How to Build a Team of AI Agents for Your Agency
Why It Matters
Most agencies are simply attaching AI to what they already do, hoping it works out. Webtopia took a different approach. They created a well-organized system where each tool has a clear purpose, every output from the agents is reviewed by a human, and adoption is directly connected to compensation. This is what truly sets apart "we use AI" from having a complete operating system for running an agency smoothly at scale. For those overseeing agent-driven purchases and creative automation, this offers a rare glimpse of what a fully integrated stack actually looks like in real practice, not just in theory.
Who This Is For
This platform is designed especially for Founders members who are running or growing their agencies, particularly those handling paid media and creative projects for various clients. If you're looking to understand how AI can genuinely integrate into your team's daily work (rather than just testing tools on their own), Tristram's framework and agent roster offer you a clear and practical blueprint to follow.
The question many of us have been asking for some time now: how to actually build an AI agent team that empowers your humans and your business. If you’re an agency owner, a marketer, or just someone curious about how AI is changing the creative industry, we’re breaking down insights from Tristram Dyer, CEO of Webtopia and OAKS, as featured on the AI DTC WTF podcast with Andrew Foxwell, Will Sartorius, and Thomas Moen.
1. The Fast and the Slow: AI’s Two Speeds in Agencies
Tristram kicks things off by sharing a simple but powerful framework: Daniel Kahneman’s “Thinking, Fast and Slow.” It’s a mindset for structuring your agency’s AI adoption:
Fast AI: Think of this as the turbocharger for your team. It’s about speeding up the repetitive, data-heavy tasks that used to bog down your best people. Fast AI takes hours-long jobs and turns them into minutes.
Slow AI: This is the thoughtful, always-on background work. Slow AI is your safety net, catching things a busy human team might miss, such as performance anomalies, budget pacing, and daily summaries.
The real magic happens when you combine both. Fast AI gives your humans superpowers, and Slow AI keeps everything running smoothly, ensuring nothing falls through the cracks.
2. Building Your Stack: The Tools and How They Work Together
Tristram’s agency runs on a slew of AI-powered tools, each with its own job to do:
Fast AI Tools
Meta MCP, Kaizen MCP, Motion MCP: These pull in real-time insights from platforms like Meta (Facebook/Instagram) and drop them into one dashboard. No more digging through a million reports just to figure out what’s working.
Creative Workflow: Claude, Arcads, Claude Design, HeyGen: Here’s how it works:
Claude drafts creative briefs and scripts, distilling what made past campaigns successful.
Arcads then generates UGC-style ads from those briefs with no need to chase clients for more content.
Claude Design takes those creative assets and spins up multiple versions, tweaking overlays and branding for different audiences.
HeyGen steps in to translate and localize the content, helping you scale campaigns across languages and markets in a flash.
Every Webtopia employee gets a paid Claude account, from junior execs to project managers, and everyone’s expected to become fluent in these tools and to experiment with them on a daily and weekly basis.
The Kaizen Layer: Connecting Data to Human Insight
Numbers alone don’t tell the whole story. Kaizen brings in qualitative data, the actual conversations with founders, marketing managers, and ops leads. It records and analyzes calls, capturing sentiment, goals, and the “why” behind client decisions. As a result, it supports the creative and strategic planning that’s rooted in both hard data and real human context.
There's a noticeable gap between simply adding ChatGPT to your workflow and building a smooth operation like Webtopia's, with clear roles for each tool, built-in human review, and incentives that keep everyone engaged. It’s not something you achieve with just a blog post. Instead, you learn by observing what other operators have built, what challenges they faced, and what they would improve. That's what the Foxwell Founders Membership is all about.
3. Tangible Results: What’s Changed
Rolling out this stack didn’t just speed things up. It made client communication clearer, creative output sharper, and teams a whole lot happier. Tristram noticed:
Faster decision-making: With all data and creative tools connected, the team can make, test, and iterate on campaigns in a fraction of the time.
Transparent client updates: AI agents generate daily and weekly summaries, pulling from six or seven data sources. Humans review and edit these before sending them off, so clients are always in the loop, and trust grows.
More time for strategy: With the grunt work off their plates, the team can finally dig into the high-impact analysis and creative thinking that actually moves the needle.
4. The AI Agent Team: Who Does What?
Webtopia’s “AI agent team” is a set of routines and bots, each with a job:
The Agent Roster
Performance Anomaly Detector: Scans all accounts every morning for major shifts in ROAS/CPA and pings the right human on Slack if something’s off. It checks trends over different time windows (3, 7, 14 days) for extra accuracy.
Budget Pacing Bot: Keeps spend on track, flagging issues before they become expensive problems.
Morning Briefing Agent (Agent 00): This is if Jarvis from Iron Man was obsessed with your ad performance. It pulls together everything you need to know about campaign health, creative wins, and budget pacing.
Meta Ads Operator: Suggests optimizations for Facebook/Instagram ads based on your agency’s SOPs, but always requires a human to approve changes.
Mia, the Google Ads Specialist: A new addition, Mia audits Google Ads accounts, surfaces winning (and losing) search terms, updates shopping titles, and flags negative keywords. She’s already improving results after just two weeks in action.
5. Seamless Integration: Where the Data Lives
Everything feeds into a secure Google BigQuery warehouse. Every agent pulls and pushes data here, so there’s a “single source of truth” for both numbers and qualitative feedback. Even Slack communications get analyzed and looped in, so nothing gets lost.
All the agent outputs, reports, and alerts get piped back into Slack. No more dashboard-hopping.
6. Humans + AI: Why It Works
Tristram is crystal clear on this: AI isn’t here to replace your team. It’s here to multiply what they can do, help them level up, and let them focus on the stuff that actually matters.
Human-in-the-loop: Every AI suggestion, from ad changes to daily reports, is reviewed and approved by a person.
Cultural fit is key: The agency hires people who are excited to use AI as a tool, not threatened by it. This “growth mindset” for his team is non-negotiable.
Continuous improvement: Feedback from staff on agent suggestions is fed back to the AI routines, making them smarter every week.
If MCPs are becoming core to how agencies operate, get ahead of it.
7. Incentivizing AI Adoption: How to Get Your Team Onboard
How do you make sure everyone’s actually using these tools?
Universal access: Every staff member gets a paid Claude account and is encouraged to experiment.
Performance-based bonuses: Bonuses are tied to client results and satisfaction, not just hours worked. If you use AI to improve client performance or free up time, you win.
Quarterly reviews: High client CSAT scores (tracked via Kaizen and Slack) feed into bonus calculations. If AI helps deliver better results faster, that’s recognized.
8. The Roadmap: What’s Next for AI Agent Teams?
Looking ahead, Tristram sees three big shifts:
Democratization: These DIY agent tools will soon be available to founders and clients directly, not just agencies.
Agentic Buying: Fully autonomous media buying, where the AI doesn’t just suggest but actually pulls the trigger on ad buys, is just around the corner. Early tests show agentic buying may be able to beat human results in some cases.
Human Expertise Remains Essential: Even as AI gets smarter, it still can’t replicate human intuition, creativity, and judgment. Agencies that blend AI’s speed with human insight will deliver the best results.
9. Will AI Replace Agencies?
The panel’s answer is a confident “no.” AI will lower the barrier to entry for running ads, but great results still require expertise. As Tristram puts it:
“You could paint your house yourself, but you probably won’t. People are always willing to pay for experts.”
AI will make agencies more efficient, more creative, and more valuable, not obsolete. The agencies that thrive will be those that embrace change, incentivize innovation, and keep the human touch front and center.
10. Final Thoughts: Your Playbook for Building an AI Agent Team
If you're ready to start, here’s your checklist, inspired by Webtopia’s journey:
Adopt the “Fast and Slow” AI mindset: Use fast AI to supercharge workflows and slow AI for oversight and strategy.
Connect your data: Use tools like Meta MCP, Kaizen, and Motion to unify quantitative and qualitative insights.
Automate the grunt work: Bring in creative agents, anomaly detectors, and daily briefing bots to free up your team.
Keep humans in the loop: Every major AI action should have a human reviewer.
Incentivize innovation: Reward staff for client results, not just hours.
Focus on client communication: Use AI to keep clients informed, involved, and trusting.
Prepare for a future where clients use agents too: Don’t just build for your agency. Build tools that empower your clients.
The agency of the future isn’t about humans versus AI, it’s about humans plus AI. The winners will be the ones who treat AI as a creative partner, not the enemy.
Craving more insights? Hit play and catch the full podcast episode here!
This is the kind of breakdown you get every episode. Subscribe to AI DTC WTF so you don't miss the next one.
