Clément Herr

90% of Private Equity firms are playing with #AI. The other 10% have embedded AI into their investment process and are and compounding EBITDA. 📍 In our latest article, we break down what’s actually happening inside #PrivateEquity firms. blog.lampi.ai/ai-adoption-in… #AIforFinance

AI agents are everywhere. But building one that actually works in production? That’s hard. At @lampi_ai, we solved real challenges with a multi-agent architecture that works reliably at scale. Find out more in our latest blog article. 🔗Link👇 #AIagents #AIforfinance

First large-scale study of AI agents actually running in production. The hype says agents are transforming everything. The data tells a different story. Researchers surveyed 306 practitioners and conducted 20 in-depth case studies across 26 domains. What they found challenges common assumptions about how production agents are built. The reality: production agents are deliberately simple and tightly constrained. 1) Patterns & Reliability - 68% execute at most 10 steps before requiring human intervention. - 47% complete fewer than 5 steps. - 70% rely on prompting off-the-shelf models without any fine-tuning. - 74% depend primarily on human evaluation. Teams intentionally trade autonomy for reliability. Why the constraints? Reliability remains the top unsolved challenge. Practitioners can't verify agent correctness at scale. Public benchmarks rarely apply to domain-specific production tasks. 75% of interviewed teams evaluate without formal benchmarks, relying on A/B testing and direct user feedback instead. 2) Model Selection The model selection pattern surprised researchers. 17 of 20 case studies use closed-source frontier models like Claude Sonnet 4, Claude Opus 4.1, and GPT o3. Open-source adoption is rare and driven by specific constraints: high-volume workloads where inference costs become prohibitive, or regulatory requirements preventing data sharing with external providers. For most teams, runtime costs are negligible compared to the human experts the agent augments. 3) Agent Frameworks Framework adoption shows a striking divergence. 61% of survey respondents use third-party frameworks like LangChain/LangGraph. But 85% of interviewed teams with production deployments build custom implementations from scratch. The reason: core agent loops are straightforward to implement with direct API calls. Teams prefer minimal, purpose-built scaffolds over dependency bloat and abstraction layers. 4) Agent Control Flow Production architectures favor predefined static workflows over open-ended autonomy. 80% of case studies use structured control flow. Agents operate within well-scoped action spaces rather than freely exploring environments. Only one case allowed unconstrained exploration, and that system runs exclusively in sandboxed environments with rigorous CI/CD verification. 5) Agent Adoption What drives agent adoption? It's simply the productivity gains. 73% deploy agents primarily to increase efficiency and reduce time on manual tasks. Organizations tolerate agents taking minutes to respond because that still outperforms human baselines by 10x or more. 66% allow response times of minutes or longer. 6) Agent Evaluation The evaluation challenge runs deeper than expected. Agent behavior breaks traditional software testing. Three case study teams report attempting but struggling to integrate agents into existing CI/CD pipelines. The challenge: nondeterminism and the difficulty of judging outputs programmatically. Creating benchmarks from scratch took one team six months to reach roughly 100 examples. 7) Human-in-the-loop Human-in-the-loop evaluation dominates at 74%. LLM-as-a-judge follows at 52%, but every interviewed team using LLM judges also employs human verification. The pattern: LLM judges assess confidence on every response, automatically accepting high-confidence outputs while routing uncertain cases to human experts. Teams also sample 5% of production runs even when the judge expresses high confidence. In summary, production agents succeed through deliberate simplicity, not sophisticated autonomy. Teams constrain agent behavior, rely on human oversight, and prioritize controllability over capability. The gap between research prototypes and production deployments reveals where the field actually stands. Paper: arxiv.org/abs/2512.04123 Learn design patterns and how to build real-world AI agents in our academy: dair-ai.thinkific.com

@danshipper @lampi_ai Confidential AI agents for financial firms! Feel free to reach out!

We’re working with a gigantic hedge fund to find the AI products that will reshape their entire operation. What are some tools they should be paying attention to? Drop the links!

What a week at @VivaTech and @NVIDIAGTC ! 🚀 It was a real pleasure to demo our confidential #AI agents! Huge thanks to our partners @Scaleway and @nvidia 🤝 and to everyone who stopped by our booth. 💬 #VivaTech2025 #AIAgents #Innovation

🚀 Lampi AI Joins the @Scaleway Growth Stage Startup Program! 🎉 This opportunity strengthens our commitment to building #AI agents that prioritize #privacy and #security and delivering a wide choice of #cloud providers to meet all demands. 🔒🇪🇺 #GenerativeAI #technology #LLM

👉 With Lampi AI Meeting Assistant, go one step further than transcription and #AI-generated summaries! ✅ Retrieve key insights 🔎 ✅ Bring meeting insights into #AIAgent workflows 🤖 ✅ Turn meetings into a shareable #knowledge hub📚 🔗For more blog.lampi.ai/from-meeting-a…

👉 What if your investment notes or market studies could be ready in minutes, not days? #AI agents make it possible! 📈 AI #agents don’t just assist — they automate. 🔗 Learn more about #AIagents in financial analysis 👉 blog.lampi.ai/the-future-of-… #GenerativeAI #technology #LLM

👉Generative #AI has emerged as a powerful way for businesses to unlock their #data. But, as #GenAI adoption grows, so do concerns around #confidentiality and #security 🔐⚠️. 📖 Explore our latest security guide to learn more: blog.lampi.ai/enterprise-ai-… #GenerativeAI #LLM

👉 Lampi AI is excited to collaborate with the @Oracle and @AmpereComputing teams and demonstrates efficiency in #AI applications in the context of a real use-case for enterprises. ✨ The latest #Oracle blog post explores the value of #CPUs in supporting retrieval augmented generation (#RAG) with vector embeddings.💡 It includes a testimonial from our experience, where we demonstrate that supervised AI #agents running on CPUs could be envisaged to autonomously perform redundant and asynchronous tasks based on RAG, such as monthly portfolio analysis, performance review, transcription categorization, customer feedback analysis, etc. ➡ Read more:  blogs.oracle.com/ai-and-datasci… #OCI #generativeAI #AIagent #LLM #AIinnovation

👉 Lampi AI is now part of the exclusive NVIDIA Inception Program! 🎉 This recognition is a testament to our innovative use of #AI. 💡 For more information: blog.lampi.ai/lampi-ai-joins… #NVIDIAInception

#VIVATECH is over for this year! 🎉✨ Europe's biggest event for #innovation and #startups exceeded our expectations. 🚀 Whether at Lampi AI's or @Scaleway 's booth (with @AmpereComputing ), or during various side events (thank you, @nvidia!), it was full of amazing meetings and exchanges! 🤝💡 Congratulations and thank you to everyone! 🙌 See you next year! 🗓️ #VivaTech2024

👉 Just 2 days until @VivaTech! ⏳ Join us to connect, network, and discuss the latest #AI trends with industry leaders from May 22 to 25! 🚀 📅 May 22 - 25 📍 Paris Expo Porte de Versailles 🎪 Hall 2 - R03-004 Can’t wait to catch up with you there! 👋

Proud to share that @lampi_ai is now part of @OVHcloud Startup Program! Excited for the journey ahead! 🚀 @OVHcloudStartup

👉 We are excited to announce that @lampi_ai has been selected for the OVHcloud Startup Program 🎉 @OVHcloud stands as a leading global cloud provider, and we are thrilled to partner with them to accelerate our development. 🌐 For more information: blog.lampi.ai/lampi-ai-joins…

Are you looking to leverage #AI into your organization?💡 While identifying AI use cases, requirements, and ROI remains challenging, it should be your first step into your AI journey. 🔍 Read more about it 📚: blog.lampi.ai/this-first-ste… #innovation #technology #LLM #GenAI

After our collaboration with @AmpereComputing at the #GoogleCloudNext 2024 in Las Vegas, where we showcased a real-time voice-to-text and voice-to-voice chatbot interaction with knowledge base running on an Ampere-based cloud instance, Lampi AI joins the second stage of Google for Startups Cloud Program! 🚀 This second stage of collaboration enables us to access the latest Google Cloud's robust resources, offering us advanced computational power, facilitating development and deployment of secure, efficient #AI solutions tailored to our clients' needs. 🛠️ In addition to our collaboration with Google Cloud, we remain committed to offering modular and flexible deployments to our clients, from dedicated cloud with French providers to on-premise deployment to ensure data sovereignty. 🇫🇷🔒 For more information: blog.lampi.ai/after-google-c…

👉 We'll be at @joinstationf tomorrow for AI Day with @FRdigitale and @InstitutPrairie! Our CEO, @GCouturas, will be here to talk about open-source and profitability with Caroline Chopinaud (@HubFranceIA) and Anais Monlong (@Iris_Capital) 💡 🤝 Feel free to contact us! We're looking forward to meeting you! For more information 👉 francedigitale.org/agenda/ai-day-… #IADay #IA #GenIA #StationF

🇪🇺MiCA update: the EU Parliament has published its final agenda for the plenary session next week. Both MiCA (comprehensive crypto regulation) and the TFR (crypto travel rule implementation) will be discussed on Wed afternoon & voted on Thursday. europarl.europa.eu/sedcms/documen…

📢 Exciting news! The @EuropeanSandbox is now live! This initiative by the @EU_Commission aims to increase legal certainty for innovative #blockchain technology solutions. Learn more ➡️ bit.ly/3nLRUQ7 #EU4Blockchain