At Anthropic, Kristen Swanson’s role is to help people get better at using AI.
Often, that means knowing when they should do the work themselves.
“I have a job teaching people how to use AI, and I probably spend at least half of my time telling people when not to use AI,” said Swanson, who leads AI fluency research and learning at the company.
The savviest AI users, she said, aren’t necessarily those who rush to it for every task. What sets them apart is their judgment about what to delegate.
That’s becoming an increasingly important workplace skill as AI finds its way into more of what people do. The challenge isn’t only learning how to use the technology. It’s knowing when it can save time, when it might create more work, and when a task is better left to a person.
Knowing what not to delegate
In some cases, Swanson said, evaluating what AI produces takes more effort than doing the task yourself. It’s what she calls a “discernment tax.”
When people hand off the right things to AI, Swanson said, the time spent reviewing its output can be worthwhile. Yet for tasks people are already adept at doing, evaluating AI’s work can eat into whatever time they hoped to save.
Properly judging what AI produces, Swanson said, “is the highest-order thinking that you can do. It’s actually exhausting.”
That’s why asking AI for help with tedious data analysis might make sense, she said. Yet having it write a social media post about something you’re knowledgeable about could leave you doing more work, “parsing through all of this stuff.”
The discernment tax isn’t the only reason to keep some work away from AI. The technology’s limitations can also make delegation risky.
“If you’re asking AI about a really niche research paper or researcher, and it didn’t have a lot of that information in its training, it might hallucinate,” said Swanson.
Bosses are also trying to draw clearer boundaries around when employees should turn to AI. While many companies are pushing workers to embrace the technology, some CEOs have warned against letting it substitute for employees’ own thinking.
Scott Stevenson, head of Spellbook, which develops an AI tool for drafting and analyzing contracts, recently told employees he wanted the “D+ version” of their ideas rather than proposals polished by AI because he wanted to see the thinking behind them and avoid bloated memos drafted using AI.
Avoiding ‘capability overhang’
Another challenge is that the line between what people should delegate and what they should do themselves can shift as AI improves, said Swanson, who oversees Claude Academy, which offers educational resources around AI. A task that makes sense for people to do themselves now might be a better fit for AI in six months.
That’s why more sophisticated AI users routinely revisit what the technology can do, Swanson said.
Doing so is a way to avoid what’s sometimes called “capability overhang,” where users’ understanding of what AI can do gets frozen in time based on earlier experiences with it. That can lead people to stick to familiar uses, even as AI models themselves become more capable.
That’s one reason, she said, Claude Academy asks users to write down some of their hardest tasks. As new models arrive, they can try those tasks again and see how well AI does.
The latest models might perform better on some tasks, though not others, Swanson said. Retesting a task can give users “a better sense of what’s possible.”
That means continuing to experiment, she said.
“Being able to say, ‘I tried this, and it didn’t work, and I’m going to try it a different way next time,’ is much more important than, ‘Do I have this feature turned on? Did I prompt this way?'” Swanson said.
Ultimately, she said, becoming fluent with AI means thinking enough about when and how to use it that those decisions become second nature, rather than simply trying to use as many prompts, features, or AI tools as possible.
“More AI is not always better,” Swanson said. “And more AI is not necessarily fluent AI.”
Read the full article here
