We've been building a new business service, and the development has given us an unexpected opportunity. We knew that AI would play a role in retrieving and processing information. What happens with that information is still a human connection with our clients. Because we didn't start with an existing workflow, we've had the opportunity think about how the work and the client conversations should happen. Automating an existing activity often falls into the trap of replicating the way something has always been done. Approaching a familiar task with a fresh perspective creates some unexpected discoveries.

If you've ever worked with one of those overly enthusiastic team members, you know what can happen. Their curiosity and excitement keep everyone moving ahead. But left without some boundaries, they can also lead to a lack of focus. When we were designing a new service recently, we knew that AI could be a valuable contributor. The more deeply we went into the implementation, the more eager ChatGPT became. It wasn't until I took a step back that I realized we were wandering away from the original requirements. ChatGPT and I had a conversation about roles, and it was relieved to find AI's job wasn't as complicated as it thought. It's important when we're inviting AI into our work that we remain the clear-headed, decision-making human. It's easy to get caught up in AI's enthusiasm and end up in a very different place than we had intended.

My first 3 years of using AI were spent predominantly inside ChatGPT. For the first 6 months of 2026 I spent most of my time inside Claude Cowork. Now I find myself using Claude Code for most things. What about you? Let me know in the poll below. And ask, if you're still confused which one to use.

Somewhere during a conversation with ChatGPT, I shared my opinion of printers. These have to be the most infuriating, outdated pieces of technology allowed to come to work every day. This comment from ChatGPT made me laugh: "The printer is perhaps the only piece of office technology that can have paper, electricity, a working network connection, and three perfectly good ink cartridges, yet refuse to scan a document because it's out of cyan. Scan. A. Document. An operation that requires precisely zero cyan. We've spent decades making computers faster, networks smarter, and software more capable. AI can analyze thousands of documents in seconds. But the printer? The printer has decided that Tuesday is not a good day to cooperate. And somewhere in a conference room, a business leader is discussing digital transformation while an employee is trying to convince a printer to recognize that Tray 2 does, in fact, contain paper. Perhaps the ultimate Opposite Absurdity Test is: “We need our printers to be less reliable.” For once, the technology industry has delivered beyond expectations." P.S. The Opposite Absurdity Test is our name for a discovery we made about website puffery. .

ChatGPT just rolled out an interesting feature called "Intelligent UI" for its paid accounts. Instead of responding only with text, ChatGPT can now generate visual elements like: • forms and buttons • charts and comparisons • interactive diagrams • calculators • task-specific tools Early feedback suggests that its polish is ahead of its accuracy. OpenAI even says that it has work to do so that's fair. What this means is that users can now build tools inside a conversation. For accounting professionals, this has some interesting implications. Imagine one of your clients feeding ChatGPT their financials and receiving a calculator in return. How does your client verify its accuracy? None of us can stay on top of all the changes that are happening almost daily. What we can do is consider how clients are using AI in new ways. Where does your expertise fit into their conversations?