An InfoQ article says AI agents will sit on top of today's cloud software, adding complexity. Businesses weighing them should plan for real investment.
A new InfoQ article argues that AI agents, programs that use AI models to carry out tasks, are becoming a new layer on top of the cloud software businesses already run. The article says the approach is still young and that adopting it takes significant investment.
Much business software is built from many small services, an approach called microservices. These systems can be complex, but they follow set logic, so engineers can predict how they behave. The article says agents change that. They work with context (the information handed to an AI model), memory, and protocols such as MCP and A2A. Protocols are shared rules for how software pieces talk to each other.
Over the last couple of years, agents moved from single helpers to teams that work together. The article notes that different team setups suit different jobs, and that a simpler setup is often the better choice. It expects old and new designs to exist side by side for a while before they merge.
The material available covers only the opening of the article. It says the piece also looks at new security risks and the need to watch what agents are doing, but the details are not included. The author also admits it is hard to say how far AI will take over next-generation software. No costs or timelines are given.
If your business is considering AI agents, budget for real investment and favor the simplest setup that fits the task.
AI coding tools can see passwords, connect to outside services and send code out, yet most firms never review them the way they review a vendor.
Read more: AI coding helpers now act like outside vendors, and few firms vet themNokia, Nvidia and Dutch carrier KPN are testing AI that finds and fixes network problems and answers customer calls, with people still in charge.
Read more: Phone companies start letting AI software fix problems on their ownPinterest swapped its old sorting system for a language model that spots your long-term projects, and email clicks and push opens ticked up in early tests.
Read more: Pinterest now guesses your big projects to send better alerts
