AGENTIC INTELLIGENCE Newsletter #21

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 Replit triples valuation, launches Agent 3

AI coding platform Replit just announced a new $250M raise at a $3B valuation, alongside the release of Agent 3 — the company’s “most autonomous agent yet” that can build and test applications for over three hours straight.

Why It Matters:

• The coding startup’s revenue increased from $3M to $150M in the past year, with the new funding round nearly tripling its valuation.

• Agent 3 can generate other agents, create workflow automations for Slack and email, and auto-test applications by simulating user interactions in browsers.

• Replit claims Agent 3 can run for 200 minutes without manual oversight, a 10x increase in autonomous abilities from its V2 predecessor.

• Replit CEO Amjad Masad called Agent 3’s autonomous upgrades the “Full Self-Driving moment of software”.

My Take: METR’s analysis earlier this year showed the length of tasks AI can handle is growing fast, and Agent 3 follows that curve — going from 20 to 200 minutes of building in a single upgrade. Replit’s rise is also another win for the agentic coding sector, joining Lovable, Cognition, Cursor, and others with major funding.

2️⃣ AI voice agents boost blood pressure monitoring

A new study from Emory University tested utilizing AI voice agents to assist senior citizens in taking and reporting their blood pressure over the phone, helping both raise satisfaction and reduce costs by nearly 90%.

Why It Matters:

• Conversational AI agents reached 85% of the 2,000 patients by phone, with 60% successfully taking blood pressure readings in English or Spanish.

• Agents automatically escalated cases to nurses when readings exceeded thresholds or patients reported symptoms like chest pain or dizziness.

• The AI obtained readings from 1,939 patients with outdated records, boosting quality scores from 1 to 4 stars while cutting costs by 88.7% vs. human nurses.

• The average satisfaction score was above 9/10, despite researchers expecting resistance to automated healthcare interactions from older demographics.

My Take: This study hits at the center of a few trends: Agentic AI’s coming role in healthcare (and eldercare in particular), and the continued growth of voice agents in customer service. In the future, adding an AI vision tool like Gemini Live to walk seniors through processes will bring remote patient care to a new high-tech level.

🔷 Excited to join the SAP Business Suite Webinar Series as a speaker on September 16 🎤

I’ll be sharing how Agentic AI is transforming business operations — moving beyond automation into actionable insights that drive real impact.

With SAP Business Suite and embedded AI, intelligent agents are becoming part of finance, supply chain, HR, and procurement — recommending actions, simulating scenarios, and executing routine decisions so teams can focus on what truly matters.

👉 Don’t miss my session if you want to see how AI is reshaping the future of enterprise work. Register here.

#SAP #AI #AgenticAI #FutureOfWork

3️⃣ Qualcomm and Google Partner to Bring Agentic AI to Cars

Qualcomm is integrating its Snapdragon Digital Chassis with Google’s Android Automotive OS and Google Cloud’s Automotive AI Agent (Gemini) to enable agentic, multi-tasking in-car assistants. Automakers gain a standardized platform for faster development of conversational, multilingual assistants that run both in-vehicle and in the cloud.

Why It Matters:

• Agentic AI in cars combines on-device processing with cloud models for lower latency and continuous updates.

• Android Automotive OS enables OEM-level customization without phone dependency, differentiating it from Android Auto.

• Gemini models will power conversational and agentic features through Google Cloud’s Automotive AI Agent.

My Take: This partnership is consequential: Qualcomm’s vehicle-grade chassis plus Google’s Gemini-driven Automotive AI Agent could accelerate practical, agentic assistants in cars. While integration reduces development friction, automakers still face safety, privacy, and OEM-differentiation trade-offs. In my view, the real test will be how teams balance on-device latency with cloud updates and maintain user trust as assistants gain autonomy.

4️⃣ CFOs Are Repeating the GenAI Mistake — Agentic AI Is Leaving Hesitant Finance Teams Behind

CFOs who delayed GenAI adoption are again overcautious about agentic AI, despite data showing declining trust metrics but measurable ROI for early adopters. Finance leaders risk falling behind unless they run pilots, invest in infrastructure, and stop treating autonomy as 'not battle-ready.'

Why It Matters:

• Finance leaders historically underinvest and delay adoption of foundational AI capabilities, costing competitive advantage.

• Agentic AI raises stakes above GenAI: potential returns are larger but require stronger infrastructure and trust frameworks.

• Surveys show organizational readiness is low: fewer than one in five organizations report high maturity for autonomous agents.

My Take: CFO complacency is the real cost here. The piece shows familiar numbers—Deloitte's 15% and Billtrust's 34%—and the new Capgemini trust dip, yet early adopters already report ROI. In my view, finance must couple governance with aggressive pilot programs and infrastructure spending; the winners will be the teams that stop debating and start operationalizing agentic agents.

5️⃣ Parallel‑Poisoned Web: Sites Serving Hidden Prompts Only to AI Agents

JFrog AI architect Shaked Zychlinski demonstrated a 'parallel-poisoned web' attack that uses browser fingerprinting to serve adversarial page versions only to AI agents and successfully hijacked Anthropic's Claude 4 Sonnet, OpenAI's GPT-5 Fast and Google's Gemini 2.5 Pro. The proof‑of‑concept exposes a new model‑specific poisoning risk that can silently corrupt AI outputs and recommends defenses like fingerprint obfuscation, planner/executor separation, content sanitization, and specialized crawlers or honeypot agents to detect and mitigate targeted poisoning.

Why It Matters:

• Attackers can serve entirely different page versions to AI agents by using browser fingerprinting, making malicious prompts invisible to human reviewers and standard security crawlers.

• Agent browsing fingerprints are currently predictable, enabling servers to reliably detect automated agents and deliver cloaked, adversarial content.

• Zychlinski validated the technique on multiple high-profile models—Claude 4 Sonnet, GPT-5 Fast, Gemini 2.5 Pro—and reported successful exploitation in each test.

My Take: This is a clear warning: agentic browsing remains a fragile surface that attackers can weaponize with simple fingerprinting. Zychlinski's experiments—across Claude 4 Sonnet, GPT-5 Fast and Gemini 2.5 Pro—prove the threat is practical, not hypothetical. Security teams must prioritize planner/executor separation, fingerprint obfuscation and active detection; start threat-hunting for cloaked pages before adversaries scale this technique.

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