Why MadKudu Went All-In on AI-Native Operations

At this point, most SaaS companies are using AI. But few are doing what Francis Brero, co-founder and CPO of MadKudu, is doing: rebuilding the way the entire company thinks, builds, and communicates around AI — not just in product, but across go-to-market, support, and operations.

We talked about what it means to be truly AI-native — not just AI-enabled. And how it’s changing the way teams write code, ship features, onboard customers, and even write documentation.

The big idea? The future won’t be owned by the best coders. It’ll be owned by the companies who can ship fastest, delegate most, and make LLMs part of the team.

Key Takeaways

• MadKudu shifted from GitHub Copilot to Cursor for AI-native dev workflows
• They feed architecture, product, and help docs directly into the dev environment
• Teams are using AI to auto-generate PRDs, TRDs, and prototypes
• Documentation is written for LLMs — not humans — to support AI agents
• Every team runs weekly AI hackathons to compound learning and capability

From Sales Copilot to AI-Centric Org

MadKudu started with a clear goal: automate B2B sales prioritization with AI. Their platform removes the grunt work from prospecting so sales reps can focus on deals that actually convert.

But now, they’re turning that same mindset inward — using AI to scale the company itself.

“We’re not just using AI to ship product. We’re rebuilding our operating system around it.”

Why They Switched to Cursor

The dev team moved from GitHub Copilot to Cursor for a few key reasons:

• You can feed it internal documentation

• It understands company-specific architecture

• It behaves more like an agent than a code completion tool

“Cursor became our senior engineer. It knew the codebase and the business logic.”

They’re feeding it:

• Internal Notion docs

• Product requirement documents

• API references

• Even customer support data

That context gives Cursor the power to solve simple bugs, prototype new features, and free up engineers to focus on systems thinking.

AI-Created Specs → AI-Written Code

Francis’s team is now using AI to:

• Write PRDs (product requirements)

• Translate those into TRDs (technical specs)

• Feed both directly into Cursor to generate code

It’s not just faster — it’s also changing how people think about their role.

“Our engineers don’t think about writing code first. They think about describing the system — then letting AI implement it.”

The Hidden Shift: Writing for Machines

One of the wildest parts of the conversation?

Their support team doesn’t write docs for humans anymore. They write them for LLMs.

“Images are great for humans — but useless for LLMs. So we write in long form. Descriptive, structured, parsable.”

That tiny shift changes how teams document everything — from onboarding to product training — because the audience isn’t just new hires anymore. It’s agents and AI tools that will act on that knowledge.

AI Moats Are Gone. Distribution Wins.

Francis was blunt: the era of defensible tech moats is over. LLMs are getting better fast, and the switching cost between tools is dropping close to zero.

“The new moat is distribution. If your product can’t distribute itself, or be part of a scalable GTM, it’s toast.”

MadKudu’s next frontier? Using its own product to drive AI-powered distribution — and embedding intelligence into every part of the customer journey.

The Advice: Don’t Wait. Build the Muscle Now.

“The barrier to doing something valuable with AI is lower than you think. But you won’t see the value until you start practicing.”

MadKudu runs weekly AI hackathons — one week with GTM, next with product, next with support. It’s not about massive wins. It’s about building the reflex.

Start small. Build fast. And compound the learning.

Final Word

“AI won’t replace your team. But the companies who train their teams to delegate to AI will move 10x faster than the ones who don’t.”

Book a Free AI Assessment if you’re trying to make that leap — from AI features to AI foundations. We’ll help you design the workflows, prompts, and process layers that actually scale.