AI agents for business finance: what they can and can't do
What AI agents can do in bookkeeping, close, collections and compliance today, where a person must stay in charge, and how to judge a product.
The short answer
In a growing number of places, yes. The EU AI Act requires AI systems that interact with people to be designed so users are told they are dealing with an AI, from 2 August 2026, and US states including California, Utah and Maine have bot disclosure rules. Elsewhere, consumer and data protection law makes misleading people about who they are talking to risky, so telling customers at the start is the safest default everywhere.
When a customer writes to your business and gets a reply, do they have a right to know whether a person wrote it? The legal answer depends on where you and your customers are. The practical answer is simpler: disclosure costs little and protects you in every market. This guide sets out the rules in general terms. It is not legal advice; check the position for your business with an adviser.
The EU AI Act
The EU AI Act (Regulation (EU) 2024/1689) includes transparency obligations in Article 50. In broad terms, providers of AI systems intended to interact directly with people must design them so the people concerned are informed that they are interacting with an AI system, unless that is obvious from the context. These obligations apply from 2 August 2026. Although the EU has agreed to delay some high-risk obligations, commentators note that most transparency duties kept their August 2026 date.
The Act applies to businesses outside the EU when their AI systems are placed on the EU market or their output is used in the EU. A UK, US or Nigerian business with customers in the EU should therefore plan to comply. Fines for breaching transparency obligations can be significant, so this is worth confirming with an adviser.
United States: state bot laws
There is no general federal disclosure requirement, but several states have rules:
| State | Rule in general terms |
|---|---|
| California | Unlawful to use a bot to communicate with a person in California with intent to mislead them about its artificial identity to incentivise a sale or influence a vote, unless it discloses it is a bot |
| Utah | Under the Artificial Intelligence Policy Act as amended in 2025, disclosure is required in certain situations, including when a consumer clearly asks, and for some high-risk interactions |
| Maine | A 2025 law targets the use of AI chatbots in trade that could mislead a consumer into believing they are dealing with a human |
| Other states | Further bills are under consideration; the picture is changing |
On top of state laws, the FTC has warned businesses not to use AI to deceive people, including about whether they are talking to a real person. Deception about the nature of the interaction, where it affects a customer's decisions, can fall under the FTC Act. Our guide to FTC rules for AI support covers the wider US position.
UK, Canada and Nigeria
The UK, Canada and Nigeria do not, at the time of writing, have a general statute requiring chatbots to identify themselves in customer support. Other rules still point the same way:
- UK. Consumer law prohibits misleading actions and omissions, now enforced directly by the CMA under the 2024 Act. Data protection law requires transparency about how personal data is processed, and the ICO expects organisations to explain their use of AI.
- Canada. The Competition Act prohibits misleading representations. In Quebec, Law 25 requires businesses to inform people when a decision about them is based exclusively on automated processing.
- Nigeria. The Nigeria Data Protection Act 2023 requires transparency about processing and gives data subjects rights relating to automated decisions. The FCCPA gives consumers a right to clear information.
None of these is a bot disclosure law as such. But a customer who later learns that "Sarah from support" was software, and that it made a decision against them, will have stronger grounds to complain.
A disclosure approach that works everywhere
- Say it at the start. A short line such as "You are chatting with our AI support assistant. A member of the team can take over at any time."
- Do not give the agent a human persona. A name is fine if it is clearly an assistant's name; a fake photo and job title are not.
- Answer truthfully if asked. "Am I talking to a real person?" must always get an honest reply.
- Offer a person. Make the route to a human clear, and honour it.
- Mark agent-written emails. If the agent sends emails, say so in the signature.
- Tell people when a decision is automated, and how to ask a person to review it.
Illustrative example. A UK online retailer with customers in France, California and Nigeria uses one disclosure line in every market and translates it into French. Rather than tracking which law applies to which customer, it meets the strictest standard everywhere. The agent's signature on emails reads "Sent by our AI support assistant, reviewed by the team where needed".
Disclosure is not the whole job
Telling customers they are talking to an AI does not make a wrong answer acceptable. The business remains responsible for what the agent says, as the Air Canada chatbot case in Canada showed. Disclosure should come with the controls that make the agent worth trusting: answers drawn from written policy, a clear route to a person, refusals and exceptions decided by people, and a record of every conversation. Customers tend to accept an AI assistant that is open about what it is and quick to hand over; they resent one that pretends.
Where Pryme fits
The Customer Support agent in a Pryme Business account answers from your policies with the source attached, so customers can check an answer rather than trust it, and it hands cases to a named person when needed. What it tells customers about being an AI is part of the setup you choose; ask Pryme about the options. The agents are built and certified on Pryme Intelligence, the governed agentic operating system they run on, and every action is on the audit trail. For what to record, see audit trails for AI agents.
Questions people ask
Does the EU AI Act require chatbots to disclose they are AI?
Broadly, yes. From 2 August 2026, AI systems intended to interact with people must be designed so users are informed, unless it is obvious from context.
Which US states require bot disclosure?
California, Utah and Maine have rules covering certain situations, and others are considering them. Check the current position with an adviser.
Can my AI agent use a human name?
A name is generally acceptable if it is clearly an assistant. Avoid anything that suggests the customer is talking to a real person.
Is there a UK law requiring chatbot disclosure?
Not a specific one at the time of writing, but misleading customers can breach consumer law, and data protection law requires transparency.
Sources
- Regulation (EU) 2024/1689 (EU AI Act), EUR-Lex · accessed Oct 2026
- Jones Walker: Yes, August 2 still matters for EU AI Act transparency duties · accessed Oct 2026
- California Business and Professions Code section 17941 (bot disclosure) · accessed Oct 2026
- Cooley: AI chatbots at the crossroads, navigating new laws and compliance risks (Oct 2025) · accessed Oct 2026
- FTC: Chatbots, deepfakes, and voice clones: AI deception for sale · accessed Oct 2026
This article is general information, not financial, legal or tax advice. Products described are provided under the terms on this site and are subject to regulation and eligibility.