When Fabien Cros joined Ducker Carlisle from Google, he expected to find teams already building sophisticated AI workflows with frameworks like LangChain and AutoGen. Instead, he discovered something far more common: organizations overwhelmed by AI possibilities, unsure where to start, and relying on small technical teams that had become impossible bottlenecks.

"Coming from Google, I was expecting everybody to be super advanced, ready to build their own solutions," Fabien recalls. "What we discovered is that most companies are completely lost about AI. It's not a criticism, it's just a fact. So we repositioned the way we think about AI entirely."

Ducker Carlisle, which had recently consolidated its Sparkwise Solutions brand under the Ducker Carlisle name, now pioneers a radically different approach to enterprise AI: the Citizen Developer Program. Built on StackAI's no-code platform and supported by centralized governance, the program enables business users across the firm to rapidly design and iterate AI agents without writing code. The results speak for themselves: over 50 agents deployed across dozens of use cases, with the company on track to achieve $1 million in annual cost savings.

"We used to be the bottleneck. I had dozens of requests coming in, and it was unsustainable. By decentralizing from a team of 5-10 people to 100 people equipped with StackAI, we went from weeks of building time to minutes."

Fabien Cros, Chief Data and AI Officer, Ducker Carlisle

The Challenge: Breaking the Centralized AI Bottleneck

Ducker Carlisle faced a problem that plagues most mid-sized consulting and professional services firms attempting AI transformation. Their data and AI team was drowning in requests while most employees waited weeks or months for basic automation workflows.

The centralized model wasn't working:

Fabien's lean tech team received dozens of requests for AI use cases across different practice areas. Building each workflow in-house with traditional development tools meant week-long backlogs. By the time one use case was deployed, ten more requests had arrived. The team couldn't scale to meet demand, and business users grew frustrated with the pace of innovation.

They noticed the decentralized alternative was failing too:

Some organizations had tried the opposite approach: giving everyone access to generic AI tools like Microsoft Copilot with minimal training or support. This rarely worked. "They install Copilot, tell people it will be magical, and nobody trains them. Nobody gives them support," Fabien explains. "People look at it, see it's not working, and then say 'AI doesn't work.' It's a complete failure."

Ducker Carlisle realized they needed something in between: a way to decentralize AI development while maintaining the expertise and governance that made solutions actually work.

The Solution: Empowering Citizen Developers with StackAI

Ducker Carlisle's answer was a structured Citizen Developer Program built on StackAI's no-code platform. The program follows a simple but powerful model: business users build 90% of their AI workflows themselves, while a central team of experts provides the critical 10% of technical support that makes solutions production-ready.

Discovery Through Google

Fabien's path to StackAI began while he was still at Google. When Google invested in StackAI, Fabien reached out to discuss integrating Gemini models into the platform. After joining Ducker Carlisle, he recognized that StackAI's approach aligned perfectly with their needs.

"We originally thought of StackAI as a quick prototyping tool," Fabien explains. "The idea was to prototype in StackAI, then develop production versions in-house. But as we discovered how low the AI maturity was across most companies, we realized StackAI could be the foundation for 80% of our use cases."

The 80/20 Model

Ducker Carlisle adopted a strategic framework: 80% of AI use cases stay within StackAI, enabling rapid deployment and iteration. The remaining 20% of highly strategic, business-critical workflows get developed in-house for maximum control and customization.

This model accomplishes several goals simultaneously. Most business needs get addressed quickly through StackAI. The internal development team focuses only on the most valuable use cases. Business users gain autonomy without sacrificing quality or security.

Program Structure: Support Without Suffocation

Now, business users can handle 90% of agent development: defining the workflow, connecting data sources, testing with real scenarios, and deploying to users. But they need expert support for the critical 10%: complex integrations, advanced routing logic, security and governance review, and optimization for production scale.

"It's really like we take this 10% expertise and help everybody do the 90%. That's the beauty of it," Fabien explains. "We're tapping into the opportunity of the technology so everybody can build. The minute they're blocked by expertise, we just infuse the prerequisite knowledge to go from zero to one."

Community-Driven Innovation

The program's strength comes from its community aspect. When one team solves a problem, others learn from it. When someone discovers a useful integration pattern, it gets shared. When a workflow delivers exceptional value, it becomes a template others can adapt.

This community approach has created a flywheel effect. More builders means more use cases. More use cases means more shared learning. More learning means better quality and faster development across the board.

Real-World Use Cases: From Prototypes to Production

Ducker Carlisle has deployed dozens of AI agents across their consulting practice. Here are some of the most impactful:

Voice of Customer Analysis: From Hours to Seconds

The Challenge: Ducker Carlisle's M&A practice conducts 10-15 customer interviews per project, generating 150-200 pages of transcripts. Writing final reports required finding specific quotes to support key findings. Consultants spent 8-12 hours per project using Ctrl+F to manually search through hundreds of pages, often settling for "good enough" quotes instead of the perfect illustration. They frequently lost track of which interview a quote came from, requiring time-consuming re-verification.

The Solution: The team built an AI agent that indexes all interview transcripts in a single Knowledge Base with automatic metadata tagging. Consultants now ask natural language questions like "Find quotes about pricing concerns" or "What did customers say about implementation challenges?" and the agent returns exact sentences with full source attribution—interview name, timestamp, participant role—in under 10 seconds. They can click any citation to verify full context in the original transcript.