A buyer asks ChatGPT which decisions your software can make on its own. By the time a sales rep sees the vendor comparison, there may already be a check beside “autonomous.” Where did that check come from? Quite possibly a sentence nobody at your company wrote.

That’s why we’d put owned media in the PR budget. Pay an editor to sit with a customer and follow one decision all the way to sign-off. The next buyer who has to defend that check in a purchasing meeting has something to bring along.

Coinbase CEO Brian Armstrong called his AI agent a “chief of staff.” The job title is easy to remember. The detail worth remembering is in State of AI’s account of Armstrong’s workday. His agents chase reviews and bring him a queue of decisions. Every change still goes through human review.

On November 10, OpenAI introduced Businesses in ChatGPT. Companies can claim a profile and update their information. Eligible businesses can also run a labeled agent. A buyer can read ChatGPT’s description of a product and ask the vendor a follow-up without leaving the app. If those answers disagree, whose account can the buyer take back to the team?

Two Campaign Agents, Two Different Promises

In September, we compared the approval rules behind a wave of AI campaign launches. Demandbase’s Mojo could take a campaign live within spending and targeting limits set by the customer, pausing at what the company called critical decision points. Multiply’s agent reviewed Google Ads accounts for wasted spend and required a person to sign off on any change that affected spending. A buyer asking which tools can manage paid media could reasonably get both names back. The person responsible for Friday’s ad budget would want the difference spelled out.

Here’s the story we’d want a vendor newsroom to publish: take one proposed budget change and follow it through a customer’s account. Show the limit the customer set, the screen the campaign manager saw and the entry in the activity log. Now the buyer has a follow-up for the demo: “What happens to this recommendation if our approver is out for the week?” That’s a question a feature list gives them little reason to ask.

Who Releases the Money?

Suppose an AI answer says an expense tool “handles reimbursements.” Accounts payable still has a Friday payment run to staff. A tool that flags duplicate receipts could save someone an afternoon. One that releases payments would change who has authority over the bank account. The word “handles” covers an uncomfortable amount of ground.

Cheryl Benoit, COO and chief data officer at Iron Gate Program Advisory, walked State of AI through a financial task that crosses a border. An agent can retain access to a system even after the rules governing its authority have changed. “Is the authority where it began, or is the authority where it ended?” she asked.

Give a reporter half an hour with the person who ran payments last Friday. Which transactions went straight through? Who got the one that needed approval? A customer who can walk through that exception has more to tell a prospective buyer than an executive restating the launch announcement.

The Question Customers Keep Asking

Kimberly Eberl, founder of The Motion Agency, told State of Brand about a consumer client that had put PR on the back burner. The prospect of customers asking chatbots for recommendations renewed its interest in press coverage. Meanwhile, the brand had a ready source of questions to investigate in the comments under its own posts.

A trade reporter decides whether any of those questions deserves a story. With an owned publication, PR can assign the one customers keep asking even if no outlet picks it up. A named customer, a link to the relevant documentation and a clear disclosure of who paid for the work give a reader grounds to judge the answer.

Our publisher, Outlever, runs newsrooms for companies as well as publishing State of Brand and State of AI. We’d settle one question before commissioning a company newsroom: can its editor print the detail that might cost the company a sale? For the buyer, that’s often the detail worth forwarding.

Keep the Story Useful After Launch

A pilot customer might have a feature in March that buyers on the standard plan still can’t get in June. Put the plan and the date next to the claim. When access expands, add a dated update at the same address. Someone in procurement may already have pasted that link into a recommendation.

As State of Brand reported on ChatGPT’s Business Hub, PR can count mentions and website clicks. It can’t see the individual conversations behind them. We’d spend the next discretionary PR dollar on the source a buyer can open after one of those conversations. The company can answer for every word on that page.

FAQ

Which product claim should PR assign first? Ask a sales engineer for the last deal where a buyer had to revise the business case after the demo. What had the buyer counted on the software doing? Start with the sentence that set that expectation.

What if the reporting turns up a limit we’d rather leave out? If it would change a buying decision, put it in the story. Give product a chance to check the facts. A customer who discovers the limit during implementation has a fair question for everyone who knew it at launch.

What should a buyer do if an AI answer and a vendor’s story disagree? Give the vendor the exact sentence and ask them to demonstrate it on the plan you’re pricing. Until they can, mark the capability unverified. Two links that trace back to the same press release amount to one company claim.

Publisher disclosure: Outlever builds and operates newsrooms for brands. It publishes State of Brand and State of AI as its own editorial properties.