Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now
Like many enterprises over the past year, Intuit Mailchimp has been experimenting with vibe coding.
Intuit Mailchimp provides email marketing and automation capabilities. It’s part of the larger Intuit organization, which has been on a steady journey with gen AI over the last several years, rolling out its own GenOS and agentic AI capabilities across its business units.
While the company has its own AI capabilities, Mailchimp has found a need in some cases to use vibe coding tools. It all started, as many things do, with trying to hit a very tight timeline.
Mailchimp needed to demonstrate a complex customer workflow to stakeholders immediately. Traditional design tools like Figma couldn’t deliver the working prototype they needed. Some Mailchimp engineers had already been quietly experimenting with AI coding tools. When the deadline pressure hit, they decided to test these tools on a real business challenge.
The AI Impact Series Returns to San Francisco – August 5
The next phase of AI is here – are you ready? Join leaders from Block, GSK, and SAP for an exclusive look at how autonomous agents are reshaping enterprise workflows – from real-time decision-making to end-to-end automation.
Secure your spot now – space is limited: https://bit.ly/3GuuPLF
“We actually had a very interesting situation where we needed to prototype some stuff for our stakeholders, almost on an immediate basis, it was a pretty complex workflow that we needed to prototype,” Shivang Shah, Chief Architect at Intuit Mailchimp told VentureBeat.
The Mailchimp engineers used vibe coding tools and were surprised by the results.
“Something like this would probably take us days to do,” Shah said. ” We were able to kind of do it in a couple of hours, which was very, very interesting.
That prototype session sparked Mailchimp’s broader adoption of AI coding tools. Now, using those tools, the company has achieved development speeds up to 40% faster while learning critical lessons about governance, tool selection and human expertise that other enterprises can immediately apply.
The evolution from Q&A to ‘do it for me’
Mailchimp’s journey reflects a broader shift in how developers interact with AI. Initially, engineers used conversational AI tools for basic guidance and algorithm suggestions.
“I think even before vibe coding became a thing, a lot of engineers were already leveraging the existing, conversational AI tools to actually do some form of – hey, is this the right algorithm for the thing that I’m trying to solve for?” Shah noted.
The paradigm fundamentally changed with modern AI vibe coding tools. Instead of simple questions and answers, the use of the tools became more about actually doing some of the coding work.
This shift from consultation to delegation represents the core value proposition that enterprises are grappling with today.
Mailchimp deliberately adopted multiple AI coding platforms instead of standardizing on one. The company uses Cursor, Windsurf, Augment, Qodo and GitHub Copilot based on a key insight about specialization.
“What we realized is, depending on the life cycle of your software development, different tools give you different benefits or different expertise, almost like having an engineer working with you,” Shah said.
This approach mirrors how enterprises deploy different specialized tools for different development phases. Companies avoid forcing a one-size-fits-all solution that may excel in some areas while underperforming in others.
The strategy emerged from practical testing rather than theoretical planning. Mailchimp discovered through usage that different tools excelled at different tasks within their development workflow.
Governance frameworks prevent AI coding chaos
Mailchimp’s most critical vibe coding lesson centers on governance. The company implemented both policy-based and process-embedded guardrails that other enterprises can adapt.
The policy framework includes responsible AI reviews for any AI-based deployment that touches customer data. Process-embedded controls ensure human oversight remains central. AI may conduct initial code reviews, but human approval is still required before any code is deployed to production.
“There’s always going to be a human in the loop,” Shah emphasized. “There’s always going to be a person who will have to refine it, we’ll have to gut check it, make sure it’s actually solving the right problem.”
This dual-layer approach addresses a common concern among enterprises. Companies want AI productivity benefits while maintaining code quality and security standards.
Context limitations require strategic prompting
Mailchimp discovered that AI coding tools face a significant limitation. The tools understand general programming patterns but lack specific knowledge of the business domain.
“AI has learned from the industry standards as much as possible, but at the same time, it might not fit in the existing user journeys that we have as a product,” Shah noted.
This insight led to a critical realization. Successful AI coding requires engineers to provide increasingly specific context through carefully crafted prompts based on their technical and business knowledge.
“You still need to understand the technologies, the business, the domain, and the system architecture, aspects of things at the end of the day, AI helps amplify what you know and what you could do with it,” Shah explained.
The practical implication for enterprises: teams need training on both the tools and on how to communicate business context to AI systems effectively.
Prototype-to-production gap remains significant
AI coding tools excel at rapid prototyping, but Mailchimp learned that prototypes don’t automatically become production-ready code. Integration complexity, security requirements and system architecture considerations still require significant human expertise.
“Just because we have a prototype in place, we should not jump to a conclusion that this can be done in X amount of time,” Shah cautioned. “Prototype does not equate to take the prototype to production.”
This lesson helps enterprises set realistic expectations about the impact of AI coding tools on development timelines. The tools significantly help with prototyping and initial development, but they’re not a magic solution for the entire software development lifecycle.
Strategic focus shift toward higher-value work
The most transformative impact wasn’t just speed. The tools enabled engineers to focus on higher-value activities. Mailchimp engineers now spend more time on system design, architecture and customer workflow integration rather than repetitive coding tasks.
“It helps us spend more time on system design and architecture,” Shah explained. “Then really, how do we integrate all the workflows together for our customers and less on the mundane tasks.”
This shift suggests that enterprises should measure AI coding success beyond productivity metrics. Companies should track the strategic value of work that human developers can now prioritize.
The bottom line for enterprises
For enterprises looking to lead in AI-enhanced development, Mailchimp’s experience demonstrates a crucial principle. Success requires treating AI coding tools as sophisticated assistants that amplify human expertise rather than replace it.
Organizations that master this balance will gain sustainable competitive advantages. They’ll achieve the right mix of technical capability with human oversight, speed with governance and productivity with quality.
For enterprises looking to adopt AI coding tools later in the cycle, Mailchimp’s journey from crisis-driven experimentation to systematic deployment provides a proven blueprint. The key insight remains consistent: AI augments human developers, but human expertise and oversight remain essential for production success.
