You're reading #DRANBLEIBEN, views & analyses on tech and society, by André Cramer. I'm a consultant, speaker and podcast host of DRANBLEIBEN and Code & Konsequenz. Learn more about me here on my website, on LinkedIn, Bluesky or Mastodon!
Welcome to a new edition of #DRANBLEIBEN!
This time it’s about something that’s been gnawing at me for a while. I remember a discussion from my time at Deutsche Telekom, back in 2024. A female colleague had posted an article on the social intranet asking why all the internal AI events were so male-dominated. That made waves. Lively, controversial, with no end of reactions. Plenty of reactions from men, too, and they showed me that nobody here had broken through the surface. The feedback was bursting with lines like “well then, dear women, do finally come to these events. Nobody’s stopping you.”
That very “just come along” showed me back then where the problem was. It treats a cultural question like a pure matter of attendance. As if women only had to enter the room and everything would be fine. The actual question, why this room feels so male to so many in the first place, stayed untouched.
Today, with more distance and a presumed maturing of the AI topic, this very pattern returns, only bigger. You probably know the usual story. Women use less AI, so they supposedly need to catch up on AI. We now have a term for it, too, the Gender AI Gap. But it’s a different gap that occupies me. It’s the male-coded culture of hype and status around AI, and the question of whether organizations are rewarding the wrong kind of AI use right now.
I really want to get to the bottom of this question. I’ve sorted out my thoughts on it, woven in my own impressions and experiences, and researched fairly extensively. I’d be glad to take you along on this journey through studies, my own observations and a few uncomfortable questions. It’ll run a little longer. So best grab yourself a coffee ☕, and let’s get going.
My next encounter with the Gender AI Gap came last year, through a study. It was “Global Evidence on Gender Gaps and Generative AI“, a paper by Nicholas Otis, Solène Delecourt, Katelynn Cranney and Rembrand Koning at Harvard Business School. It bundles 18 studies with more than 140,000 people worldwide, and its finding is uncomfortably clear. Women use generative AI less often than men almost everywhere, across regions, industries and professions. One spot in particular struck me. The gap doesn’t disappear once women and men have the same access to the tools. It stays. So a lack of access doesn’t explain it. The authors expressly note that the causes are not yet settled.
I then used the topic and these very insights from the study in several leadership trainings. And I got very positive responses. Several times women came up to me afterward and thanked me for even bringing up gender and AI at all. That was unusual, they said, they hadn’t expected it and hadn’t experienced it that way before. A thank-you for something that should be a given, that perhaps already says something quite important about the situation.
The topic then came back to me two weeks ago via LinkedIn. A post by Colette Rückert-Hennen, CHRO at EnBW and an HR influencer or “Top Voice”, “The future of AI needs more women“, popped up in my feed through a comment by Robert Franken, a transformation consultant and gender-equality activist I hold in high regard. Rückert-Hennen’s post does something understandable and well meant. It tells the gap as a matter of courage and skill. Women use AI less often, she argues, and thereby risk falling out of the loop and out of sight, so they should become braver, experiment, help shape things.
But that framing is exactly what bothers me. And honestly, it makes me a little angry, too. My irritation is about the reflex behind it, the one that passes for thought leadership just because a “Top Voice” voices it. The reflex simply dumps the whole onus onto women. It reads holding back as falling behind.
Do we really want to meet the enormous cultural challenges AI puts on us with the same old reflexes, pulled straight out of mothballs? Be brave, try harder, make yourself visible! Or can we finally admit that what’s needed here is honest culture work? I want to get at the causes. The endless tinkering with symptoms gets us nowhere. What affects me most here is that such advice to women so often comes from women themselves, and indeed from powerful women inside the system. My appeal is and stays the same. Introducing AI has to be treated as a holistic and comprehensive culture project. If we do that, we’ll get to different conclusions on the Gender AI Gap as well.
If we don’t do that, we leave out the question that seems superior to me. Catch up to what, exactly? And toward what?
The only one who nudged the discussion in exactly this direction was Robert Franken. His point ran roughly like this. Participation in what, exactly? AI systems, he said, are not neutral, and skepticism can be a legitimate judgment. Women therefore belong in AI development, in governance and design. Participation in shaping it, not just in using it.
That hits a nerve. But from there I’d like to go a step further and ask, precisely, about the culture. Which form of joining in are we celebrating very loudly right now, and might it be pushing many women away?
I tried this in a comment and threw in a different view … without resonance. Not too surprising. A thought like this is quite hard to like without positioning yourself against the implicit culture of hype and bosses. It’s socially risky, because it names power. And naming power may get read on LinkedIn, but it rarely gets public support. Perhaps the algorithm doesn’t much care for power-critical notes either.
Let’s stay with this irritation for a moment, because it’s the real starting point of this Deep-Dive. From my consulting work and from my own circle of friends and acquaintances, I regularly hear about women whom the catch-up story doesn’t fit at all. They’ve been using AI for long. Confidently, naturally, productively. And they’re annoyed all the same. The technology bothers them least of all. What grates is the noise around it. The constant tool bragging, the efficiency fantasy, the loud “AI-first” talk, the equation of lots of output with good work.
So the Gender AI Gap may show more than the fact that women use AI less often. It may also show that many organizations reward a form of AI use that repels quite a few women, for good reasons. Loud, status-driven, efficiency-fixated, blind to quality.
An honest aside belongs here, though. The usage gap is real and well documented. On average, women use AI less often, and no anecdote of mine wipes that away. At the same time I know enough women who use AI very intensively and are simply fed up with all this AI bragging.
And I have a hunch. I can’t fully prove it, but the studies I’ve cited point this way. It strikes me that this culture is itself one of the reasons for the lower usage. Anyone who finds the noise, the bragging and the efficiency fixation off-putting might keep their distance from the start. A kind of “leave me alone with your AI and have at it yourselves”. Friction turns into withdrawal, and the withdrawal then shows up in the statistics as a gap.
Why the gap stays even with equal access, the Harvard Business School synthesis left open, as I said. This very why is the exciting question here.
First the numbers, because the basic observation is solid. The Lean In survey from March 2026, among roughly 1,000 US adults, finds 78 percent usage among men and 73 percent among women, and for daily use 33 versus 27 percent. That’s no abyss, but a stable pattern.
The strongest answer to this so far comes from a British study from early 2026, “Women Worry, Men Adopt“, by Fabian Stephany and Jedrzej Duszynski, among others at the Oxford Internet Institute. They turn the usual reading around. On the basis of representative UK data from 2023 and 2024, around 8,000 people, they reach a clear finding. Women use GenAI less often on average because they regard societal risks differently.
In the study, Stephany and Duszynski built an index out of worries about mental health, privacy, climate effects and labor-market consequences. This index explains between 9 and 18 percent of the usage differences and is among the strongest predictors of all for women. For young women it weighs more heavily than digital education or formal qualification. The researchers speak of an “other-oriented” caution, an attitude geared more toward the consequences for others. And they hold that this caution can be a legitimate judgment.
This is the first crucial point. Skepticism is no deficit. It is rather a form of early problem detection. Anyone who takes data labor, deepfakes, algorithmic discrimination, energy hunger and creative devaluation seriously earlier on is not being fearful. These people simply see more than others.
Catherine Connelly, a professor at the DeGroote School of Business in Hamilton, sharpened this in an interview. Women, she says, more often ask questions about quality, ethics and consequences, while some colleagues looked mainly at the gains in speed. In environments where AI gets framed as unquestionable progress, exactly these people fall under suspicion. They are quickly seen as the ones putting on the brakes, as “anti-tech”.
And with that we’re in the middle of the real question.
AI is of course not inherently male. That would be a flat claim, and it would be wrong. But in many organizational contexts AI gets coded in a male fashion. It works wonderfully as a trophy, because it promises everything at once. An edge, control, speed, scale, automation, ownership of the future, you name it.
Whoever knows the newest tools can perform modernity. Whoever talks constantly about agents, automation and prompting can mark competence. Whoever demands “AI-first” can simulate leadership. Whoever produces more output can claim productivity. And whoever dismisses skeptical objections as a brake can crown himself a brave man of the future.
We still lack words for this behavior. Let me try something like AI status performance. Or, put more bluntly, AI bragging. I don’t mean legitimate curiosity about new tools. I rather mean the social use of AI as a marker of superiority. Tool name-dropping, early-adopter posturing, aggressive “AI-first”, talking about automation as a leadership signal, mistaking volume for skill.
There’s hardly any direct research on this exact phenomenon yet. But the theoretical foundation is there and I find it sturdy. The sociologist Judy Wajcman has shown for decades that technology is no neutral thing that happens to drop into our social world. It is gender-coded in design and use, and it helps shape relationships and orders of power. In “Feminism Confronts AI“ Wajcman recently applied this directly to AI.
Fitting with this is a small interview study from early 2026, “Living in the tensions“, which examines gender as performance in STEM. The women interviewed report three recurring patterns. The stereotype that women are less analytical. The pressure to conform to male-coded norms. And the compulsion to constantly prove belonging visibly. Belonging has to be performed there, over and over again.
And the broader critique of tech-bro culture supplies the vocabulary for an environment in which aggression, a fetish for abstraction and contempt for caring orientations get read as strength. One study makes the connection especially tangible. In “The Gender Gap in Scholarly Self-Promotion on Social Media“, Hao Peng, Misha Teplitskiy, Daniel M. Romero and Emőke-Ágnes Horvát examined scholarly self-promotion on Twitter. Over six years, 23 million tweets on 2.8 million papers by 3.5 million researchers. Their finding: Women promote their own work around 28 percent less often than men. The gap grows with performance and status and is largest among productive women at top institutions. And even when women become more assertive, visibility for it is lower that it is for men.
The authors draw a clear consequence from this. Fix the institutions, not the women! That’s the crux of it. Academic platforms reward a male-coded style of self-presentation. Whoever doesn’t serve it disappears, regardless of the quality of the work.
Applied to AI, it’s just the same. It’s not only about who uses AI. It is about who can visibly stage AI use as competence, future-readiness and leadership, and who gets recognized for it.
Here the Lean In survey gets interesting again, read against its own punchline. Lean In, after all, finds more than the usage gap. Among employees who use AI, 27 percent of men report praise for that use, but only 18 percent of women. 37 percent of men feel encouraged by superiors to use AI, for women it’s 30 percent. Women are almost twice as likely to believe AI will cost women jobs, and they worry far more often about being seen as cheaters when they use AI.
Sheryl Sandberg, founder of Lean In and former COO of Meta, reads this as a familiar bias. Men, she says, tend to get praised for effort, women tend to get criticized, and superiors often don’t even notice that they encourage men more. Her conclusion, though, stays the old reflex, the same one Colette Rückert-Hennen calls out. Women should make themselves more visible.
But the data tells a different morale. If the same practice produces status for men and barely any for women, then the problem surely isn’t women’s restraint. Then it’s the reward structure. The Gender AI Gap is, after all, also a gap in recognition and visibility.
And with that, the whole story of women’s deficient self-assessment tips over for me. Yes, many women underestimate themselves, that’s well documented. A study from the University of Pittsburgh shows it for female engineering students. Their self-efficacy runs lower even though their grades are as good or better than the men’s. But the other side usually goes unnamed. Where women underestimate themselves, others overestimate themselves. What’s decisive here is perception. The same attitude gets read differently depending on who shows it. Women who check carefully count as hesitant. Men who assert quickly count as visionary. Women who demand quality count as difficult. Men who produce volume count as productive. Women who name risks count as skeptical. Men who ignore risks count as bold.
We flag low confidence as a problem. Overconfidence, though, we celebrate as leadership. That is the real asymmetry. And I’ll be honest, I find it very hard to bear.
That leaves the point where my observations and the reports passed on to me, above all from creative, communications and consulting contexts, sit closest to the research. With a sentence I keep hearing: “My people, supervisors, bosses keep telling me, just do that (even) more with AI.”
It sounds pragmatic, it often carries far more. That speed beats quality. That creative work is treated as freely shortenable. That process, experience and a sense of style count for little. That leadership sees efficiency and overlooks the actual work that makes good results possible in the first place. That AI use itself becomes the proof of performance, whether or not it helps the concrete matter of work.
There’s now a term for the consequences: Workslop. It was coined by Kate Niederhoffer, Jeffrey Hancock and colleagues at BetterUp Labs and the Stanford Social Media Lab, in a piece for the Harvard Business Review in September 2025. Workslop means AI-generated work that looks professional but lacks the substance. Smooth, plausible, quickly produced, and not really useful for the task. You’ll surely know this. I hardly know a single person in knowledge work who hasn’t already run into it.
In their survey of 1,150 US employees, 40 percent said they had received such outputs in the previous month. On average each case cost just under two hours of rework. More than half reacted annoyed, a fifth even offended. So the decisive mechanism is simple. Workslop saves no work, it shifts it. Someone else has to check, interpret, correct, rethink, repair. Never before in history have we been able to produce superficially professional-looking work this fast, faster than we need to check it. My impression is that an invisible rework economy is taking shape right here.
One more qualification belongs here, otherwise I’d be less than honest. The workslop figures come from an industry survey for an HBR format, not from a peer-reviewed study. The term is brilliant marketing and a plausible observation at the same time. As hard proof it won’t do in this form, but as vocabulary for a real phenomenon it will.
A prominent voice has also just given this invisible work a name from above. Aaron Levie, CEO of the cloud provider Box, recently spoke of an “AI psychosis” in the executive suites. Medicine knows the term as something clinical. Levie uses it as a metaphor. In his experience, he says, leaders systematically overestimate AI because they sit too far from the actual work. They see the happy path, the finished generated contract, the slick prototype. But they don’t see the ten, twenty steps that come after. Checking the code before it goes live, reconciling the contract clauses, repairing things… invisible to the executive levels.
That is exactly the rework we’re talking about here, only from a bird’s-eye view. Whoever shouts “just do more with AI” often sits exactly where Levie locates the AI psychosis. High enough up to see the polished result. Far enough away to overlook the correction work.
What I find piquant is Levie’s therapy. He recommends that leaders use AI far more intensively themselves. D’oh! More usage as the cure… right. Hardly surprising from a tech CEO, of course. And yet it’s striking how the standard answer is almost always the same, no matter the question. Women use too little? Use more! Executives overestimate the tech? Use more! Maybe “use more” just isn’t the answer to every question. (Hint, it absolutely isn’t.)
There’s a question that stays astonishingly unanswered in the research: Who actually cleans all this up? Whether checking, correcting and context work get distributed unequally by gender in the AI age has barely been measured so far. But one suspicion is working inside me. If loud AI use brings visibility and quiet rework stays invisible, then a new division of labor emerges.
Some perform innovation while others secure quality. Some generate volume and others clear away the workslop. Some talk about tools. Others turn unfinished material into usable work. Some get visibility. The others do the judgment work.
Please note: I’m not talking about all women and all men. Of course not. There’s the man who quietly secures quality, and the woman who loudly plays the tool game (I know those too). Individuals say little.
What this is about are behavior patterns, gender-coded roles. It has nothing to do with biology, but a great deal to do with social experience and reward. And we’ve long known these patterns from labor research. Self-promotion and recognition, care work, error avoidance, invisible organizational work. Who gets to be fast and unfinished, and who has to deliver carefully and in a way that connects?
So the suspicion doesn’t amount to a simple “men make the mess, women clear it away”. That would be too simple. The thought is subtler. A culture that rewards loud use and leaves quiet rework invisible pulls exactly these familiar patterns back in. Regardless of who does what in any given case.
I want to bundle the whole thought into one term, and I consider it the real payoff of this research. I call it the judgment gap.
It shifts the focus away from the usage rate. Most studies measure who uses AI how often. But I haven’t yet seen where anyone measures who uses AI well, who spots bad outputs. Who checks context. Who judges when AI raises quality and when it merely produces volume. Who decides when the machine helps and when you’re better off contradicting it.
That is precisely the skill that counts in the AI age, as I see it (and I talk, write, discuss and argue wherever I can to build awareness of it). But it is also the skill that, in my experience, still gets lost in the status game in far too many organizations. What gets sold as a future skill is often tool fetishism with a management hookup. The real future skill, for me, lies in discernment.
The judgment gap is therefore not a technical question. It is a cultural and organizational-sociological one. It asks which kind of appropriation an organization rewards. Is it loud tool know-how, buzzword fluency and quantity? Or rather careful, critical, responsible application?
I want to stay honest about where the thesis stands after all this. The building blocks are well documented, like the adoption gap or the role of risk perception. The self-promotion and recognition gap. The workslop experience and the quality problem. Theoretically, the reading of AI as male-coded status performance connects strongly.
What’s still missing is the direct measurement of the core. So far there seems to be no clean research that asks directly: Do women use AI productively while rejecting the male-coded culture of hype and status around it? I very much hope this gap gets closed soon.
But one observation fits like a key into a lock. The authors of the self-promotion study around Hao Peng expressly recommend fixing the institutions instead of the women. Lean In recommends that women make themselves more visible. Two answers on the same issue. One addresses the structure. The other the individual, the single woman. For me it’s settled, I find the first more convincing. I don’t need a single second to think about it.
#DRANBLEIBEN s free for you, but for me there's a whole lot of work and heart in it. If you feel like supporting me a little in that, I'm grateful for any support. You can give it, for example, through a voluntary paid subscription or a small donation for a coffee.
kostenpflichtiges Abo abschließen
The Gender AI Gap is real, and it is more than a usage or a competence problem. Very likely it is also a reaction, namely to an AI culture in which male-coded hype, tool fetishism, self-promotion and promises of efficiency connect better than care, context awareness and quality judgment.
If that’s right, then the reflex “come on, women, use more AI!” clearly misses the problem. It asks women to adapt to a culture whose very standards are questionable. My sharper demand goes to the organizations. Stop rewarding AI use as an end in itself! Learn to recognize quiet, good AI use just as readily as loud AI bragging. Build cultures where judgment counts, not volume.
And so I want something from the people who shape exactly these cultures. From CHROs, from the Top Voices who set the tone on LinkedIn.
Take Colette Rückert-Hennen, whose post opened this piece. She is anything but a poor address for the topic. As CHRO of EnBW she has anchored diversity seriously for years in recruiting, development and organizational design, with concrete goals, not just symbolic politics. And yet her voice stays inside a frame that is typical for many women in top management. Modern DEI vocabulary, reform from within, progressive enough to look modern, and pragmatic enough not to open up any fundamental questions of power.
And right now, that is no longer enough for me. When something as upending as AI enters the working world, we need people who are braver than that. Now is the chance to overcome old patterns. And it is the danger of coding all the disadvantages of the past deep into the system once more, if we don’t steer decisively against it in the AI tsunami. Far too many speak from inside the system, with its narrow, hard-coded perspectives. And of all people, those who may be showing the smartest responses right now are told they have to become different. I’d rather have them as allies who change the system itself. That closes the circle back to the trigger of this piece, the well-meant appeal that women should become braver.
Because here the burden of proof flips. Usually women are supposed to prove that they’re AI-ready. In my ideal world, though, it would be the companies that have to prove they are mature, level-headed and capable of judgment enough to introduce AI well in a time of unprecedented hype.
Maybe we’ve been talking about the wrong gap the whole time. What I derive from all this is that there’s an important gap between AI bragging and AI judgment. We need to talk about that one far more. And I’m certain, and I have at least anecdotal evidence, that quite a few women have plenty of appetite for AI. What they understandably have less appetite for is what the work cultures described here make of it.
A word on my own behalf, fully in the spirit of this piece. On June 26 I’m running another Lunch & Learn on “Introducing AI is culture work” (German language). The Gender AI Gap is just one example among many there. It’s about the larger thesis: Whoever wants to introduce AI well can’t get around the culture. If you like that thought, feel free to come along!
And stay on it! Bleib’ dran!
Warmly
André
That's it for today. Please do forward or recommend the newsletter ➡️ ✉️, it would help me a lot. Thanks for your attention and see you next time!
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.