Agents Orchestrate. Trust Emerges. Systems Transform.

What happened
IBM unveiled Bob, an orchestration platform embedding specialized, role-based AI agents throughout the software development lifecycle. Bob dynamically routes tasks to the best models, enforces compliance, and enables human oversight at key decision points. In testing, it compressed a 30-day Java upgrade to just three days and 80,000+ IBM employees reported a 45% productivity boost.

Why it matters
Bob demonstrates how agentic orchestration can restructure multi-step enterprise processes without sacrificing auditability or risk controls, the missing link for regulated industries.

What’s next
IBM is rolling Bob out globally as SaaS, with on-premises support coming for regulated industries and a migration path for legacy code assistant users.

What happened
The FIDO Alliance formed an Agentic Authentication Working Group and a Payments Technical Working Group to create open standards for AI agent interactions and commerce. Initial contributions from Google’s Agent Payments Protocol and Mastercard’s Verifiable Intent will form the foundation of specifications that define how AI agents authenticate, act and transact on behalf of users. The effort addresses gaps in user instruction, agent authentication and trusted delegation, which are challenges because current authentication models are built for human interactions.

Why it matters
Without trusted standards, users might have to share credentials and service providers lack reliable ways to verify intent. FIDO’s initiative introduces phishing‑resistant mechanisms, agent authentication and verifiable authorization, laying the groundwork for a projected $5 trillion agentic commerce market. By bringing industry players together, it aims to make agent‑driven transactions secure and interoperable as AI agents go mainstream.

What’s next
The two working groups will develop specifications, review contributions like AP2 and Verifiable Intent and collaborate with other standards bodies; FIDO plans to publish updates as the standards evolve. Broad participation suggests trusted agentic protocols could emerge soon, enabling secure autonomous purchasing and delegation at scale.

What happened
AWS introduced Amazon Connect Decisions, an agentic supply‑chain planning tool that bundles more than 25 specialized tools into AI agents called “teammates.” The agents centralize data from multiple systems, triage thousands of alerts, translate demand forecasts and provide chat and visual interfaces so planners can ask questions and receive recommendations. Built on Amazon’s experience managing over 400 million SKUs, the system learns from user actions to improve decisions.

Why it matters
Supply‑chain teams struggle to aggregate data and act quickly. Connect Decisions compresses decision cycles from days to hours, offering root‑cause analysis and alert prioritization. By continuously learning from users, the agents could reduce inventory costs and speed responses to demand fluctuations.

What’s next
AWS is rolling out the service and integrating it with existing enterprise planning systems; as more organizations adopt it, the AI teammates will learn from diverse patterns and further optimize supply chains. The product signals a broader trend of embedding agentic AI into operational infrastructure.

What happened
CAS, a division of the American Chemical Society, launched CAS Newton, a conversational, multi‑agent AI integrated into SciFinder and BioFinder. Grounded in 150 years of curated literature, Newton uses a multi‑agent framework to connect tools for spectra prediction, structure search and functional‑group reactivity prediction, retaining context as complex queries evolve.

Why it matters
Researchers typically sift through dozens of search results. Newton’s architecture delivers directly actionable answers, enabling scientists to ask multipart questions and receive contextual responses. By keeping user queries private and audit‑able, it aims to provide trustworthy AI assistance while avoiding hallucination.

What’s next
CAS plans secure integrations so organizations can use Newton alongside proprietary data and is developing industry‑specific enhancements. If widely adopted, multi‑agent AI could accelerate drug discovery and materials research by seamlessly connecting data and tools.

What happened
At the Laserfiche Empower conference, the company introduced AI Agents that let users automate complex, multi‑step document tasks via natural‑language prompts. Through the Smart Chat interface, the agents can route contracts, tag records, flag overdue invoices and act on document data while inheriting user permissions within Laserfiche’s governance framework. They combine AI‑powered content analysis with intelligent automation to surface conditions across repositories.

Why it matters
Enterprises spend significant time on manual document management. AI Agents offload mundane work while maintaining compliance, allowing organizations to modernize operations without overburdening IT. By democratizing automation, non‑technical users can design workflows quickly and free up resources.

What’s next
Laserfiche plans to expand the agents beyond one‑time actions into background processes and embedded business workflows, where they monitor systems and act autonomously. General availability for all Laserfiche Cloud users starts May 7, 2026.

What happened
PTC released Windchill AI Assistant, a generative AI chat interface embedded in its product lifecycle management platform. Users can ask natural‑language questions to find information, review summaries and surface key details from complex product documents while the assistant maintains access controls and references sources.

Why it matters
Product teams often waste hours digging through technical documents. The Windchill AI assistant reduces search time by delivering contextual answers and summarizations, improving productivity and helping engineers apply insights faster. Grounding responses in existing data and enforcing security builds trust in AI‑driven PLM.

What’s next
PTC plans to extend the assistant to other domains like parts and change management and embed AI‑driven actions directly into workflows. The broader vision is an intelligent product lifecycle where AI helps manage complex data end‑to‑end.

What happened
Transporeon, a Trimble company, launched Natural Language Search for its Autonomous Procurement platform, enabling carriers to type conversational queries such as “reefer loads from Houston to Atlanta in 48 hours” and instantly see matching loads with AI‑powered Buy‑It‑Now offers. The generative feature interprets user intent with over 90% accuracy, eliminating the need for multi‑step filtering.

Why it matters
Freight procurement traditionally requires navigating complex filters; the conversational interface reduces search time and has increased query‑to‑booking conversion by up to 25% in pilot programs. It exemplifies how generative AI can streamline logistics and improve marketplace efficiency.

What’s next
After successful pilots involving 790 carriers, the feature is now available to all Autonomous Procurement users across Europe and North America. Its adoption could encourage other transportation platforms to integrate AI‑driven search and booking.

What happened
At the SAS Innovate conference, SAS announced a set of platform updates including AI Navigator—a SaaS tool to inventory, monitor and govern AI models and agents—and a new Supply Chain Agent that automates sales and operations planning via conversational interfaces. The agent simulates scenarios, forecasts demand and adjusts supply strategies in near real time, while AI Navigator aims to mitigate risks from “shadow AI” by providing a unified view of AI assets.

Why it matters
Many enterprises struggle to scale AI responsibly. SAS’s combination of governance tooling, embedded copilots and an agent framework moves organizations from isolated experiments to production‑scale deployments. Continuous planning agents could make supply chains more agile and data‑driven.

What’s next
AI Navigator is in private preview with general availability expected in Q3 2026. SAS plans to expand its portfolio of industry‑specific agents and expose analytics capabilities to external agents via an open protocol, signaling a future of interconnected, governed AI ecosystems.

What happened
UniX AI’s Panther service robot has begun real‑home deployment, completing chores such as waking users, making beds, preparing breakfast, cleaning and organizing objects. Equipped with mass‑produced 8‑degree‑of‑freedom robotic arms, an omnidirectional four‑wheel base and 2070 TOPS of computing power, Panther navigates narrow spaces and handles payloads up to 12 kg.

Why it matters
Unlike earlier humanoid robots, Panther uses proprietary AI modules—UniFlex for cross‑space generalization, UniTouch for multimodal perception and UniCortex for long‑horizon planning—to understand unfamiliar spaces, assess objects and decompose tasks. Its move from lab tests to real homes signals a step toward consumer‑ready robotics.

What’s next
UniX AI’s second‑generation product achieved stable mass production in 2025, and the company plans to expand Panther to the U.S., Europe and the Middle East. Success could pressure rivals like 1X Technologies and Figure AI, which have yet to deploy fully autonomous robots in homes.

What happened
Innoviz, a leading automotive LiDAR company, is expanding into defense and homeland security, repurposing its high-performance sensors for mission-critical environments. Its InnovizSMART and InnovizTwo Ultra Long-Range systems originally built for autonomous vehicles, can detect objects up to 1 kilometer away and generate real-time 3D maps of complex environments.

Why it matters
This is a classic “civilian tech → defense upgrade” moment. Physical AI systems—whether drones, robots, or surveillance platforms—depend on accurate, real-time perception to operate safely. Innoviz’s sensors offer long-range detection, weather resilience, and high-resolution spatial awareness, making them viable for perimeter security, drone detection, and battlefield situational awareness. In short: better sensors → better world models → smarter autonomous systems.

What’s next
Innoviz is already shipping products and targeting governments, airports, and security agencies. Expect more crossover: automotive-grade hardware is becoming the backbone of defense AI systems. As physical AI scales, the real competition won’t just be models—it’ll be who sees the world best.

Agents are no longer experiments—they’re being embedded into the core systems that run enterprises, science, commerce, and even the home. The real shift isn’t intelligence—it’s orchestration, trust, and integration at scale, where the winners will be those who control how agents act, decide, and transact.

Run a “mini agent marketplace” test using tools you already have. Take a real task—like analyzing a contract, planning inventory, or researching a molecule—and route it across 2–3 different models (e.g., ChatGPT, Claude, Gemini) for different roles: one generates, one critiques, one summarizes.

Then add a simple “trust layer” require each step to cite sources or justify decisions before passing to the next. You’ll start to feel what platforms like IBM’s orchestration layer and FIDO’s emerging standards are solving—coordinating agents is easy, trusting them is the hard part.