Imagine your business needs a report. One artificial intelligence assistant can prepare it, but someone suggests having five collaborate instead. It sounds promising: more assistants should mean a better result.
The catch is that you could pay much more for essentially the same report.
That was the finding of research by Vals AI, a company that evaluates artificial intelligence systems. Its tests compared two advanced models, Claude Opus 5.5 and GPT-6 Sol, operating on their own and in teams.
Using teams was 1.8 to 5.1 times more expensive than using one agent. But just one of the four comparisons produced a statistically significant gain in quality.
When the models operated at their highest reasoning capacity, teaming up offered no meaningful benefit.
In short, putting more artificial intelligence on the job does not guarantee better answers.
Why it matters
Think of a restaurant using AI to organize information about its suppliers. Or a real estate business drafting property descriptions.
In either case, it is tempting to assume that several assistants checking one another's work will do a better job. But if the result is essentially unchanged, what does that additional spending actually buy you?
It may buy you speed.
Noam Brown, an OpenAI researcher, explained that having several agents work together can shorten the time needed for certain tasks, while also raising costs.
In the examples he gave, using four agents cut completion time in half while doubling the expense. Increasing the number to 16 brought a slight drop in efficiency.
The crucial distinction is this: getting something done sooner does not mean doing it better.
Some tasks are also harder to divide than others. Online research and mathematical problem-solving allow several agents to work at the same time. Writing something that must remain coherent from beginning to end, however, does not necessarily benefit in the same way.
Another expense is less obvious: keeping the assistants coordinated.
When several systems need to exchange information, check results and reach agreement, that coordination uses up some of their resources. Eric Provencher, an OpenAI developer, has cautioned about this problem.
For a small business, the extra complexity can create a cost that is difficult to justify.
What it means for your business
The takeaway is not that using several agents is a bad idea. It is that you need a specific reason to use them.
If you run a clinic and use AI to draft communications, first assess whether one assistant meets your needs.
If you own a store and need to research numerous products from different suppliers, several agents could help you collect information sooner.
But a more sophisticated solution is not necessarily a more cost-effective one.
Before paying for systems that promise to have multiple assistants work simultaneously, ask what improvement you can expect. Will the answers be more accurate? Will you wait less? What will that difference cost?
This week, pick a task you already handle with artificial intelligence. Try it with one assistant first, then with several if that option is available. Compare completion time, output quality and cost.
You may find that your business does not need additional agents. It needs to get more out of the one it already uses.