Thomas L.

🏢 Client: Datadome - Cybersecurity Leader (250+ employees, international presence)

🎯 Mission: Deploy an AI agent orchestration infrastructure accessible to all business teams, without security risks or IT dependency.

──────────────────────────

🔴 PROBLEM

Datadome wanted to industrialize the deployment of AI agents across its departments (Legal, Sales, Support, Engineering) but without a centralized orchestration infrastructure, each AI initiative created a silo, a security risk, and a dependency on IT teams.

Result: Business teams could not create AI agents themselves. Zero scalability.

──────────────────────────

🟢 SOLUTION: AgentOS per department (via Dust.tt)

Deployment of an AgentOS, an orchestration meta-agent per department, built on Dust.tt, which transforms AI agent deployment into an autonomous and secure process.

──────────────────────────

⚙️ TECHNICAL ACHIEVEMENTS

• AgentOS (Dust.tt) - orchestration meta-agent enabling the creation of specialized AI agents without technical expertise

• Integrated feedback system - employees provide feedback directly to the agent; IT admins iterate in 2 minutes via the meta-agent

• Native AI Governance - company rules, access controls, and deployment validation integrated from the design phase (critical for Datadome: cybersecurity sector)

• AI Agent Fleet Monitoring - usage rate by department, detection of underperforming agents

• Agent-to-Agent Protocol (MCP / Model Context Protocol) - multi-agent coordination between departments: a Legal agent can query a Support agent, a Sales agent can trigger an Ops agent

──────────────────────────

🛠️ STACK & TECHNOLOGIES

Dust.tt · AgentOS · Model Context Protocol (MCP) · Agent-to-Agent orchestration

Multi-agent systems · AI Governance · Enterprise AI deployment · LLM

🏢 Family & Co - M&A & Family Business Governance Firm (15 people)

🔴 Problem: 8 disconnected business tools, chronic re-entry of data, complex sales cycles from 1 month to 14 years without automation. Consultants spent more time managing information than advising, with no capitalization of business knowledge between assignments.

🟢 Solution: After a complete audit of all stakeholders, construction and deployment of an ecosystem of AI agents + interconnected automations on Make.com, Dust, Zoho CRM, Pappers.fr. Improvement of the sales process. End-to-end mission: friction mapping, workflow architecture, production deployment, and team training for autonomy.

⚙️ Achievements:

• Internal AI Agent (RAG) on 500+ methodological resources - contextualized answers for the team and clients, built on Dust

• Automated Pappers.fr Pipeline: extraction of legal data → automatic scoring → decision to integrate into the sales funnel

• Complete sales automation: prospect scoring, Dropcontact/Lusha enrichment, sector monitoring, personalized emailing sequences, Zoho CRM workflows

• Transcription + AI synthesis of client interviews: automatic thematic consolidation and executive summary generation

• Dynamic templates for client presentations powered by CRM data

• Synchronization of the 8 tools → 1 source of truth via Make.com (end of double entries)

• Automatic multi-level Slack notifications based on project status

📊 Results:

✅ Zero manual re-entry between the 8 tools

✅ 500+ methodological resources activated by AI for the entire team

✅ Each interview → usable summary in < 5 minutes

✅ Team of 15 people autonomous in post-mission maintenance

🛠️ Stack: Make.com · n8n · Dust · Relevance AI · Zapier · Zoho CRM · Pappers.fr · Dropcontact · Lusha · Slack · RAG · AI Agent · LLM · AI Automation

🧠PSŌM: Instagram Content Automation by AI Agents for a Psychonutrition Startup

PSŌM (formerly Origine Care) is a young deeptech startup founded by Dr. Guillaume Fond, a researcher specializing in psychonutrition. They develop scientific dietary supplements for brain health (L'Essentiel, Le Probiotique).

Their problem:with a small team (<15 people), they spent 12-15 hours per week manually creating content for 2 Instagram accounts. Not viable for a startup that needs to move fast, and difficult to maintain a "premium" visual consistency when everything is done by hand under pressure.

I implemented a complete AI infrastructure that automates the entire Instagram content production chain, from A to Z.

An AI agent that orchestrates the entire workflow: Upon receiving the text post from the team, it triggers content creation, coordinates the generation of premium visuals via the Midjourney API, and then breaks down the text, assigning it automatically to the carousel slides.

>> 13 hours per week recovered. The team can focus on product development instead of making Instagram carousels.

All visuals respect PSŌM's graphic charter and premium positioning.

The system runs autonomously and can handle 25+ posts per week if needed, without adding any workload to the team.

Ready to scale if PSŌM accelerates.

Sector:HealthTech / DeepTech / Psychonutrition

Keywords:AI Agent, AI Infrastructure, Instagram Content Automation, Visual AI Generation, Content Production Automation, Health Marketing Automation, AI Workflow Automation, Midjourney, AI Brand Content, Make.com, Dust.tt, Instagram Automation, Carousel Automation