14 essential AI "I Can ..." skills every undergrad should have (Updated)

Employers are increasingly looking for students who show good judgment with AI, can demonstrate their AI skills in an interview, and are adept at directing AI.

Back in April, I published a list of essential AI skills I thought every undergraduate student should have — “I can …” skills they can pitch to employers in interviews.

Given the speed with which AI is being adopted and used in the workplace, though — not to mention the new capabilities that seem to be coming online by the week — I thought it worth updating the list.

Some of the biggest trends to emerge since the previous list are a shift in what employers look for, and specifically a shift from AI literacy alone toward judgment and critical thinking (they want people who can think about how they use AI, and assess what it does and produces); a move toward employers asking interviewees to demonstrate how they use AI on the spot; and the growing importance of being able to direct AI apps and agents — in essence, to delegate smartly to AI.

The original list reflected these to some extent, but it was beginning to feel outdated. And so here’s the updated list of 14 essential AI “I can …” skills (with more context below):

1. I can choose the right AI tool or platform for a specific task, and explain why.

2. I can tell AI clearly what I want, give it the context it needs, and describe what a good result looks like.

3. I can explore ideas through back-and-forth conversations with AI, and push back when it fixates on my first idea or just tells me what I want to hear.

4. I can research a topic with AI and check its work, including verifying that its sources are real and actually support what it says, and that the facts it presents and its reasoning hold up.

5. I can edit and improve AI-generated work, even when it already looks polished, until it meets my standards (and my employer’s or institution’s) and the needs of the people it’s intended for.

6. I can safely and effectively work with an AI agent on multi-step tasks, responsibly setting what it can access and do on its own, and checking its work.

7. I can turn a repetitive or routine task into a reusable AI workflow or assistant, and improve it over time.

8. I can use AI to analyze data and draw out useful insights while checking its outputs and protecting private and sensitive information.

9. I can use AI to turn complex information into clear summaries, visuals, and slides, and spot where they oversimplify or are misleading.

10. I can disclose how I’ve used AI in line with the relevant policies and the expectations of people who use my work, and take full responsibility for anything I produce with it.

11. I can explain what AI is good at and what it’s bad at (including bias, errors, and risks), and how these shape the way I use it.

12. I can use AI creatively and imaginatively to open up new possibilities and opportunities.

13. I can explain how I balance curiosity, care, clarity, and intentionality in deciding when and how to use AI.

14. I can use AI to learn how to use AI, and keep learning as the tools change.

A bit more context

AI use in the workplace — and what employers expect of the people they hire — is changing fast, so it wouldn’t surprise me if this list needs another update in a few months. But one thing is already clear: expectations are shifting in ways that students, especially those graduating over the next year, need to know about.

It’s also a shift that I worry many students are missing. In July, the National Association of Colleges and Employers (NACE) reported that employers now say more than a third of entry-level jobs require AI skills. Yet nearly a third of the graduating seniors NACE surveyed said AI skills would be of little or no importance to their careers.

Three trends in particular stand out here, and each is reflected in the updated list.

Employers want judgment more than tool knowledge

Being able to use AI tools still matters it seems, but it’s increasingly taken as a given. What employers really want to know is whether someone can think critically about what AI does or gives them when they use it.

Microsoft’s 2026 Work Trend Index, for instance, surveyed 20,000 people who use AI at work. Asked which human skills are becoming more important as AI takes on more work, they put quality control of AI output and critical thinking at the top. And NACE reported in June that employers want new hires who can prompt AI well, check what it produces, and judge when it is and isn’t the right tool.

When Strada surveyed nearly 1,500 executives and senior talent leaders about entry-level hires earlier this year, critical thinking and communication came out as two of the most important skills. In contrast, AI literacy came last. Of course that doesn’t mean AI skills don’t matter, but it does means that employers want them grounded in good judgment.

And there’s good reason for this. Research from Anthropic published in February of this year found that people are less likely to question AI’s output when it produces polished, finished-looking work. This is precisely why so many of the skills on the list are about checking, questioning, and improving what AI produces.

Employers are asking candidates to show their AI skills in interviews

Saying you’re “good with AI” on a resume is, it seems, no longer enough. A growing number of employers want to see you actually use it. McKinsey, for instance, has piloted interviews where candidates use the firm’s own AI tool in live case exercises, and are assessed on their judgment — and how they iterate. And Meta’s technical interviews now look at how candidates manage and verify AI-assisted work.

Some employers are going further, and handing candidates flawed AI output to see who spots the problem. And the software company Zapier, which has published its hiring rubric for AI fluency, now watches candidates work with AI in real time. The reasoning (in part): the company cares more about how candidates think and iterate than about how polished the end result looks. The company also looks at trajectory: where someone started with AI, what they tried and dropped, and how their approach has changed.

This is why the skills above are framed as “I can …” skills. For each one, if you’re a student reading this, it’s worth having a real example you can talk through — or better still, demonstrate — along with a sense of how your approach has evolved.

The bar is moving from using AI to directing it

AI agents, essentially AI that can take actions for you, such as editing files, sending messages, carrying out autonomous research and actions on your behalf, or working through multi-step tasks — are fast becoming part of everyday work. As a result, employers increasingly expect new hires to direct AI, not just chat with it.

Zapier’s updated rubric, for example, expects every new hire to have AI built into their core work through repeatable systems, not one-off prompts. It also treats accountability as a core part of AI fluency. As Zapier puts it: “With AI, you can delegate the work, but not the accountability.”

Microsoft’s research is pointing in the same direction. The most effective AI users, it seems, are clear about the outcome they want and what “good” looks like. And they know when a task calls for asking AI, working alongside it, or handing it off entirely. And in a Cognizant/Pearson study, 96% of HR leaders said they expect entry-level roles to evolve into jobs that involve supervising or managing AI within five years. This aligns very much with my own perspective, which is increasingly leaning toward how managers learn to manage AI as well as people.

And to be clear, these expectations are here now. In just the past few weeks (days in the case of OpenAI), Meta has extended its Muse agent into workplace software, and OpenAI has just launched Dots — agents that users give a goal and set limits on what they can do on their own.

Of course, handing work to AI comes with real risks. And with the speed at which things are moving, these threaten to become serious liabilities to employers and employees alike if people don’t develop then necessary skills fast. This is precisely why the list includes forward-looking skills designed to instill agility and the ability to continuously learn and adapt in students, including being clear about what you want (2), working safely with agents (6), building reusable workflows (7), and taking full responsibility for what you produce (10).

The common thread here is that employers are looking less for people who can use AI, and more for people who can think with it, check it, direct it, and take responsibility for the results.

These are all learnable skills — and hopefully the list above is a good place to start.

AI Use Statement

As with the previous list, this update was researched, brainstormed, and iteratively drafted, with the trusty help of Claude (Opus 5.5 Max). Final decisions, editing (you can probably tell), validation, and sign-off, were all mine.