← All posts

November 5, 2026 · 2 min read

Only 6% of people want AI to decide for them. Does your AI agent suggest or decide?

New Okta research found that only 6% of consumers want AI to make final decisions for them, while 37% prefer AI suggestions and decide themselves. Here is how a small business can set up its AI agent to suggest first and let people choose.

When your AI agent helps a customer, does it make the decision, or does it hand them a good suggestion?

That small difference matters more than most of us think. Okta, an identity software company, just released its 2026 Customer Identity Trends Report. It surveyed 5,250 consumers, plus 910 developers and senior business leaders, across seven markets.

Here is how people said they want AI to help them:

  • Only 6% want AI to make final decisions and act entirely on their behalf.
  • 32% are comfortable letting AI handle routine tasks while they keep control of important decisions.
  • 37% prefer to get AI suggestions and make the decision themselves.
  • 25% want little or no AI involvement.

The report also found that 52% of consumers believe AI in apps and websites is advancing too quickly. And when people can approve an action first or undo it afterward, their willingness to let AI act rises above 50% for many uses.

On the business side, nearly 60% of organizations are introducing AI agents that take actions, and 80% of those efforts are running behind schedule.

Why this matters for a small business

Many of us picture an AI agent as something that does the whole job alone. But most of your customers are asking for something gentler. They want help, and they want to stay in charge. A coach, a consultant, or an attorney already knows this. Clients trust you because you guide them and let them decide.

Your AI agent can work the same way. Let it do the busy work, like drafting, sorting, and preparing options. Let people make the final call. That is what I mean by automate the process, not the relationship.

What I would do this week

  • Pick one task your AI agent handles with customers and change it from "decide" to "suggest," such as offering three meeting times instead of booking one.
  • Add a simple approve step before anything that costs money or changes an appointment.
  • Ask two or three customers how much they want AI to do for them, and listen to their answer.

My takeaway: Most people do not want AI to decide for them. They want it to help them decide. Build your AI agent to suggest first, and let your customers stay in charge.

Sources

  1. Closing the AI trust gap: Why 80% of agentic commerce projects are stalled (and how to fix it) · Okta (vendor research, November 4, 2026)
  2. Okta Survey Finds Consumers Want Approval Controls Over Autonomous AI Agents · CDO Magazine (November 5, 2026)
  3. Most consumers aren't ready to let AI agents shop unsupervised · Help Net Security (November 5, 2026)