Nearly half of lawyers in a new Australian survey say their clients are arriving with inaccurate information obtained from AI. At the same time, healthcare providers report that AI-informed patients ask more thoughtful questions than they used to.
Both of these things are happening at once. And they change what the first conversation with a customer looks like.
The research: better questions, wrong assumptions
On October 30, the Victorian Legal Services Board and Commissioner published findings from a survey of 2,221 lawyers in Victoria, Australia. Among client-facing lawyers, 46% reported clients arriving with inaccurate or misleading information obtained from AI, and 41% had received AI-generated documents or correspondence that needed correction. Only 10% reported clients arriving with a better understanding of their legal issues because of AI.
Meanwhile, 46% said clients' AI use increased the time and effort required to deliver services, while only 19% said it reduced the work.
The pattern shows up in healthcare too. In a Zocdoc-commissioned U.S. survey published in March, 83% of participating providers said they had corrected AI-sourced misinformation during visits, and 63% said patient AI use added time to appointments. Yet 77% reported that AI-informed patients asked more thoughtful questions, and 76% reported greater patient engagement.
Patients can ask better questions while bringing incorrect assumptions. Lawyers can encounter clients who've done more homework but need more time checking it. Preparation and understanding aren't the same thing.
Three incomplete briefs
Think of it as three people with different information walking into the same conversation.
The customer knows what happened, what they're worried about, and what outcome they want. The AI knows what the customer chose to tell it, plus whatever it could retrieve or generate. The professional knows the field — but often can't reach a sound recommendation until they've examined the particular situation.
Those briefs can disagree. Someone seeking a divorce attorney after asking AI to estimate a settlement may not realize the model lacks crucial facts about finances or jurisdiction. The attorney's first contribution is identifying what changes the advice.
An auto repair customer may arrive asking for a transmission replacement because a chatbot matched a symptom to an expensive failure. A mechanic still needs to inspect and diagnose the vehicle. A family researching senior living may be convinced a particular community is suitable based on published amenities, only to learn during a care assessment that the resident needs a different level of support.
These examples illustrate the difference between information about a provider and information about the customer's problem. A verified operating record can help establish that a contractor has completed similar jobs or that a practice has the relevant credentials. It cannot establish what work should be performed for someone it has never examined.
The embarrassment factor
There's a complication that makes this more interesting than a simple story about correcting chatbot errors. People may be more willing to explain embarrassing details to AI than to another person.
In research by Abigail Sussman, Eric So, and Fiona Yang, participants generally favored human advisers when expertise mattered, but their preference shifted toward AI when disclosing a problem felt embarrassing. In one experiment described by MIT Sloan, participants offered justifications for their behavior more often to human advisers than to AI advisers: 33% versus 15%.
Zocdoc found that 23% of surveyed patients said they had concealed their AI use from a provider, and another 24% said they would consider doing so. A patient may be comfortable discussing a concern with a chatbot but embarrassed to reveal the conversation to a doctor. A legal client might give their attorney a tidier account than the one they gave ChatGPT.
What the AI heard and what the professional hears may not match.
That doesn't mean an agent should automatically forward a private conversation. Sensitive medical, legal, and financial information requires deliberate consent and appropriate safeguards. The opportunity is to help customers communicate what's relevant on their terms — not to turn a private conversation into an automatically distributed dossier.
The first meeting should test the brief, not confirm the booking
AI commerce promises a shorter path to buying. For many services, the more useful outcome is a better-prepared first meeting. The agent can research relevant qualifications, identify likely providers, and help the customer articulate priorities. Then the professional has to determine whether the customer's initial description is accurate and whether the proposed solution makes sense.
That handoff should distinguish five things:
- The customer's goal
- Facts independently established
- Assumptions supplied by AI
- Questions that require professional inspection or judgment
- Decisions the customer has not yet authorized
In home services, this might separate property facts from a proposed equipment replacement. In medicine, symptoms from a chatbot's diagnosis. In law, documented events from a legal theory.
The provider has responsibilities here too. Researching a problem shouldn't make customers feel dismissed. A useful first question may be: "What have you already learned, and which parts are you unsure about?" Asking that without judgment can reveal both the customer's expectations and the assumptions that need checking.
Why measurement has to go past the booking
This is also why measuring AI's influence on service businesses can't stop at mentions or booked appointments. An agent can successfully schedule a consultation without understanding what the customer will ultimately need. Whether that appointment becomes a suitable engagement depends on information neither side possessed when the shortlist was assembled.
Forrester's June 2026 consumer research found that roughly three-quarters of surveyed online adults in the U.S., UK, and Canada were uncomfortable with AI autonomously purchasing and paying, even with preset rules. Gartner's May survey likewise found greater willingness to let AI narrow choices than make purchase decisions. Both studies concern shopping — not hiring attorneys, doctors, or contractors.
To establish what happens in services, we'd need something different: consenting customers followed from their original AI research through consultation and, where appropriate, engagement. We would document which assumptions they brought in, what the provider corrected, and who made the final decision.
What this means for your business
The service economy isn't one big checkout flow. Sometimes the work is routine enough that an agent can arrange it within clear boundaries. Sometimes the consultation is where the job is finally understood.
If AI is going to help consumers choose good professionals, two things need to be true. First, the information AI uses to build its shortlist needs to be accurate and verifiable — not scraped from an outdated website and presented as fact. That's what a TrustRecord exists to provide: structured, verified data about what your business actually does, so AI starts with evidence instead of assumptions.
Second, the first conversation between a customer and a professional needs to be treated as what it is: the moment where the real problem gets defined. AI can prepare that conversation. It shouldn't be expected to replace it.
Sources: Victorian Legal Services Board and Commissioner, Oct 30, 2026; Zocdoc/Censuswide, Mar 24, 2026; Sussman, So & Yang research; MIT Sloan, Aug 31, 2026; Forrester, Jun 15, 2026; Gartner, May 27, 2026.