Let’s begin with the time period “agent” itself. Proper now, it’s being slapped on all the pieces from easy scripts to classy AI workflows. There’s no shared definition, which leaves loads of room for firms to market fundamental automation as one thing rather more superior. That type of “agentwashing” doesn’t simply confuse clients; it invitations disappointment. We don’t essentially want a inflexible customary, however we do want clearer expectations about what these programs are imagined to do, how autonomously they function, and the way reliably they carry out.
And reliability is the following massive problem. Most of in the present day’s brokers are powered by giant language fashions (LLMs), which generate probabilistic responses. These programs are highly effective, however they’re additionally unpredictable. They will make issues up, go off observe, or fail in refined methods—particularly once they’re requested to finish multistep duties, pulling in exterior instruments and chaining LLM responses collectively. A current instance: Customers of Cursor, a preferred AI programming assistant, had been instructed by an automatic assist agent that they couldn’t use the software program on multiple gadget. There have been widespread complaints and stories of customers canceling their subscriptions. But it surely turned out the coverage didn’t exist. The AI had invented it.
In enterprise settings, this sort of mistake might create immense harm. We have to cease treating LLMs as standalone merchandise and begin constructing full programs round them—programs that account for uncertainty, monitor outputs, handle prices, and layer in guardrails for security and accuracy. These measures might help be certain that the output adheres to the necessities expressed by the person, obeys the corporate’s insurance policies relating to entry to data, respects privateness points, and so forth. Some firms, together with AI21 (which I cofounded and which has obtained funding from Google), are already transferring in that route, wrapping language fashions in additional deliberate, structured architectures. Our newest launch, Maestro, is designed for enterprise reliability, combining LLMs with firm knowledge, public data, and different instruments to make sure reliable outputs.
Nonetheless, even the neatest agent received’t be helpful in a vacuum. For the agent mannequin to work, totally different brokers have to cooperate (reserving your journey, checking the climate, submitting your expense report) with out fixed human supervision. That’s the place Google’s A2A protocol is available in. It’s meant to be a common language that lets brokers share what they’ll do and divide up duties. In precept, it’s an ideal thought.
In observe, A2A nonetheless falls brief. It defines how brokers speak to one another, however not what they really imply. If one agent says it might probably present “wind circumstances,” one other has to guess whether or not that’s helpful for evaluating climate on a flight route. With out a shared vocabulary or context, coordination turns into brittle. We’ve seen this drawback earlier than in distributed computing. Fixing it at scale is much from trivial.
