What's happening in the world of AI agents
Every week we distill the signal from the people who actually know what's coming — into clear, actionable reads. To understand why agents are already replacing tasks, not people, and how to put one to work in your business.
Start here
MCP and A2A under the same roof: for the first time, agents can talk to each other without depending on any single vendor
On August 20, 2026, Google formally transferred the A2A protocol to the Agentic AI Foundation (AAIF) — the same foundation that already hosts Anthropic's MCP. For the first time, the two standards that define how agents connect to tools and how they coordinate with each other share neutral governance, backed by over 250 organizations. Closing that gap isn't a PR move: it's the missing piece for companies building agents that work together without vendor lock-in.
Gemini hit one billion users this week. Only 23% of companies have agents running.
On August 11, Sundar Pichai announced that Gemini reached one billion monthly active users. That same week, Deloitte published the stat that matters most to any business without agents yet: only 23% use them today in any real capacity. 74% will within two years. The curve now has a deadline.
The AI agent that doesn't introduce itself as AI is now illegal in Europe — and here's what that changes
On August 2, 2026, Article 50 of the EU AI Regulation became enforceable: every AI system that speaks with people must identify itself as AI from the first message. The same week, Microsoft reported 40 million active agents — outnumbering its 30 million human Copilot users. Two data points that define a new minimum standard for any company deploying agents.
88% of companies already had an incident with their AI agents — and 82% of their executives think they're protected
Two 2026 studies with more than 2,600 senior technical leaders surveyed reveal the same gap: 96% of companies already have agents deployed, but nearly half operate without real oversight. The risk doesn't come from the agent that fails — it comes from the one that works without anyone watching.
AI is no longer just talking — it's working: 2.4 billion tasks completed by agents
Salesforce coined a new unit of measure: the Agentic Work Unit (AWU) — a discrete task an AI agent completes autonomously: a decision made, a record updated, a workflow triggered. Their last quarter delivered 771 million of these units. Cumulative total: 2.4 billion. Not conversations — finished work.
The AI agent most companies use doesn't code — it consolidates, drafts, and organizes
Anthropic analyzed 1.2 million sessions of its autonomous agent tool and found that only 8.7% involve software development. The largest block — 33.4% — is pure operational work: reports, checklists, consolidating scattered data. The myth of agents as a tech team tool breaks apart with data from the lab itself.
AI agents grew 15x — and only 1 in 5 companies knows how to benefit
Microsoft just released its two most important annual studies: active agents in its ecosystem grew 15x in one year. The question is no longer whether agents work — it's that most organizations aren't ready to capture that value.
AI agent spending grew 139% in one year. It's no longer innovation — it's infrastructure.
Gartner documents $206.5 billion in AI agent software in 2026: a 139% jump in one year, three times faster than the overall AI market. Warner Bros., Yahoo and Tyson Foods are no longer piloting — they're operating. Here's what changed this week and what it means for any business.
What an employee really costs in Mexico vs. a synthetic employee
The salary is only 60% of the real cost of hiring in Mexico. The other 40% no one counts — until December. Here's what changes with a monthly-fee model that includes no benefits.
OpenAI and Anthropic spent $250M on agent consultants. The signal is clear.
In two coordinated moves, the world's most influential AI labs publicly admitted what many already sensed: the bottleneck is no longer the model — it's who knows how to deploy it inside a real business.
5 questions to ask before hiring an AI agent provider
95% of AI projects generate no measurable return. The difference between the 5% that scale and the 95% that don't almost always comes down to five questions most companies never ask before signing.
Agent teams are now enterprise standard: Salesforce activates multi-agent orchestration today
Starting today, 18,000 companies can deploy full AI agent teams with a click. The market has reached its inflection point — but Gartner warns that 40% of projects will fail without proper governance.
AI agents stopped being an experiment: in 2026 they are already employees
80% of enterprise apps shipped this quarter already have an agent inside. Here is what changed — and why your competitors are already hiring synthetic employees.
5 signs your company needs an AI agent this month — not in 6 months
Most companies that hire an AI agent do so 8 months after the pain was already obvious. Here are the five signals that mean the moment is now.
95% of AI pilots never reach production. Here's what the 5% that do are doing differently.
The most-cited study of the year doesn't say AI doesn't work. It says something more uncomfortable: the problem is almost never the model. It's where —and how— it's deployed.
MCP: how an open protocol became the USB-C of AI agents
In under a year it went from an internal Anthropic experiment to a standard adopted by OpenAI, Google and Microsoft. Why it matters for anyone connecting an agent to their business.
Voice agents crossed the line: 2026 is the year the phone actually resolves
Voice stopped being the IVR you hated. The 2026 numbers show agents that understand, resolve and cost a fraction — and why the phone channel matters again.
Which process in your company should an agent run?
Talk to Luz, our agent, instantly.
Free · instant · No commitment