The workplace impact of AI isn’t just a matter of whether, but when and which jobs get replaced or reinvented by artificial intelligence. AI is already rewriting the rules of how we work together—by reshaping our interactions with other humans.

That’s the flip side of a story I published last week in The Wall Street Journal, A Guide to Collaborating With ChatGPT for Work. The Journal story focuses on the ways we can get better results from AIs by treating them more like human colleagues than like software tools—for example, by approaching them the way you’d approach a junior member of your team, and by providing them with specific, constructive feedback.

In mapping out this advice, however, I found myself reflecting on how this very approach—working with AIs as if they’re quasi-humans—almost inevitably rewrites our relationships with actual humans.


Bad news for the new kids?

Beyond the results we can coax from a chat when we get conversational, the potential impact of AIs on our human relationships gives us another reason to treat AIs with more courtesy than they perhaps require. Get too comfortable barking commands at your AI helpers, and that lack of consideration may seep into your text messages with humans who actually do care if you say please and thank you, or if you ask them about their bandwidth before loading them up with tasks.

Indeed, the impact of AI on collaboration is most likely to be felt in our relationships with the junior colleagues who are most susceptible to displacement or deskilling. After all, the types of tasks we are most likely to assign to AIs are the kinds of tasks many of us typically assign to junior colleagues: things like information gathering, writing first drafts, synthesizing notes or sorting out logistics.

I worry a lot about how the rise of AI will affect opportunities for people who are new to the workforce today, or joining the workforce over the next decade, because how long will managers and organizations be willing to invest in training entry-level employees (who will just move onto bigger and better things) when the alternative is to train a single AI just once?


A new path for entry-level employees

Without entry-level opportunities, however, there will be no way for younger or less experienced employees to acquire the skills and relationships that set them up for gradual advancement into the mid-level and senior roles that can’t (yet) be done by AIs. While some people might see that as a reason to keep giving grunt work to young humans instead of (even younger) AIs, it’s an even better argument for rethinking what entry-level work should look like.

It’s time to stop giving grunt work to junior employees so they can “learn the ropes” and eventually advance to more-demanding work. The grunt-work model of advancement is predicated on the idea that junior team members can learn by simply watching their senior colleagues…which is a great way of absolving mid-level and senior employees from the responsibility to actually teach and mentor. For proof that this model is outdated, consider that we still describe it in terms of people “working their way up from the mailroom”—even though mail has long since been displaced by email!


From grunt work to mentorship

If AIs can run the (e)mailroom and take care of more and more of the grunt work, that’s not a reason to dispense with junior employees; it’s an argument for scaffolding our junior colleagues’ learning and growth with mentorship and training programs that provide for much more rapid advancement. Yes, that’s more work for mid-level employees—but training up new employees (with their AI helpers!) means that mid-level employees will be able to hand over some of their tasks to juniors, and make more room on their plates for mentoring and coaching.

In the best-case scenario, younger employees have earlier and more positive experiences with work, forming strong and better relationships with their colleagues and employers: If grunt work is replaced with challenging, engaging work from the earliest days of employment, there’s less opportunity for juniors to build up resentments based on the way they’re piled up with dull and tedious tasks.

This points to a larger opportunity: In redefining the ways we work with our fellow humans, AIs may actually make us better managers, mentors, and colleagues. That’s especially true if we cultivate a practice of reflecting on what we’re learning from working with AIs and thinking about what this might teach us about working with other humans.


Getting there with AI

Precisely because working with AIs is still such a novel experience, it’s a great opportunity to see your collaboration habits in a new light or to hone new management skills. Here are four ways to reach a new level of collegiality through what you learn from working with ChatGPT or other AIs:

    1. Get confident. Recognize that the fear or discomfort you feel at an AI potentially outshining you—or even taking your job!—may be the same reason you feel hostile towards a brilliant new hire or a superstar colleague. If you feel uncomfortable with other people or AIs shining, you’ll never feel fully comfortable on a team. You need to work on this from the inside out, by getting comfortable and confident in your own contributions.
    2. Get specific.Generative AIs delight us when we tell them to write or draw in the style of a particular genre or example. But this is more than a party trick: It’s a great lesson in how to get the results we want from a co-worker! Try being comparably specific with your human colleagues by showing them examples of work that reflect what you’re trying to get them to take on. Just as an AI succeeds in being “micromanaged” where many humans wouldn’t, some humans also flourish when they get more detailed instructions.
    3. Get skeptical. Working with an AI is a great way to fine-tune our bullshit detectors, so we learn to distinguish between confidence and accuracy. The authority with which AIs rattle off wholly inaccurate bios and answers should help us recognize that how polished or confident someone sounds should not be taken as an indicator of whether they’re right.
    4. Get courageous. When I initially asked ChatGPT to minimally revise this newsletter for me, it went to town and started on a whole new draft. Three paragraphs in, I cut it off: “You are doing much too much editing. Keep this 95% the same, and just tweak a tiny bit.” This experience led me to think: Boy, I wish I could say that to some of the people I have worked with! And then I wondered, well…what would happen if I did? Learning to be more direct and courageous in our communication with human colleagues could be another great spin-off from working with AIs.

And now, a word from the AI itself

I hate writing conclusions. Like, HATE HATE HATE.
So I decided to let ChatGPT do it…and to be honest, the results give me two more “gets”:

  • Get comfortable being transparent about who (or what) is helping you with your work, and
  • Get realistic about when an AI just can’t do the job.

But I think ChatGPT’s draft conclusion makes that second point better than I can:

Conclusion: Embracing Al for a collaborative future

The emergence of Al in the workplace presents us with an opportunity to redefine how we collaborate and interact with each other. Instead of perceiving Al as a threat to human jobs or a catalyst for dehumanizing interactions, we should leverage Al’s potential to enhance our human relationships, promote better mentorship, and create more meaningful work experiences…

In embracing Al as a collaborative partner, we can foster a workplace where both humans and Al thrive together, ultimately leading to a more efficient, dynamic, and positive work environment.

This post was originally featured in the Thrive at Work newsletter. Subscribe here to be the first to receive updates and insights on the new workplace.