What Happened:
Whataburger deployed a "composition agent" that automates marketing campaign creation from Figma to Braze, reducing build time from 4 hours to minutes per campaign and saving approximately 25 hours weekly across their typical six campaigns per week.
Taco Bell deployed a QA agent that executes over 200 test scripts in under one minute, cutting quality assurance time from 2+ hours to seconds per campaign and eliminating late-night work for their team while reducing error rates.
Stitch built these implementations run on top of the tools the teams already use. Using n8n as the workflow layer, the agents pull from Figma, reference project tasks in ClickUp/Jira, and build directly into Braze via API. This was presented at Braze Forge 2025.
Our Take:
Taco Bell and Whataburger did a full end-to-end analysis of their marketing campaign workflow, and importantly, not every step was a good candidate for AI. Ideation, offer strategy, creative direction, and brand voice required human context and taste, so those pieces remained in-house. But two steps stood out as repetitive, rules-based, and slowing things down:
Asset composition (turning approved designs into Braze builds), and
Pre-launch QA (checking rendering, targeting, deliverability, and configuration).
It’s worth noting that the financial investment wasn’t trivial. Each agent required approximately 300 hours from three team members to build, with costs ranging from $25,000 to $250,000 depending on complexity. But they were happy with the ROI. 25+ hours saved weekly translates to roughly 1,300 hours annually per brand, freeing their lifecycle teams for strategy and innovation rather than administrative tasks.
The agents run inside n8n - an AI agent builder that’s seen tremendous growth (example view shown above).
On the technical side, concerns around security and autonomous action were mitigated with guardrails put on agents. Agents were designed not to schedule or deploy campaigns, and a human was always the final approver.
To address the challenge of platform API limitations (like Braze not having an image upload API), Stitch leveraged a tool called Puppeteer. This tool mimics human interaction on a browser, allowing the agents to log in and upload images, but with permissions strictly pared down to only the specific tasks required.
So are these agents replacing jobs? Troy Vo said his Taco Bell team had zero concerns about AI taking their roles. Instead, eliminating 'mundane, redundant work' gave them space for the strategic thinking that actually requires human judgment.
Looking ahead, both brands plan to adopt more agents. Whataburger will implement QA automation while Taco Bell hopes to add composition automation. This could shorten their full campaign cycles from 14 days to 6-8 days.
If you want to replicate this:
Start with a four-week pilot on one high-frequency campaign (e.g., weekly offers).
Use your existing stack and out-of-the box tools (Braze + n8n + Figma/ClickUp/Jira). Avoid net-new tools at first to keep security reviews light.
“In the next three to five years,” says Stitch’s Bobby Tichy, “AI and automation will really heavily take over the campaign process - specifically the repetitive, copy-paste things we’re used to today.”
For the full breakdown, watch the full webinar from Stich here.