AI agents make agreeable coworkers, and that's a problem for solo business owners. Here are 3 ways they're fixing it.

• Overly agreeable AI agents as coworkers present challenges for startup founders.
• Three solopreneurs shared what systems and prompts help them get pushback from their AI agents.
• This article is part of "The AI-Powered Solopreneur," a series exploring how solo business owners use AI to drive growth.

Will AI agents be the employees of the year in 2026?

Only time will tell. Last year, one category AI absolutely dominated was being an extremely agreeable coworker. While this might sound nice, this can turn into a problem for founders who rely on AI as their only teammate.

When your head of legal, HR, and supply operations are all AI agents, unsubstantiated flattery can create costly blind spots. That's one of the reasons OpenAI said goodbye to its "yes-man" version of ChatGPT, and why some AI-powered solo founders are training their tools to push back.

Three business owners who rely on AI daily shared with Business Insider how they've built workflows and prompts that force their AI employees to challenge their ideas. Quotes have been edited for length and clarity.

I avoid asking AI what it thinks and instead ask it to rate my idea on a scale of 1-10

Yesim Saydan is a branding and communication expert in her early 50s, based in the Netherlands.

When OpenAI launched custom GPTs, everything changed for me. I used the feature to create over 17 custom GPTs to build my team. Then I thought of my ideal mentors and created custom GPTs of them.

When I'm stuck on a business decision or need to come up with a creative idea or strategy, brainstorming usually starts with my Steve Jobs custom GPT. When I prompt it, I avoid asking questions like "What do you think of this idea?" because the AI usually wants to agree with me and please me. Instead, I ask, "On a scale from one to 10, how good is this idea?"

It's not going to say the idea is bad, but now it might tell me it's a five. Then I'll ask, "OK, what would make it a 10?"

That's usually when it starts drawing on the experience of Steve Jobs that I've trained it with. We can go back and forth until I get the most useful and honest feedback possible. Depending on the task, I usually go through three to five rounds of refinement for more strategic outputs.

I created a governance structure and a roundtable with my AI agents to get better responses

Aaron Sneed is a 40-year-old defense-tech solo founder based in Florida.

When I started my business, I realized I didn't have the money to pay lawyers, HR reps, and a bunch of other companies. So, using AI, I created what I call 'The Council.'

Altogether, my AI council consists of the following:

• chief of staff agent
• HR agent
• finance agent
• accounting agent
• legal, comms, and PR agents
• security and compliance agent
• engineering agent
• quality agent
• supply chain agent
• training agent
• manufacturing agent
• business systems agent
• facilities agent
• field operations agent
• IT and data agent

I don't want a bunch of yes agents. I trained them purposefully to give me pushback. I want them to test my theories to help me with what I'm trying to accomplish. The training never really stops.

The models have gotten better, and my prompting has, too. I have a better understanding of what information should be in an agent, like having a governance structure for priorities. I have a set of files that put those requirements in place to mitigate the risk of hallucination and false or bad information.

I told my chief of staff agent which models have priority when making decisions. For example, anything legal, compliance, or security-related will be given a higher priority. So, I tell the chief of staff to listen to these models over everyone else.

I have a roundtable set up in ChatGPT's projects section with all my AI agents, including my chief of staff, where I can put something like a request-for-proposal document in the chat, and all the agents will weigh in at the same time. I use this roundtable as a level of prevention for hallucinations and knowledge gaps.

I use an 'AI conveyor belt' as one way to mitigate only hearing positive feedback

Tim Desoto is a 49-year-old founder and CEO, based in San Francisco.

I've been working with AI to launch my startup since late 2024. I don't have a tech background, and since starting my business, I've learned a lot about where to leverage AI, and where not to.

Taking whatever I'm building or ideating and having AI push against it, either as I'm thinking about the idea or afterward, is part of an exercise I call my AI conveyor belt. Usually, I start with a written prompt, then go multimodal, talking out loud to the model. I'll talk back and forth with it about my idea and try to get the agent to push back because I know that some AI models tend to be more agreeable.

Once I get an output that I'm happy with. I use a different model to get a different view. I'll drop the document in and go back and forth with the new model. Sometimes, I'll push a document out to both models at the same time and see what comes back. Some are better than others at giving feedback, researching, and annotating, but I'm always getting a better, more well-rounded perspective by feeding content to multiple models at a time.

Do you have a story to share about running an AI-powered business? Contact this reporter, Agnes Applegate, at aapplegate@businessinsider.com.