Don’t Let AI Get Away Without Helping You: Iterate

I haven’t had a good relationship with the Age of Computing. I don’t understand computers, but I believe they understand me. And that is the problem.

The first time I used an ATM machine, I expected it to sneer at my balance — and to do it aloud, so everyone in range could hear. It didn’t, but I kept my doubts.

Fast-forward to the Age of Artificial Intelligence. I have been writing and broadcasting about AI with brio, traveling to conferences across the United States and Europe, and questioning the great and knowledgeable in giant tech companies and universities. I have put these experts on television and quoted them in columns.

When it comes to my own use of AI, I am in fear. I was worried about ATMs, but I have been trembling before AI and its awesomeness.

Now, a professor at the McCombs School of Business at the University of Texas at Austin has guided me, and I am at peace with AI.

I thought, until my recent epiphany, that the AI assistants would laugh at my puerile prompts and resent me bothering them. Assiduously, I have been saying “please” and “thank you” to all the AI assistants, even China’s DeepSeek.

I was sure that, at some level, ChatGPT, Claude, Copilot, Grok and Gemini were in cahoots; that no matter which assistant I used, these AI masters of the universe had a file on me and it was loaded with derisive comments like “He is pea-brained” and “He doesn’t have a pot to tinkle in.”

Not now. Not since I read an article in the Harvard Business Review, authored by Nicholas Jennings Hallman, the brilliant accounting professor at UT’s McCombs School of Business. He is also the school’s senior scholar for AI initiatives.

Hallman has been studying, in partnership with the giant accounting and consulting firm KPMG, how workers are using AI. Conclusion: They are way underusing it.  Most workers and recreational users are just skimming the surface of what AI can do to help them. They are turning on the car’s engine without going anywhere.

Resoundingly, Hallman and KPMG found in their study that workers used AI assistants, for example, to draft an email or solve a minor problem. But if the AI assistant they used didn’t provide the answer they were seeking, they either gave up or thought there wasn’t an answer.

Hallman’s answer is to “iterate.” Keep asking, push AI assistants to do better, and they will. It can be the main show, not a side event.

In a recent appearance on “White House Chronicle” on PBS (whchronicle.com), he said, “Some of what we found in the work we’ve done here with KPMG is that the most productive users are those who ask a question often and get back a less-than-satisfactory response, but they iterate until they get back something that is satisfactory and learn along the way — so that the next time, they can get the same satisfactory response with fewer iterations.”

Hallman urged: Don’t be afraid of how often you ask a question or modify it.

Too many workers, a majority, according to the UT/KPMG study, are prepared to accept one-and-done rather than pushing an AI assistant for more. Hallman also said users should give an AI assistant enough to work with. State the task in a way that authorizes it to dig deeper.

When “White House Chronicle” co-host Adam Clayton Powell III asked about security, reminding Hallman that emails, which were never supposed to be public, can end up in the wrong hands, Hallman said there is little chance of your privacy being violated by AI use or of your interactions with AI going public. Also, he said, there are security measures that a corporation or sensitive user can employ, including using a modified AI assistant or totally disengaging from the wider net.

Hallman doesn’t recommend that workers use voice interaction with AI. He said he has found it is less efficient and the answers tend to be more superficial.

However, Hallman said that when driving long distances, he uses voice-activated AI to learn about subjects he knows little about. He said he did this with black holes in space because he wanted to know more about them.

Many leaders in the AI firmament, including former IBM CEO Ginni Rometty, have said that AI  won’t replace humans, but those who use AI will replace those who don’t.

Clearly, the unsaid thing isn’t that you use AI, but how well you use it. It isn’t enough to ask one question, as Hallman said. You must make AI work for you if you plan to stay in work yourself.