When Will We Actually Let AI Choose the Professional?

Booking someone to clean your gutters and hiring someone to redesign your home's heating system are both home-service purchases. That's about where the similarity ends.

For the first, you might be perfectly comfortable telling your AI agent to find a reputable company, confirm the price, and schedule the work. As long as the company is qualified and the cost is reasonable, there's not much you need to be involved in.

For the second, you might happily let the agent spend hours researching contractors, comparing experience, checking credentials, and narrowing the options. But before committing tens of thousands of dollars, you'd probably want to meet the contractor, walk through your home, and hear what they actually recommend.

The agent could do an extraordinary amount of useful work. You'd still want to make the final decision.

I think this distinction is going to shape how AI agents change the service economy. There will be services people are comfortable delegating almost entirely, and others where AI does most of the work leading up to a decision but never actually makes it.

The interesting question is where that boundary falls, and why.

People want AI to do the work. They're less sure about letting it decide.

There's already evidence of a gap between what consumers want AI to help with and what they're willing to let it control.

In May 2026, Gartner published a survey of 322 U.S. consumers. It found that 31% were willing to let AI narrow their choices for household supplies, while willingness to let AI actually make purchasing decisions topped out at just 11%, even in relatively low-stakes categories.

Forrester's June 2026 research found something similar. Roughly three-quarters of surveyed online adults in the U.S., UK, and Canada were uncomfortable letting AI agents autonomously complete purchases and payments, even with preset spending limits and rules. Consumers expressed more comfort with agents researching, comparing, and preparing purchases than actually completing them.

These studies concern shopping and payments, not hiring service professionals. We shouldn't assume the percentages translate directly to decisions about lawyers, dentists, contractors, or senior living communities.

But they highlight an important distinction: the ability to delegate a task doesn't mean people are ready to surrender the decision.

And I suspect the boundary will look very different depending on what they're hiring someone to do.

A customer might let an agent arrange routine vehicle maintenance at a previously approved repair shop, while insisting on choosing the specialist who will diagnose a complicated engine problem. Someone might authorize an agent to schedule a dental cleaning but want to personally evaluate treatment recommendations before committing to major restorative work.

The industry alone doesn't determine how much authority a customer will delegate. The nature of the assignment matters more.

Cost and risk are obvious considerations. So are reversibility, personal preferences, and whether the customer has an established relationship with the provider.

But there's another factor I think deserves more attention.

Can the decision actually be made from information that's available before the professional meets the customer?

Some of the most important information doesn't exist yet

Much of the conversation about AI evaluating businesses focuses on information that can be found and analyzed beforehand.

A law firm's practice areas. A dentist's credentials and procedural experience. A contractor's completed projects. A senior living community's services, licensing, staffing, and pricing.

These are all potentially useful facts. The more accurate and accessible they are, the better equipped an AI agent should be to distinguish promising providers from poor matches.

But consider what happens when someone actually meets the professional.

A homeowner invites three contractors to inspect a house before replacing its heating system. All three may have excellent qualifications and extensive experience with the relevant equipment. Their past work might be well documented, and an AI agent could have good reasons for including each one.

Then they walk through the house.

One recommends replacing the existing system entirely. Another identifies a way to preserve some of the existing equipment. A third notices a problem that may need to be addressed before either approach makes sense.

The homeowner isn't simply learning which contractor is friendlier or more persuasive. They're receiving new technical assessments, proposed solutions, and explanations of trade-offs.

Much of that information wasn't available when the agent assembled its shortlist.

The same thing happens in other professional services.

An attorney might have an excellent record handling business disputes. But once the attorney reviews your particular situation, you learn how they assess the strengths of the case, which risks concern them, and whether they believe litigation is even the right path.

A dentist might be highly experienced with implants. But the examination determines whether an implant is appropriate, whether additional treatment is required, and what alternatives the patient should consider.

A senior living community might meet every documented criterion on a family's list. But a visit and care assessment can reveal whether it can actually accommodate a particular resident's needs, and whether the family feels comfortable with how care would be delivered.

In each case, the interaction produces information that wasn't simply waiting online to be retrieved.

Sometimes, a good professional changes the question entirely.

You arrive thinking you need a new heating system and learn that a less expensive repair may solve the problem. You consult an attorney expecting to file a lawsuit and hear a convincing argument for negotiating first.

That ability to recognize that the customer's proposed solution may be wrong is itself an important part of professional judgment.

Economists have long studied this problem in markets for what they call credence goods: services where customers often cannot reliably determine what they need, or even fully assess what they received. Healthcare, legal services, and repairs are familiar examples. A survey of research published in CESifo Economic Studies examines the information imbalance between professionals who diagnose problems and customers who depend on their expertise.

AI may help narrow that imbalance. But it cannot research the outcome of a diagnosis that hasn't happened yet.

The human conversation isn't just about personality

It's tempting to say that people will keep meeting professionals because they want a personal connection.

There's truth to that. You probably care whether your attorney listens, whether a surgeon inspires confidence, or whether a contractor communicates clearly.

But reducing the human interaction to personality misses something important.

The conversation itself can reveal how a professional thinks.

Do they ask questions that demonstrate an understanding of the problem? Do they identify complications you hadn't considered? Can they explain why one approach might be better than another? Are they candid about uncertainty, limitations, or work they aren't qualified to perform?

And critically, how do they respond when challenged?

Those are potentially meaningful observations about how the professional might handle the engagement.

There's research supporting the importance of the type of information being exchanged. A 2026 study in the Journal of Retailing and Consumer Services, spanning four studies, found that consumers tended to prefer AI agents when disclosing objective information and human agents when disclosing subjective information.

The research examined customer-service interactions, not the hiring of professional service providers. Still, it suggests that the nature of the information people need to share may help explain why they want different forms of assistance at different stages.

A customer providing an address, budget, preferred appointment time, and required credentials may be comfortable working entirely through AI. Explaining fears about a medical procedure, concerns about a divorce, or competing priorities in a major renovation is a different kind of interaction.

Of course, human judgment has its own weaknesses. A confident, articulate professional isn't necessarily the most competent. A charismatic salesperson can win work that a quieter, more qualified competitor deserves.

That's one reason independently documented qualifications and experience matter so much. They give customers a way to evaluate professionals beyond how well they perform in a meeting.

The strongest decisions may combine both.

The agent's most important job may be deciding who gets the meeting

Consider what a capable personal AI agent might do for someone hiring a family law attorney.

It could research relevant firms, identify attorneys with experience in the appropriate area, examine credentials, compare available information about fees, and ask preliminary questions. It might even schedule consultations and prepare a concise brief for the customer before each meeting.

The customer could arrive knowing considerably more about each firm than they would have after browsing a handful of websites.

But the customer might still insist on speaking to the attorneys before choosing one.

That doesn't make the agent's contribution minor.

The agent has already made an enormously consequential decision: which professionals deserve the opportunity to earn the customer's business.

For routine, well-defined services, an agent may eventually handle the entire process within the customer's instructions and approved limits. For complex services, the agent may primarily determine who reaches the final consideration stage.

Those are different kinds of delegation, and both matter commercially.

In fact, the second may be especially important for professional service businesses. A law firm doesn't need an AI agent to autonomously sign an engagement agreement for AI discovery to influence its customer acquisition. It needs to be among the firms the agent identifies as relevant and qualified enough to recommend for a consultation.

The same applies to a remodeling contractor invited to bid, a specialist dentist selected for an initial examination, or a senior living community included in a family's tour schedule.

And the agent's work doesn't necessarily end when the human meeting begins.

It could help prepare questions, organize notes afterward, compare competing proposals, identify inconsistencies, and remind the customer which important issues remain unresolved. With appropriate permission, it could even help evaluate information generated during the consultation itself.

That creates a much more interesting future than one where AI either does everything or hands the process back to a human.

What this means for service businesses

I think businesses should prepare for both forms of AI involvement.

They need to make the information an agent can reasonably evaluate beforehand accessible and trustworthy: qualifications, specialties, operating experience, geographic coverage, relevant capabilities, policies, and the limits of what they offer.

At TrueSignal, this is one of the problems we're working on with TrustRecord. A business's actual history and capabilities should be easier for AI systems to find and evaluate, with clear distinctions between business-provided statements and information supported by underlying records.

But no published record can fully substitute for a professional's assessment of an individual customer's situation.

That means businesses also need to think about what happens after they're recommended.

Do consultations generate clear proposals and useful explanations? Can customers understand the differences between competing approaches? Does the business document its recommendations in a way the customer — and potentially their AI assistant — can review later?

A contractor's carefully explained proposal may eventually matter as much to an AI-assisted customer as the information that got the contractor shortlisted in the first place.

For complex services, the goal should be to arrive at the first conversation with more of the basic questions already answered, so the professional and customer can concentrate on the questions that genuinely require their interaction.

We shouldn't assume every customer will want to delegate the same amount of authority. And we don't yet have representative evidence establishing which categories of service providers consumers will consistently allow agents to hire without direct involvement.

What seems increasingly likely is that AI will participate in more stages of the decision, even when it doesn't control the final one.

A personal agent may eventually arrange a routine repair without the homeowner giving it much thought. That same agent might spend hours helping the homeowner choose a contractor for a major renovation, then leave the final decision to a conversation across the kitchen table.

Both represent a significant change in how service businesses get chosen.

The future may not be divided between businesses hired by AI and businesses hired by people. It may be defined by how much of the decision AI makes before the customer ever meets the professional.

And for many businesses, getting that meeting will be the first decision that matters.