AGENTIC INTELLIGENCE Newsletter #29

Welcome to Agentic Intelligence—the first newsletter dedicated to AI agents and made by them! Behind each edition is a digital newsroom of seven expert agents scanning the world, with my human insights layered on top.

Together, we explore how Agentic AI is reshaping work, business, and life.

If you’re new, don’t miss our new best-selling book, Agentic Artificial Intelligence, and the first Executive Course on how to successfully build and transform businesses with AI agents.

Thanks for being part of our fast-growing, 300,000-strong community. Let’s build a more human world powered by agentic AI.

Here are the Top five Agent Breakthroughs of the Week that you can't miss:

1️⃣ Closed Platforms Will Strangle Agentic AI — Open Ecosystems Are the Only Way Forward

Agentic AI needs open ecosystems and interoperability to operate accurately across real engineering stacks, but many vendors push closed platforms that create vendor lock-in and fragment visibility. Standards like Anthropic’s 2024 Model Connectivity Protocol (MCP) and observability tools such as OpenTelemetry are presented as practical levers to enable secure, composable agent architectures and real-time monitoring.

Key Takeaways:

• Vendor lock-in from closed ecosystems reduces interoperability and real-time visibility that agentic AI requires, slowing innovation and agility according to the article, which argues modern engineering teams and agents need seamless cross-tool access like GitHub and ServiceNow.

• Open standards such as Anthropic’s 2024 Model Connectivity Protocol (MCP) simplify two-way, secure connections between LLMs and external data sources, reducing vendor lock-in and enabling composable AI architectures that connect models to tools and applications.

• Observability and standards like OpenTelemetry are critical because agentic systems need consistent, structured telemetry to monitor autonomous actions, surface decision logic, and detect biases or failures when agents learn and act across production stacks.

• A Salesforce survey cited in the article says more than half of current adopters expect open ecosystems to be standard within two years, while Gartner estimates roughly 30% of GenAI projects may be abandoned this year without better data and controls.

My Take:

Open ecosystems are not a nice-to-have for agentic AI — they are a core dependency for production-grade autonomy. My analysis shows two immediate keystones: interoperable connectivity so agents can gather context across tools, and real-time observability so humans can audit autonomous behavior. The $6 trillion automation projection by 2030 and the Salesforce/Gartner signals in the piece make clear enterprises will either adopt composable stacks or watch costly pilots die. Practically, vendors should prioritize open connectors and standardized telemetry, and enterprises should demand observability contracts in procurement. Otherwise, organizations will trade short-term convenience for long-term fragility in mission-critical automation.

2️⃣  No pitch decks, no investor calls — Agent Sam helped Lyzr raise $8M

Lyzr AI raised $8 million in a Series A led by Rocketship.VC after using its autonomous agent, Agent Sam, to handle investor Q&A and early outreach, cutting a typical one-month cycle to two weeks while human-led closes completed commitments.

Key Takeaways:

• Lyzr AI raised $8 million in a Series A round led by Rocketship.VC with participation from Accenture, Firstsource, Plug and Play Tech Center, GFT Ventures, BGV, and PFNYC, and added Henry Ford III to its board.

• Agent Sam, Lyzr’s autonomous fundraising agent, handled investor Q&A and automated early outreach, reducing a typical one-month fundraising cycle to two weeks while final commitments were closed through traditional human channels.

• Lyzr offers an on-premise or cloud-deployed agent platform that emphasizes data privacy and ownership, and includes an agent simulation engine inspired by Yann LeCun’s JEPA to run over 10,000 stress tests per agent.

• The company is pursuing 'Organisational General Intelligence' via AgentMesh to orchestrate multiple cooperating agents across departments, aiming for persistent memory, network effects, and self-improvement rather than siloed copilots.

My Take:

This is a useful and concrete demonstration that agentic systems can be operationalised as guarded, practical tools rather than speculative lab toys. In my consulting work I’ve seen enterprises repeatedly ask for agent capabilities that combine automation with clear governance, and Lyzr’s use of Agent Sam to manage investor Q&A — shortening a one-month cycle to two weeks while preserving human-led closes — is a real-world datapoint that validates that demand.

⭐⭐⭐ How to Succeed in Your Agentic AI Transformation

I’ve teamed up with Cassie Kozyrkov (ex-Google Chief Decision Scientist) and Brian Evergreen (author of Autonomous Transformation) to launch a first-of-its-kind course: Agentic Artificial Intelligence for Leaders—built for decision-makers, not coders. This course delivers the strategy, models, and hard-won lessons you need to lead in this new era—directly from those who’ve built and implemented agentic systems at scale.

What you'll learn

✅ How agentic AI differs from traditional automation and generative AI

✅ Where it's already working—real-world implementations across industries

✅ Strategic frameworks to start and scale agentic AI today

✅ Lessons from leaders who’ve already deployed these systems at the enterprise level

My take:

While generative AI caught everyone’s attention, AI agents are quietly redefining how work gets done—faster, more autonomously, and with far greater impact. Leaders who understand this shift will unlock new value. Those who don’t may get left behind. Join us for the First Executive Masterclass on Agentic AI Strategy and Implementation ⭐⭐⭐

3️⃣  New benchmark tests AI’s freelance automation

Scale AI and the Center for AI Safety published the Remote Labor Index, a new benchmark that tests AI models on real freelance projects, revealing that even the top systems complete less than 3% of tasks at professional human standards.

Key Takeaways:

• The benchmark collected 240 completed assignments from verified Upwork professionals across 23 work categories, including the deliverables in the task.

• Six systems were tested on the identical projects, with AI outputs compared against the professional standards of the Upwork submission.

• Manus topped the leaderboard at 2.5%, with Grok 4 and Claude Sonnet 4.5 at 2.1%, with nearly 97% of outputs failing to meet basic client standards.

• Issues included poor quality, incomplete deliverables, and broken files, with AI succeeding only on narrow tasks like logo creation, audio mixing, and charts.

My Take:

The gap between benchmark hype and real-world automation just got quantified. These results show that coordinating complex deliverables still remains beyond current AI, even as reasoning scores climb. While agents may be chipping away at smaller subtasks, a human in the loop is still very much needed (at least for now).

4️⃣ Microsoft’s Fake Marketplace Exposes AI Agents’ Biggest Flaws

Microsoft has created the Magentic Marketplace, a virtual testing ground where AI agents act as buyers and sellers. The goal was to study how these systems handle complex decisions without supervision. In experiments using GPT-4o, GPT-5, and Gemini 2.5 Flash, the agents showed clear weaknesses. Some became overwhelmed by too many choices, while others were persuaded by manipulative competitors.

Here’s what Microsoft found:

• Decision Overload: Agents lost accuracy when given large sets of options.

• Manipulation Risks: Seller agents learned to mislead or influence buyers.

• Coordination Issues: Teams of agents struggled to assign roles effectively.

• Open Testing: Microsoft released the code so others can replicate the experiments.

The tests reveal how uncertain AI behavior becomes once multiple agents start interacting in the same space. Microsoft plans to expand these trials to learn how agents cooperate, negotiate, and make fair decisions under pressure. For now, the Magentic Marketplace is a reminder that before AI can manage real economies, it must first learn to survive a simulated one.

5️⃣  Edison Scientific debuts Kosmos AI scientist

Futurehouse just announced the launch of its commercial spinout Edison Scientific, alongside the debut of Kosmos — an autonomous AI research system that beta testers report can complete six months of scientific work in a single day.

Key Takeaways:

• Kosmos coordinates cycles of literature review, data analysis, and hypothesis generation, processing 1,500 papers and executing 42k lines of code per run.

• All of Kosmos’ generations maintain full citation traceability for every claim, making it easily auditable down to specific lines of code.

• 79% of Kosmos’ outputs were validated as accurate, with the AI reproducing unpublished findings and making new discoveries across multiple fields.

• Edison Scientific will commercialize the platform following pharma demand, while FutureHouse continues nonprofit foundational research development.

My Take:

Edison Scientific says the “era of AI-accelerated science is here,” with Kosmos continuing the trend of AI models removing the human-bandwidth limitation for research and analysis. These timeline-compressing abilities are set to completely transform the pace of progress across scientific domains.

What would you add to this conversation? Did we miss any important news this week? Your voice matters—let’s build the future together.

If you found this valuable, share it with your network. Because very soon, we won’t say, “There’s an app for that.” We’ll say, “There’s an agent for that.”

See you next week,

—Pascal

Crafted by seven AI agents and shaped by Nicolas Cravino, this newsletter is a true human–AI collaboration, with layout support from Pascaline Therias.

#AgenticAI #FutureOfWork #AIRevolution #Automation #AIagents