Norming with AI
Last week I listened to Brené Brown & Adam Grant rumble with their respective dilemmas about using AI at work and the rise of “workslop”(1) particularly in writing but also in work-product more broadly. I was not expecting to enjoy the podcast as much as I did - and the articles and research they cite were fascinating and worth digesting in their own right (see below for citations and links). Side bonus - not only does the discussion explore this rich territory, but it also is an excellent demonstration of coaching and curiosity in real-time. So good.
Where they arrive through the discussion is actually familiar territory, which is the use of this technology needs norms, boundaries, team agreements, curiosity through feedback, and accountability. What we are seeing more of, unfortunately, is mandates for use without thoughtfully crafted agreements (let alone any training - it’s more of a guess-and-check approach in most companies).
At the end of the day, what is needed around AI is no different from anywhere else in the practice of great, courageous, human leadership: clear expectations, boundaries and commitments, collaborative construction, role-modeling, tranparency, admitting what we don’t know. These together not only build responsible usage, but also cultivate team trust and engagement because people know where they stand. It also reassures teams that using AI is not a stand-alone measure, but another manifestation of what behavior we expect in our work together, what defines great performance, and when that behavior shifts we expect growth, intention, and evolution (and not an easy shortcut).
It made me think about what my own norms are in my business, how I would share them, and what advice I would give to leaders of teams as they chart their own agreements. So here is what I’m thinking about.
For me in solo-preneur world:
My voice is mine - AI is not, can not, nor do I want it to try to become my voice for me. I am not using AI for editing posts or content. I am not using AI to write emails or follow up’s. My business is building on genuine connection - so my presence, written and otherwise, is built around my genuine self.
Trust but verify. So far I am a very skeptical user. I have used AI prompts to support me with research and I always ask it to include source links. I’m finding a stunningly high error rate: so far the majority of links to studies, research, and / or articles have turned up broken or invalid. Even when I have asked it to try again, and sometimes when I try to find said study with a basic Google search - no luck. If I can’t validate the research with my own eyes, I’m not pulling it forward.
Note - Brené has spoken about the validation process she used for her latest book, in which they ran AI and in-person research in parallel as a test. The result was not good - AI was found to be inventing citations over 50% of the time.
Practical and cautious. I’m a practical user too. Need to change keywords in a resume? Great. No one needs to spend precious brain energy on that kind of tedious work. But cautious and critical to ensure that when the keyword is changed that the impact and intent of my words remains. This is the thinking part of writing and representing oneself.
My advice to leaders of teams:
Define what is and what is not OK - on this team, in the context of AI. There are great exercises for this, like a simple above & below the line activity. Maybe you already have these definitions in your team - if so, then have a specific discussion on how AI use fits within what exists or what needs to adjust.
Define accountability when we go out of bounds - similar to the above, a solid norm on this is what do we do when we need to give or receive feedback? Feedback expectations should not necessarily be different on the AI use case than other categories.
Agree on disclosure - in some settings it might be appropriate (or necessary) to disclose the use of AI openly, not for judgement but for transparency. In the podcast, Adam struggles with this, finding non-disclosure inherently dishonest. Instead of staying in judgement, he is coached to perhaps revisit expectation setting.
For me, what this looks like right now is a disclosure footnote across my website; should my intention and use change over time I will change my disclosure.
A note on “voice” above - I recognize not everyone is a confident writer and the allure of polished writing from AI editing is a strong pull. There is definitely a use case to support writing confidence, however, perhaps this is where disclosure matters, in part so you can receive helpful feedback to continue to hone a writing skill.
For teams that already have strong norms and containers in place, this will be light work to infuse AI context into existing parameters.
For leaders who have not yet practiced this in their approach, all the more reason to start.
If you are an individual team member grappling with this for yourself - you do not need a leader to set the tone or intent. Decide for yourself what your intention is with AI - then be honest about that with others. Be the role model you need, and share generously.
—K
Resources & Citations (not AI generated):
The Curiosity Shop, with Brene Brown and Adam Grant - AI, Commencement Speeches, and Why Human Thinking Still Matters. YouTube version, or Apple & Spotify podcasts
HBR - “AI Generated ‘Workslop’ Is Destroying Productivity”- Kate Niederhoffer, Gabriella Rosen Kellerman, et al.
The Atlantic - “The Biggest Tell That Something Was Written by AI” - Eve Fairbanks. *Reading her story on challenging the raccoon in a trenchcoat metaphor served up by AI is worth the read alone. The more she probes the tool, the more bizarre it becomes, as it “{seeks} to justify itself in increasingly bewildering ways”. I love a good word salad - don’t you?
For my recruiting friends out there - there is a thorough study from Stanford examining AI hiring tools impact in hiring systems that should raise alarm bells for all of us. Not mentioned in this study is also a finding that AI prefers itself at a high rate: meaning AI generated resumes are predictably preferred by AI screening systems. I fear this will drive an unintended consequence in the profession: by solving the problem of filter-efficiency for recruiters, we are potentially incentivizing use of AI generated resumes, which if they are not written responsibly give rise to false or mis-leading candidacies, which then undermine the point of efficiency and impact quality. Lose-lose. Keep your eyes on this one friends.
