Matthieu Rousseau

Matthieu Jonglez from Progress talking about how to use AI with #semantic #agentic #RAG on structured and unstructure data from a companies’s private systems “The greater the context, the greater the insights” #progressdatasummit

56

2 commentaires

GPT-4.1 vs GPT-5 in SuiteAnalyzer: Speed or Depth? 🤔 In side-by-side tests with the same financial data, GPT-5 delivered far richer insights than GPT-4.1 — but took longer to respond. In my latest blog post, I explain why this trade-off is expected, how costs compare, and when to choose each model. https://lnkd.in/eVRkJpwf

27

8 commentaires

Processing 5M+ reviews taught us that volume reveals what prototypes hide. Small-scale tests with clean data work perfectly. Production scale with messy reality exposes every architectural weakness. → Inconsistent data formats multiply: PDFs, images, handwritten text, emails ↳ Single-format prototypes miss 60% of production complexity ↳ Multi-modal extraction becomes mandatory, not optional → Performance degrades non-linearly: 100 documents work, 10K documents break ↳ Sequential processing creates bottlenecks ↳ Parallel multi-agent orchestration becomes the only viable approach → User expectations shift: "Why can't it handle this edge case?" ↳ Every exception requires graceful handling ↳ No-code platforms enable business users to adapt without engineering This is why SimplAI built no-code multi-agent orchestration as core capability. Enterprise adoption requires platforms that non-technical users can customize and deploy. We've seen 4000+ merchants and 40+ use cases built by teams without AI engineering backgrounds. When business users can build and deploy AI workflows, adoption accelerates 10X. Are your AI platforms accessible to the people who understand the workflows best? 🔥 Want more breakdowns like this? Follow along for insights on: → No-code agentic platforms that democratize enterprise AI → Platform design decisions that enable business user adoption PS: Whenever you're ready, take SimplAI for a spin at our website or book a demo: https://lnkd.in/g3JkSJHb

27

Forget plug-and-play AI. We’re entering the era of agent-native systems and this is the new stack. 👇 Let’s break it down: 🔹 The Agent-Native Stack (MCP Edition) 1️⃣ Interfaces → Where agents live → Claude, LibreChat, Cursor, Windsurf → Not just chat — structured workflows, tool use, memory 2️⃣ MCP Servers → Where agents work → Databases (Supabase, Clickhouse) → Email (Resend), Docs (Notion), Tickets (Linear), Payments (Stripe) → All agent-accessible. All programmable. 3️⃣ Execution Environments → Where stuff gets done → Browserbase, ScrapNARA, E2B → Agents browse, click, extract, interact — securely and at scale 4️⃣ Coordination & Monitoring → What makes it reliable → AgentDesk, Braintrust, Grafana → Track flows, manage fallbacks, handle complexity 5️⃣ Marketplace Layer → Plug-ins + deployment infra → Mintlify, Glama, MCP.so, Smithery → Launch-ready agents + serverless hosting 🧠 If you're building ops-ready automation: → Don't stop at “can it call an API?” → Ask: Can it run workflows, recover, monitor, adapt? This isn’t just “tools with AI.” It’s AI that runs your tools. End-to-end. Autonomous. Coordinated. 📌 Save this stack. 📥 DM if you're exploring agent infra, orchestration, or automation beyond the prototype stage. Follow Dennis for weekly drops on building smart systems that scale.

39

27 commentaires

A few years ago, 90% of our revenue came from Airtable customers. We lived in constant fear they'd destroy us overnight. For years, this drove our strategic decisions: • Tried to diversify away from Airtable • Built our own data platform • Constantly worried about the day they'd pull the rug Then it happened. Airtable changed their pricing and API terms. We scrambled. Made changes. Warned customers. And... mostly everyone was fine. Some customers had to switch plans. But the impact was nowhere near what we'd imagined. Then Airtable built portals - one of our main features. We thought this was the end. But people still come to us. Our portals work better. Our pricing makes more sense. Even after they built what we already had, we kept winning. Big lesson there: The existential threats you spend years worrying about usually aren't as bad as you think. And the defensive moves you make because of those fears often hurt more than the actual threat would. We wasted time and energy building around a risk that turned out to be manageable. Focus on making customers happy instead of protecting against imaginary disasters.

34

2 commentaires

🚨 The new AI Agent update in n8n just changed the game for no-code orchestration. You can now build fully modular agent armies — where each AI agent lives inside your main agent as a tool. No more scattered workflows. No more Execute Workflow nodes. Just clean, centralized AI orchestration. In my latest tutorial, I show you how I built a system that: 🤖 Uses one master orchestration agent to control all sub-agents 📩 Automates emails, calendars, expense tracking, and company knowledge 🧠 Assigns memory, models, and tools per agent — inside a single workflow 📂 Removes the need for multiple builds by letting agents act as tools 📈 Scales perfectly for client work, personal use, or AI assistant businesses Watch the full build + demo here 👉 https://lnkd.in/gDgfRfW6 Want to learn how to create AI automations like this — and package them into real client offers? Join the AI Workshop community — where I share full blueprints, sales strategies, and how to turn automations into income. 👉 https://lnkd.in/gnNJqH9u #AI #Automation #n8n #nocode #AIagents #VoiceAI #AIWorkshop #OpenAI #Claude #Gemini #AgentOps #WorkflowDesign

89

6 commentaires