Most marketing teams are stuck in “user” mode. They prompt ChatGPT for copy. They ask Midjourney for images. They treat AI like a vending machine — insert request, get output.
Zapier’s CMO Dan Slagen just proved there’s another way.
At MAICON 2026, Slagen shared what happened when he forced his entire marketing team to stop using AI and start building with it. The playbook: three weeks learning a coding agent, five internal coaches on call, a two-week build-a-thon where everyone shipped something real, and personalized feedback on every single project.
The result? A marketing team that builds internal dashboards, automates reporting pipelines, and prototypes product features — without waiting for engineering.
Here’s the playbook, why it works, and how to run it yourself.
The Wake-Up Call
Slagen didn’t start with a memo. He brought in an external AI power user to demo what was possible.
The reaction was visceral. People were stunned. Confused. Some were genuinely upset.
“I’m not a fan of leading by fear,” Slagen said, “but every once in a while you need to make people understand how big this moment is.”
That moment: the bar for “good with AI” got dramatically higher at the end of 2025. When non-technical people first got their hands on coding agents like Cursor, Claude Code, and Codex, the old adoption playbook — watch a tutorial, write better prompts — stopped working entirely.
The Three-Week Ramp
Week one through three: every marketer picked one coding agent and went deep. Cursor, Claude Code, or Codex — their choice. No mandated tool, just mandated time.
Five internal power users became always-on coaches. Stuck on a loop? Can’t get the API call right? Ping a coach. They’re unblocked in minutes, not days.
This is the first place most companies fail. They buy licenses. They don’t assign buddies.
The Build-A-Thon
Two weeks. One rule: ship something.
Projects ranged from internal dashboards that replaced manual spreadsheet work to fully functional product prototypes. One marketer built a competitive intelligence tracker that pulls pricing data nightly. Another built a content calendar that auto-generates social variants from a single brief.
Every project got a video walkthrough. Slagen watched all 70 of them and wrote a personalized response to each person.
Not a thumbs-up. A real response: what worked, what to try next, where to level up.
Tiered Skill Placement
The team self-placed into three tiers:
Tier 1 — Comfortable: Can build end-to-end workflows, handle API auth, debug independently. Next challenge: architect multi-step agent systems.
Tier 2 — Building: Shipped the build-a-thon project but needed coach help. Next challenge: own a recurring automation without hand-holding.
Tier 3 — Started: Completed the ramp, built something basic. Next challenge: go deeper on one tool instead of dabbling in three.
No one felt outclassed. Everyone had a clear next step.
Why This Works
Three things Slagen got right that most leaders miss:
1. Structure beats enthusiasm. “Learn AI” is a wish. “Three weeks on Cursor, then build something in two weeks” is a plan.
2. Coaching beats content. Tutorials age badly. A colleague who unblocks you in real time compounds.
3. Public accountability beats private effort. Video walkthroughs + personalized feedback = nobody ghosts the program.
The Hard Truth
Slagen’s closing warning: no LLM, platform, or software is coming to save your team.
The shift to AI builders relies on people, structure, and culture. Companies that change how they work will benefit both quantitatively (speed, output) and qualitatively (agency, retention). Companies that don’t will keep prompting and wondering why the gap keeps widening.
Your Move: Run the Playbook This Quarter
You don’t need Zapier’s budget. You need the structure.
Week 1-3: Ramp
- Pick one coding agent per person (Cursor, Claude Code, Codex)
- Assign one coach per 3-4 learners
- Protect the time — no “real work” excuses
Week 4-5: Build-A-Thon
- Everyone ships one thing that solves a real annoyance
- Record a 5-minute walkthrough
- Leadership watches and responds to every single one
Week 6: Place and Progress
- Self-place into three tiers
- Assign tier-appropriate next challenges
- Repeat quarterly
The Meta-Lesson
The capability overhang is real. Your team — like OpenAI’s 200M weekly student users — is likely tapping 1-10% of what’s possible.
The gap isn’t tooling. It’s structure.
*Source: Dan Slagen, “From AI Users to AI Builders: The Next Phase of AI in Marketing [MAICON 2026]” — Marketing AI Institute
