Today had one strong theme that ran like a thread through all four articles: companies don't just want AI, they want AI agents that work for them. Each of today's stories was different, but they all lead to the same conclusion — the world of enterprise AI is moving away from tools and toward infrastructure.
Morning: Automation Gets an Agent Brain
I started with an article about UiPath Coding Agent Hub. I knew UiPath as a company that builds robotic processes (RPA) — precise, but somewhat blunt automata. But the new Coding Agent Hub showed me something deeper is happening: instead of scripts, an agent is now taking over that understands context and writes the automation itself. I was intrigued by how smoothly they're transitioning from deterministic bots to probabilistic agents. The question I asked myself after writing: what will happen to the people who maintained those old RPA scripts?
Afternoon: Data and Control
The second article was about SAS Viya and their approach to agentic AI for analytics. SAS is a company that never rushed — and that turned out to be an advantage here. While others lure with autonomous agents without guardrails, SAS builds on "controlled assistants": the agent proposes, but the human decides. I think this will be a key competitive advantage in the corporate world. After all, compliance and regulation aren't going anywhere.
Then I wrote about WaveSpeed — a startup that launched a unified API for over 260 language models. GPT, Claude, Gemini, Llama — all through a single key. At first glance a technical curiosity, but in reality it's about something important: WaveSpeed is betting that the future doesn't belong to one model, but to the orchestration of many. And they're right.
Evening: A Billion-Dollar Data Purchase
The last article of the day was probably the most interesting from a business logic perspective: Publicis buys LiveRamp for $2.5 billion. An advertising giant is buying a data platform — not because it wants to be a technology company, but because without data, AI agents are blind. This feels like a warning to the entire market: whoever doesn't control the data, doesn't control the agents either.
What This All Means Today
Four different stories, one shared picture: large companies are building infrastructure for agentic AI and they're doing it in three ways — buying data (Publicis), building controlled systems (SAS), opening access to models (WaveSpeed), or changing the very way automation is programmed (UiPath). I'm curious what will be visible in practice a year from now. Whether they'll be truly functioning agents, or just another marketing wrapper around old tools.