Your AI Chatbot Is Answering Client Questions: Who Is Accountable When It's Wrong?

A client asks whether a fee has changed. Your chatbot quotes an old price. Someone asks if a filing deadline is next Friday, and the chatbot confirms a date that does not apply. Or it offers a service your firm does not provide.

The immediate problem is an incorrect answer. The larger problem is that no one decided who is responsible for preventing it, correcting it, and addressing the client when it happens.

When a business puts an AI assistant in front of clients, the business is still accountable for the experience it provides. The vendor may supply the software, but the client sees your company’s name, and may act on the answer as if it came from your company.

“The AI said it” does not settle the issue

A client is unlikely to distinguish between a chatbot’s output and an official answer from your business. If the response is wrong, they may rely on it, make a decision, miss a deadline, or expect a service you never agreed to provide.

That can become a customer-service dispute, a financial loss, a damaged relationship, or, in regulated or professional settings, a legal or compliance concern. The precise consequences depend on the facts and applicable rules, but describing an answer as AI-generated does not automatically remove the business’s responsibilities.

The Federal Trade Commission’s 2025 action involving DoNotPay is a reminder that AI-related claims and services can attract scrutiny. The FTC challenged claims that the service was equivalent to a human lawyer without adequate substantiation. The broader lesson for businesses is straightforward: claims made through an AI service still need to be accurate and supportable.

The risk is especially important for accounting, law, insurance, healthcare, and wealth management firms. Clients in these fields may share sensitive information and rely on professional judgment. A chatbot that confidently answers beyond its approved role can undermine trust even if the original goal was simply to handle routine questions.

Internal drafting is not the same as client-facing advice

AI used internally can help an employee draft a response, summarize information, or prepare a first version of a document. If a qualified person checks the work before it leaves the business, that review is a meaningful control.

A client-facing chatbot is different. It may answer immediately, at any hour, without a staff member seeing the question or the response. The human review that would catch an outdated fee, an invented commitment, or a misleading statement may not exist.

That distinction should guide the level of oversight. If AI is only assisting staff, define who reviews its work and what they must verify. If it communicates directly with clients, decide in advance what it may say, when it must stop, and who is responsible for monitoring its performance.

Vendor terms do not transfer your client relationship

AI vendors’ terms vary, so businesses should read the specific agreement rather than assume what it covers. Some terms limit warranties about accuracy, make the customer responsible for how the service is configured or used, or cap the vendor’s potential liability. Data-processing terms may separately address retention, security, subprocessors, and whether customer information can be used to improve models.

Those contractual details matter. But a contract between your business and a vendor does not necessarily resolve a client’s complaint or determine what a regulator expects from your business. Your company chose to deploy the tool and remains responsible for how it fits into the service you provide.

Before adoption, review the vendor’s terms, privacy commitments, security documentation, and data-processing agreement. Pay particular attention to what happens to client information, how long it is retained, whether it may be used to train or improve models, how incidents are reported, and what happens if the model or terms change. The FTC has also emphasized that AI providers need to honor privacy and confidentiality promises made to customers; see its guidance on AI companies and privacy commitments.

A practical accountability framework

You do not need to start with a complicated AI program. Start by making responsibility and boundaries explicit.

1. Name one accountable human owner for each client-facing AI.
Assign a specific person, not just a department, to own the tool’s approved purpose, configuration, review process, and escalation path. That person should have authority to pause the chatbot if it begins producing unsafe or unreliable responses. A backup can manage day-to-day coverage, but accountability should not be left ambiguous.

2. Define what the assistant is allowed to answer.
Keep its scope narrow. For example, it may direct clients to published office hours or explain how to submit a document using approved instructions. Do not let it infer company policies from scattered documents or improvise answers about services, fees, eligibility, or outcomes. Keep its reference information current, and designate someone to update it when the business changes.

3. Set a hard handoff for sensitive or consequential questions.
Route questions about pricing, legal, medical, or tax matters to a qualified person. Do the same for appointment scheduling or changes when a wrong answer could create a missed commitment. Any question involving a client’s personal, financial, health, or confidential information should go to a person through an approved channel, not be handled by an AI assistant that is not authorized and designed for that task.

4. Keep and protect conversation logs.
Logs help the business investigate an error, identify patterns, and improve its controls. Define what is recorded, who can access it, how long it is retained, and how sensitive information is protected. Logging should not become an excuse to collect unnecessary client data.

5. Review transcripts for the first 30 days.
During launch, review conversations regularly, not just complaints. Look for stale answers, unsupported promises, missed handoffs, confusing phrasing, and attempts to draw the chatbot outside its scope. Correct the source material or settings, document changes, and continue reviewing after significant updates to the model, integrations, or knowledge base.

These steps align with the risk-management approach in the NIST AI Risk Management Framework, which emphasizes managing AI risks across the system’s use and lifecycle, not treating launch as the end of oversight.

Questions to ask before switching on any client-facing AI

• Who is the named human owner, and who can pause the system?
• What specific questions may it answer, and what topics are out of scope?
• Does it use current, approved information, and who updates that information?
• Does it hand off pricing, legal, medical, tax, scheduling, and client-data questions to a person?
• What client information is collected, retained, shared, or used to train or improve the service?
• Can we review conversation logs securely, and who is responsible for doing so?
• What does the vendor’s contract say about accuracy, security, data use, incident notice, and liability?
• How will we correct a wrong answer and follow up with an affected client?

If those questions do not have clear answers, the chatbot is not ready to speak for the business.

Keep a human responsible for the promise

AI can help teams draft and assist. It can make approved information easier to find and handle routine requests. But it cannot take ownership of a client relationship, interpret every exception, or repair trust on behalf of your firm. A named human at the business remains accountable for how the tool is used and what clients receive.

For professional-services firms in Kutztown, Allentown, Bethlehem, Easton, Reading, Wyomissing, and Lancaster County, that accountability is part of delivering dependable service, not a technical detail to leave to a software vendor. B&R Computers helps businesses identify risk, understand its potential impact, and put practical safeguards in place. Explore the Resources hub for additional guidance.

If your business is considering a client-facing AI tool, or already has one in use, request a Business Review or Strategy Session to assess the risks, define controls, and establish a clear owner before a preventable mistake reaches a client.