claude Projects, Custom GPTs, or Gemini Gems + Deep Research
You'll leave with a personalised AI advisor that knows your background, your Q1 goals, and the patterns that hold you back. Not a chatbot that forgets who you are every session.
What you'll build
- An AI project loaded with your background research, goals, and working context
- Custom system instructions generated through a guided 7-question framework
- A daily workflow where the coach helps you prioritise, make decisions, and stay accountable
What you'll learn
- How AI projects work: the three layers (instructions, knowledge, and conversation) that make any agent useful
- Using Deep Research to build comprehensive knowledge bases your agent can draw on
- Why context beats prompting — and the document structures that make agents actually remember what matters
- Designing constraints and guardrails so your agent challenges you instead of just agreeing
What you won't get anywhere else
I'm sharing two highly refined mega-prompts I've developed over months of daily use: an Interview Prompt that interrogates your communication style, decision-making patterns, and blind spots to generate deeply personalised system instructions. And a Personal OKR Prompt that helps you get crystal clarity on your goals, priorities, and what you're actually optimising for. These aren't generic templates. People I've delivered these to have described the experience as “life-changing”.
Friday: Security & Working with Agents Safely
Before you wire AI into your CRM and calendar, you need to know the risks. We cover how to vet MCP servers (official vs DIY), prompt injection attacks and how to defend against them, permission scoping (what your agent should and shouldn't access), and a decision framework for different risk levels. You'll leave knowing exactly which connections are safe to make in Week 2.
GTM (Pipeline) Agent
MCP Servers (Attio, Todoist, Calendar) & native integrations
Your AI goes from smart-but-blind to wired into your actual tools. We use a GTM pipeline as the example, but the techniques apply to any workflow: connecting your CRM, task manager, calendar, or any platform so your agent can read, act, and automate across them.
What you'll build
- A pipeline review workflow: stale deals, next actions, priority calls
- Automated call prep with deep research pulled into your CRM
- Post-meeting follow-ups: deal stage updates, task creation with full context
- A morning check-in that cross-references your goals, pipeline, and task list
What you'll learn
- MCP (Model Context Protocol): the standard for connecting AI to any live platform — CRM, tasks, calendar, databases, and more
- The safe-first-test pattern for tool integrations: read before you write, verify before you edit — applies to every tool you'll ever connect
- Chaining tools together: how to make platforms that don't integrate directly work together through your agent
- Permission scoping and security boundaries: controlling what your agent can see, do, and modify across any connected system
What you won't get anywhere else
Until a few months ago, this kind of system would have taken weeks of custom development. Now you can build it in a single session. I've been running this setup daily since late 2025 across consulting, coaching, and sales, and you'll see live demos of the real workflows I use every day: pipeline reviews, call prep with automatic research, post-meeting follow-ups that update your CRM and create tasks in one go. An MCP security specialist also joins to cover the risks most tutorials skip entirely.
Friday: Prompt Engineering & Meta-Prompting
Most people write prompts by trial and error. This session teaches you to be systematic. We cover three meta-prompting patterns (having AI critique and improve its own instructions), how to diagnose bad outputs and fix them at the root cause, and refining your system instructions so your agents get sharper over time instead of staying static.
Team Agent
Slack, Teams, or Telegram
Give your AI a permanent home where your team works. Build an always-on agent with persistent memory that's accessible to everyone — turning the personal capability you built in Weeks 1–2 into a shared team resource.
What you'll build
- An always-on AI teammate in Slack (or Teams/Telegram) that anyone on your team can message
- An agent loaded with your knowledge base from Weeks 1–2 — company context, playbooks, and workflows
- Proactive scheduled briefings that reach out to your team without being asked
What you'll learn
- The difference between a personal tool and a shared team resource — and why it matters for adoption
- Building agents with persistent memory that get smarter over time
- Configuring always-on agents accessible to your whole team from the tools they already use
- Proactive agent behaviour: scheduled briefings, automated check-ins, and agents that reach out first
What you won't get anywhere else
This is the scaling moment. Weeks 1–2 made you more capable. Week 3 makes your team capable. Most AI courses stop at personal productivity — we go further and show you how to deploy an agent that your entire team can use from the tools they already live in. You'll leave this session with a shared AI teammate that has persistent memory, knows your business context, and is available 24/7 without anyone needing to learn a new tool.
Friday: Proactive Briefings
Your agents so far wait for you to start a conversation. This session makes them proactive. You'll configure your team agent to send Monday morning briefings in Slack — pulling from your CRM, calendar, and task list to give your team a personalised start to the week. By the end, your agent works for the team even when nobody's talking to it.
Developer Agent
Claude Code, Lovable, Replit, or Cowork
Write a one-shot PRD, hand it to an AI coding agent, and deploy a live prototype in 90 minutes — then experience the power of running multiple agents in parallel. No coding required. Software companies get a working product. Physical product companies get a validation page to measure demand.
What you'll build
- A product agent for writing exceptional one-shot PRDs — specs so clear that anyone (human or AI) can build from them without clarifying questions
- A live, deployed prototype or validation page built from your spec
- Experience managing a team of parallel agents working on different tasks simultaneously
- Something you can share with users, investors, or stakeholders the same day
What you'll learn
- The one-shot spec format: writing briefs so clear that anyone — a developer, designer, agency, or AI coding agent — can build from them without clarifying questions
- How AI coding agents work and when to use each: Claude Code for custom builds, Lovable for landing pages, Replit for quick prototypes
- Managing an autonomous agent: when to interrupt, when to let it run, and how to course-correct without starting over
- Running multiple agents in parallel — the skill that turns a single-threaded workflow into a team of agents working simultaneously
- Going from idea to live, shareable URL — the full deployment loop you'll reuse for every future prototype
What you won't get anywhere else
I've logged 100+ hours with AI coding agents over the past few months, building prototypes, internal tools, landing pages, and full applications. You're getting the distilled lessons from all of that: which tools to use when, the prompting patterns that actually work, the mistakes that waste hours, and the shortcuts that save them. Plus, you'll experience running parallel agents — a glimpse of where this is all heading. This isn't a tutorial you could follow on YouTube. It's guided, hands-on building with someone who's done it enough to know where things break.
Friday: Showcase & Open Office Hours
The finale. Participants demo what they built to the cohort, we review completion stats, and run an open Q&A covering what to build next, how to keep iterating, and where to take your agents from here. You'll leave with a clear picture of what's possible beyond the cohort.
What's included
8 live sessions • 4 agents • 2 hours of 1:1 coaching • 4 weeks
4x Live Build Sessions
Tuesdays, 2.5 hours
Each week you build a working AI agent from scratch, live, with the cohort. Strategy, GTM, product, and prototyping.
4x Tactics Sessions
Fridays, 60 mins
30 minutes of focused teaching on the techniques that make your agents better, followed by 30 minutes of open Q&A with the cohort.
1:1 Coaching with James
Weekly, 30 mins
Weekly personalised session on your specific use case. Get unstuck, go deeper, or ask about anything related to AI.
Office Hours
Built into every Friday session. Open Q&A with the cohort — bring your questions, see what everyone else is building.
Private Community
Four weeks inside a group with James and the cohort. Share notes, stay up to date, and get help between sessions.
Full Recordings
Everything is recorded. Catch up on anything you miss, revisit anything you want to go deeper on.
Prompts, Templates & Frameworks
The mega-prompts, system instructions, and agent frameworks you build during the programme are yours to keep and adapt.
Partner Perks
Discounts on Granola, Attio and more.
Works with Claude, ChatGPT, Gemini, and whatever AI tools you already use.
Ready to build?
4 weeks. 4 agents. Yours to keep forever.
Questions? Email hello@jamesmcaulay.co.uk