Your Website Chatbot Speaks for Your Company. Is It Reading the Right Files?

The helper on your website now talks back

For years the chat box on a company website was a form that sent a message to someone's inbox. Today it can answer on its own. Tools from Azure AI, AWS AI and Google AI make it simple to point a chatbot at your website and files and let it reply to customers right away.

That is useful. It also means a piece of software is now explaining your prices, your return rules and your service terms to customers, around the clock, in your company's name. What it says depends on what it reads.

What the evidence says

In February 2024 the Civil Resolution Tribunal of British Columbia decided Moffatt v. Air Canada. After a grandparent died, a customer named Jake Moffatt asked the airline's website chatbot about bereavement fares. The bot said the reduced fare could be claimed after booking, within 90 days of the ticket being issued. Air Canada's own bereavement page said the policy did not apply once travel was complete. When Moffatt applied, the airline refused.

Air Canada argued that the chatbot was a separate legal entity responsible for its own actions. The tribunal member called that a remarkable submission. The decision says the chatbot is still just a part of Air Canada's website and that the airline is responsible for all the information on its website, whether it comes from a static page or a chatbot. It added that a customer should not have to double check one part of a website against another. Air Canada was ordered to pay $812.02 in damages, interest and fees.

In March 2024 The Markup reported on the MyCity chatbot launched by New York City to help business owners. The city said it would draw on more than 2,000 NYC Business web pages, and it ran on Azure AI services. The Markup found it told users that bosses could take workers' tips, that landlords did not have to accept housing vouchers and that restaurants could go cash free. All three answers were wrong under city law. When ten Markup staffers asked the same housing question, all ten got the wrong answer. The bot carried a small note that it may occasionally produce incorrect, harmful or biased content.

The National Institute of Standards and Technology names this risk in its guidance on generative AI, NIST AI 600 1. It calls it confabulation, which it defines as the production of confidently stated but erroneous or false content.

Why chatbots give customers the wrong answer

Old and new versions sit side by side

Most companies have three versions of the return policy floating around. The current page, last year's PDF and a sales sheet from a promotion that ended. A bot that reads all three may pick the wrong one.

Internal files get mixed in

When a bot is pointed at a whole shared drive, it can read draft pricing, staff notes and exception rules meant for managers. Customers then hear terms that were never offered to them.

Nobody owns the source material

Policies change when a manager sends an email. The website gets updated weeks later, if at all. The chatbot keeps reading whatever is there.

How to get your documents ready before the bot goes live

1. 1Decide exactly which pages and files the chatbot may read and keep internal drives out of reach.
2. 2Gather every version of your prices, return rules, warranties and service terms and keep only the current one.
3. 3Give each customer facing policy a named owner and a review date.
4. 4Make sure the website and the bot read from the same source so a change in one shows up in both.
5. 5Write down the questions customers ask most and test the bot against them before launch and after every policy change.
6. 6Set clear topics the bot should hand to a person, such as refunds, legal terms and anything about money owed.
7. 7Keep a log of bot conversations and review a sample every month for wrong answers.

This work pays off even without a chatbot. Clean, current documents help your staff give the same answer every time, and they make your website easier to trust.

What this means if you run a company in Oklahoma

Oklahoma companies of every size are being offered chatbots by their website host, their phone system or their software vendor. Many can be switched on in an afternoon. The documents behind them usually cannot be cleaned up that fast.

A bank, a clinic, a utility contractor or a retailer that launches a bot on messy files is taking on risk it may not see until a customer screenshots a wrong answer. Doing the document work first is cheaper than explaining a mistake later.

At Oklahoma Data Partners, we help Oklahoma companies sort out which documents are current, give each one an owner and set up a single trusted source that a website and a chatbot can both read. If you are thinking about putting a chatbot in front of customers, book a consultation.