Benoit Courty: Boost your existing IT solutions with AI (The Scale Project – ep. #3)

 

Benoît Courty is a Senior Data Scientist at the French National Assembly, where he develops innovative solutions to help lawmakers simulate the impact of legislation.

An engineer by training and former CTO of an autonomous drone start-up, he built his expertise at the intersection of software development, deep learning, and public sector data. He is also president of the CodeCarbon association, an open-source project that measures the carbon footprint of algorithms — a pioneering initiative at the crossroads of AI and environmental responsibility.

In this episode, Benoît shares a pragmatic vision of AI: augmenting existing software without rebuilding it from scratch, using MCP protocols and connected LLMs, leveraging APIs already available in the enterprise ecosystem.

 

 In this episode, you will learn:

1. Connect AI to existing software in just a few hours

➤ With the MCP protocol, even basic API documentation may be enough to turn your CRM, simulator, or internal tool into an intelligent assistant.

2. Automate complex tasks without rebuilding your tech stack

➤ LLMs allow you to interact with internal systems using natural language — no need for complex pre-processing or data transformation.

3. Turn dashboards into intelligent copilots

➤ AI can interpret data in context, simulate scenarios, interact with tools, and deliver recommendations in clear language.

4. Combine tech performance with environmental responsibility

➤ Discover CodeCarbon, an open-source tool that automatically measures the carbon impact of your AI workflows — key to ESG strategies

Whether you’re a business leader, CIO, or innovation project manager, this episode shows you how to leverage AI today — with limited resources and without starting from scratch.

Links:

Linkedin: https://www.linkedin.com/in/bcourty/ 

LexImpact : https://leximpact.an.fr 

CodeCarbon : https://codecarbon.io 

Here are a few highlights from our conversation

1. You can enhance existing software with AI — no need to rebuild.

With MCP, connecting internal tools like CRMs or simulators to LLMs is quick and lightweight.

“You can connect an LLM to existing software in just a few hours — no need to rebuild.” – Benoît Courty

  • Use APIs already available in your systems
  • Quickly transform tools into AI-powered assistants
  • Start testing with a prototype in hours

2. LLMs can interact directly with business data — no complex pipelines.

They can extract insights directly from semi-structured or raw data formats

“No need to pre-process the data — just connect it as-is to the AI.” – Benoît Courty

  • Save time with zero ingestion pipelines
  • Let LLMs work directly with your raw data
  • Maintain your existing architecture.

 

3. AI can operate across multiple tools based on user intent.

LLMs can orchestrate toolchains depending on what the user wants to achieve.

 “AI can chain tools, understand what the user wants, and act accordingly.” – Benoît Courty

  • Create smooth and intelligent user flows
  • Connect CRM, simulators, messaging, and more
  • Automate tasks with contextual intelligence

 

4. You don’t need a full-time AI team to get started.

Small teams can build working prototypes using modern, open-source tooling.

“Prototyping an AI agent linked to an existing tool can take just a few hours.” – Benoît Courty

  • Perfect for SMBs or lean innovation teams
  • Rapid integration with open-source and low-code
  • Build a solid POC in days, not months

 

5. You can now measure your AI model’s carbon footprint.

CodeCarbon helps track energy usage and CO₂ emissions at the code level.

“You can measure an algorithm’s CO₂ emissions in just a few lines of code.”– Benoît Courty

  • A concrete ESG performance indicator
  • Visualize the footprint of every project
  • Integrates natively into VS Code

 

6. AI is an extension of your business — not a separate layer.

Collaboration with business users is key to delivering value.

“You need to understand business needs for AI to be truly useful.” – Benoît Courty

  • AI only works if it reflects real use cases
  • Collaboration between tech and business is essential
  • Empower decision-making, don’t add unnecessary complexity

 

7. Experimentation is the fastest route to AI maturity.

Start small, iterate fast, and focus on operational value.

“A prototype built in a few days — and it’s already delivering value.” – Benoît Courty

  • Focus on one use case at a time
  • Iterate using real-world feedback
  • Optimize for results, not perfection

 

8. AI must be framed to avoid hallucinations.

MCP provides structure, ensuring the model uses only verified data.

 “MCP frames responses — we rely on real data.” – Benoît Courty

  • Prevent models from generating fake information
  • Ensure reliable output with structured access
  • Reduce legal and reputational risk

__

Viorel Bucur is the co-founder of Upscale Paris and the podcast “The Scale Project.” An entrepreneur and ICF-certified team coach, he brings over ten years of experience in behavioral sciences, organizational systems, and tech entrepreneurship. He supports leaders and organizations in their digital transformation, AI adoption, and organizational change.

Passionate about human potential and leadership development, Viorel is committed to shaping a new generation of conscious and high-impact leaders, guiding them through transformational journeys that redefine how they work, learn, and lead.