🎙️ EP62: Tech Is Not Neutral: Mara Bolis on Building an Equitable AI Future… | InvestHer Podcast

Hi there. Welcome or welcome back to InvestHer Podcast. I'm Anne Ravanona, your host and funding friend, and I'm so glad you're here. InvestHer Podcast exists for one reason: to share the real, unfiltered stories of funded women founders, investors, and ecosystem changemakers from around the world, and to give them the visibility and airtime they deserve. Because here's what we know: women are building extraordinary businesses. And yet their stories are still too often missing from the conversation.

We're here to change that.

Every episode, I sit down with someone who's in the trenches of building, funding, or backing women-founded ventures. And we talk about the real stuff, the pivots, the rejections, the fundraising lessons nobody tells you about, and the wins, because there are plenty of those too.

Now, I can't wait for you to hear this.

Let's dive in. AI might be the biggest opportunity of our lifetime and the fastest way to lock women out of the future if we stay on the sidelines. AI is being sold to us as neutral and inevitable, but when you layer it onto an already unequal world, it can supercharge injustice just as fast as opportunity. Today's guest, Mara Bolas, founder and CEO of First Prompt and fellow at Harvard's Berkman Klein Center for Internet and Society, has spent her career proving that systems are never neutral, from global finance to development policy. In this episode, we unpack how AI is really being built, why women are using it less at work, and how Mara's idea of fierce ambivalence can help you use these tools powerfully without handing over your power. Enjoy.

So thank you everybody for joining us for this week's episode of InvestHer Podcast. And boy, oh boy, are you ready for something amazing, for an amazing lady. Today I want to welcome here Mara Bollutz, who is founder and CEO of First Prompt and fellow at Harvard's Berkman Klein Center for Internet and Society. Welcome, welcome, welcome, Mara.

Oh, I'm delighted to be here, Anne. Thank you so much for inviting me on.

Well, I'm so excited. I can't wait for our listeners to hear everything you have to say because when we first spoke last week, I got so excited, as I said to you earlier on, I could have talked to you forever. Now, what is amazing about you, Mara, is that you have such a breadth of experience. And I want to start by letting the listeners know a little bit about yourself. So can you tell us your background, where it all started? And then we're going to deep dive into everything AI, women, gender, business, staying ahead and, and staying in the game. Over to you.

All of my favorite topics. So, thank you so much again for having me on. Yes, it's been a circuitous path, and it has taken me through a couple of different disciplines. So, I started my career actually very interested in international development, and I got a degree from Georgetown in international affairs and began working in international development finance at a company called— well, it's a US organization, a US government organization called— it was called OPIC at the time. Now it's called the Development Finance Corporation. And while I was there, I became very interested in the— I was working in the political risk insurance side, and I became very interested in the finance angle specifically. So I went to get my MBA in finance. So I was in the finance field for a while, and my I'll jump forward to when I was working at the Federal Reserve Bank of New York in the Emerging Markets Group. And that experience really allowed me to kind of combine the micro of a transaction internationally to the macro of how financial systems work as a whole and as disparate but connected parts. And one thing that I found really fascinating during that time is that I was doing a lot of work looking at these oligopolistic financial systems in many emerging markets where you have a number of big financial players. They're all funding these very big companies, right? In this sort of incestuous web that ends up creating financial crises oftentimes when one company goes down and then takes all the other companies with it. And then at the same time, there's this rise of microfinance, right? And then, but then there's this already evident gap of the missing middle, which probably many of your, many of your listeners find themselves in sometimes when they're like, microfinance is not appropriate for me.

Mm-hmm.

But then I'm not a massive corporate. And so where do I get the type of financing that's right for my business? And that became so structurally clear to me in that work. I left the Federal Reserve Bank in New York. A few jobs later, I ended up at Oxfam, at the International Development Organization, Oxfam. And there I took kind of that background and then applied it to, or started looking at women's economic empowerment initiatives. And with the understanding that, um, that there is this, um, unfair, um, unfair, unequal distribution of resources from my perspective.

Beautifully said. I like the way you say it like that, um, unequal and unfair distribution of resources, because when it comes to access to finance for women entrepreneurs, it's not about funding them just because they're women. It's because They need access to resources to grow the same way as male entrepreneurs do. And that currently is not the case because of the system. I love the way you got that systemic thinking. When you went to Oxfam, did you— and you started to work on women's economic empowerment, was that part of your mandate or did that become part of it? You folded it in because of your passion?

I folded it in because of my passion. I was lucky enough to be invited into a leadership training program on the advocacy side. I was— I did not have an advocacy background. background at all on the advocacy side. And there we did some really deep work about kind of our why individually. And I realized that a lot of what brought me to my work was actually being an only child in a relationship where my parents— where my father had really all of the decision-making power because he was the person who earned the money.

Yeah.

And it was super clear. I mean, my mom would say, well, I can't possibly make that decision about purchases large and small because she said, I'm not the one who makes the money. And she would get judged for that. I mean, by my father and by other people. And so that kind of kernel of unfairness where she was never— she was serving such an important role in terms of being the center of our family, you know, making sure that we were— we had everything that we needed. And the fact that she didn't get the respect that I thought she deserved just hurt me. And then, you know, when I sort of made that connection. I was like, oh, Sam, where the— that's sort of the beginning of all of the reasons why women don't get financed, they don't get funding, their businesses aren't taken as seriously, um, when their capabilities and their instincts are so sharp, right?

Yeah.

And then you, and then you, then you like leap forward to the research and find that women's default rates are lower than men's.

Yep. Yep.

And you just think, what on earth is going on here? Right? So it became— once I became kind of both connected at my heart, but also looking at the research, it just became like a messianic sort of journey to try to figure out ways in which I could bring a gender lens to different systems of inequality, both because it's the right thing to do and because it's the smart thing to do.

And so when we— and I, that resonates with me deeply as well from my personal experience, which I won't go into here. I've said it in other podcast episodes, but you know, what we see when we're growing up does shape our belief system.

Yeah.

And therefore what we think we are owed or what we can ask for or what, what, what are our rights. And, and that actually cannot be correct when we're seeing it very young. In our conversation, you mentioned a story about a woman in Guatemala. Tell our listeners that story because it really, really resonated with me.

Yeah, so we were in Guatemala visiting. Uh, we would fund the women's savings group— groups programs, um, which, if, um, for your listeners, if you're not aware, they're, they're called different things in different countries, but the basic The principle of it is that women will get together and they will save, save and lend to each other on terms that they themselves determine.

Mm-hmm.

So it's an alternative to the formal financial system where they're often treated unfairly, and they create their own systems and microcosms in order to build power together. And they're very, very cool. Women will— they're elected to leadership positions, and they practice public speaking skills, and many of them then lend money to each other to start their own businesses. And it's just— this is— it's a transformative thing to be a part of and a very inspirational thing to watch and to fund. In any event, we had gone to Guatemala to go visit this programming, and we met this woman. And she said that before she joined the savings group, she was— it was a very rural location in Guatemala, remote. She said that she had never once said her name out loud.

Wow.

Because her life had been so private and so insular that she never met anyone who didn't already know who she was. And then she joined the savings group, which then invited her participation with women outside of her immediate village, right?

Hmm.

And where for the first time she had to kind of establish her individual identity. And the transformation was so massive in her that someone then on the sideline, one of her colleagues, her friends said, and now we call her the Tiger.

Which is amazing. Full circle moment.

Yeah, it's beautiful. And they're like, she won't stop talking though. And that, that like question of not only never being asked your name, but also never being asked, so what do you want to do with your life? Imagine never being asked. Imagine never, ever anyone ever asking you, what do you want to do with your life? It's something that we take for granted. But for so many women, it's just expected.

Like you're going to be in service of others.

You're going to be practically servile, right? Um, unless you are— you find an opportunity to kind of break out of that closed system, um, and then, and then, um, you know, have a, have a, have an experience like, like the one that she had.

Ah, so true. And I think it's important to set the scene as to what— where your why comes from, uh, in order to better understand where it's brought you to, what you're work now. So tell us a bit more then from that experience at Oxfam. You then went on to work at, uh, Harvard Kennedy School, so you were teaching there. So tell us a bit more. And then that was right at the ChatGPT explosion moment, you were saying. So, okay, so you're on the cusp of, of AI— well, not AI, but it'd be going, going mainstream. So tell us about that moment and how it helped shape and pivot you into what you've become now and your work.

Well, I'll tell you one story that I didn't tell you the other day, but I think you'll appreciate it. So what I was doing in the Kennedy School is I was teaching a student seminar on bringing a gender lens to development policy and practice. And very inspired by Caroline Corrado Perez, who wrote the book Invisible Women, who I know it's on your shelf as well.

It's on my shelf, yeah.

And kind of bringing that lens into the policy arena. And I was teaching— I'll get to the AI part, but I think you'll appreciate the story. My sort of my getting back to my messianic purpose, it's to bring a gender lens into systems that are seen as gender agnostic. And so that was very much the case at Kennedy School, that the gender lens is not particularly evident. And I was teaching the seminar and the Prime Minister, former Prime Minister of Côte d'Ivoire, Patrick Achi, is a good friend. He was planning on teaching a seminar as well. And he had gone to the leaders of the head of the Center for International Development where we were both fellows. And he said, I don't really know how to teach a seminar. I've never done it before in my— And the head of Center for International Development said, oh, well, Mara is just starting hers. Why don't you sit on hers and you can you know, evaluate for yourself.

We have to do it.

So he comes into the class and he sits there at the back of the room. And I'm teaching this class on analyzing from a gender perspective and sex-disaggregated data and, you know, and how women experience these different— these systems differently, et cetera. And sitting very patiently. And then I come back to class, I come back to him at the end of class and he said, Mara, I don't think I should be in this class. And I said, why? And he said, I'm the only man. I'm probably making people uncomfortable. And I said, no, I think you should come back. And not only should you come back, but now you know how so many professional women feel every single day.

Not just one hour.

And then he said, and he looked at me, to his credit, he looked at me and he said, yes, now I do know that. He could have been mad. I mean, how many people say something like that to a former prime minister? He couldn't even be mad. And he came back and he stayed through the entire class, the whole course. And then we started a program together in Côte d'Ivoire in the summer because he had recognized that he hadn't, while he was prime minister, he was still serving in office but in a different capacity.

Yeah.

That she hadn't paid enough attention to the unique needs of women.

So this is what I— this is—

so getting back to the AI question, there are all of these spaces that are misinterpreted as being neutral, right? And often when you put the technical layer on it, not only is it interpreted as being neutral, but also a little bit too challenging for the average person to understand. Right. So were you to think there might be bias in finance or there might be bias in AI, you're likely to get a tap on the head and be told, well, you don't really understand how it works.

Yeah.

Right. I've seen this so much in the finance realm. I set up, when I was at Oxfam, a small impact investing fund that was a loan guarantee facility. And we would go in to speak to banks that We were looking to partner with, and we— the patronizing responses that we would get where people would say, we don't discriminate. And then we'd say, well, did you look at the— do you look at the data from a sex assessment? No, we don't look at the data from sex. We just know that we don't discriminate.

Oh, and I can tell you a story. I'm like, when I was 21 in France, And I went to a bank to set up a bank account, and I set up my normal bank account, and I said I wanted to set up a savings account. I said, because one day I'm going to become an entrepreneur. And the bank manager laughed in my face.

Oh.

Like, he literally said— like, he just laughed. And of course, he didn't get my business. I went to another bank who didn't laugh in my face when I said that. Yeah. And that was like Okay, that's 30 years ago or something now, but like, that's not long ago. That's not long ago. So, and that's one of the many reasons why I also work in this field, because I've faced this sexism every day in every capacity— in home life, work life, community life, services life, customer life. It's just systemic, endemic, baked in. Because as Julia— no, sorry, Caroline, that you just mentioned her book, she would like, would say like, this is a man's world built by men for men. It has never been built for women. Going back to your earlier point where we women have been seen as to be there to be beautiful, pretty, in service, make babies, make families, and serve the man and let him go and do the career. Now, We are 2026 at the recording of this podcast. And unfortunately, in many parts of the world, and even in developed and, and developing countries, that's still the case. It's absolutely like, certainly when it comes from a societal perspective, maybe not technically, women have achieved a lot. But we're so far still. So this is why it's important that you— what you're going to share with us around AI, because that bias is baked into every single system of our lives from the minute we wake up to the minute we go to bed. And let's remind people what a bias is. A bias is when you make a decision based on a perception, not on fact.

Uh-huh.

And so it's those perceptions and assumptions that just get embedded, embedded, and repeated, repeated. So I just wanted to put that in there before we go any further, to say that how important it is that in the different instances where you have, in the different institutions, organizations you've been working and circles you've been working in, you've been spotting those things, calling them out in an appropriate and catalytic way.

Yeah.

You know, of course, not in the face. It's like we cannot affect change on the system if we're there pointing fingers and just burning our bras. We gotta do it differently. So, oh, and now I'm being a bit facetious and a bit of it silly because, yeah, again, that's probably a huge assumption of what a feminist is. And again, I remind everybody, a feminist is somebody who believes in equal rights and equal resources for men and women. It doesn't mean that we believe women are better than men or that men are better than women. So I just put that in and remind ourselves those bases before we go any further.

Absolutely, absolutely, absolutely. And so, right, so when I was at Kennedy School, it was at the beginning of the ChatGPT moment. And I was just so surprised that the dialogue around, you know, arguably one of the most elite public policy circles in the US—

And in the world.

It was zero about inequality.

Yeah.

When it was just so evident to me that you're layering this instrument that's going to supercharge intelligence for some, for those who have access to it. And then, and you're layering it into an environment which is already in the US very unstable. Economic inequality is fueling that instability. We are in a regressive cycle in this country.

Mm-hmm.

Where the gains that women have made is seen in many circles as going too far. We have lost the right to choose and have control over our own reproduction. And now you're putting in— now you're inserting into this equation, um, a set of capabilities that will replace jobs, right?

Mm-hmm.

And in environments where jobs are scarce, Many people still believe that men deserve those jobs more than women.

Mm-hmm.

And I am convinced that what will happen is that women are going to face even harsher, more sexist labor market environments when those jobs become scarce in every single economy. And that is what terrifies me.

Real quick, Anne here. InvestHer Summit 2026 is coming to Paris in October, connecting global scaling women entrepreneurs with investors, corporates, and others to scale faster. Save 20% with early bird tickets until the end of July. Join us at investhersummit.com. That's investheorsummit.com.

So tell us a bit more about First Prompt, why you created it, what you do, and then loop it back into, okay, how— what can we do?

Yes, we can do a lot. Um, so, so, right, so as I was identifying this issue So I'm of the belief that it's very difficult to be a credible advocate on a particular topic if you don't understand that topic very deeply.

Mm-hmm.

And so as I kind of saw this unfolding in AI, I was like, oh, I need to understand how these tools work, what's the technical nature of how they're developed, Anyway, so I was looking for routes in as a non-technical person, and I found the routes that were available to me either very— not either, both very isolating and isolated and kind of landing in an impractical place, meaning I felt like I was asked to absorb all of this computer science kind of theory, When really I just kind of want— I mean, I need to know some of that as an advocate in the space. I need to know some of that because then you have— then you identify how bias exists and is replicated, etc. But as a user, I was like, this is just— it's not Microsoft Word, but it's also not getting a computer science degree from Stanford University. It's something closer to Microsoft Word. And I could not find the pathways that I wanted, so I created them. So I learned as much as I could about the tools, and then I started creating these kind of very inspired by the savings group structure, these flat environments where women could work together to practice the tools in community. And we were learning from each other. There was, it was a non-expert structure, very kind of peer-based. Right, where we're kind of sharing best practices, admitting when we don't know, etc. So I started doing that, and that sort of ramped up over time to be done within organizations and advising, um, uh, leaders who are trying to figure out how to roll out within their own, um, as leaders leading the transformation with their own organizations. That was one piece. But then the other piece is really important for me to connect that to the research and to the advocacy. To both be able to collect information about how women are experiencing this moment differently.

Mm-hmm.

And to then be projecting that out to like dispel this idea that women's experiences of GenAI are neutral.

Mm-hmm.

But in fact, women have different preferences and they see different risks.

Mm-hmm.

Than men do, particularly kind of elite men who've been involved in building all these things. And then a little bit separate and apart from First Prompt, then I started with a good friend, my, my own podcast, which is called Women's Planning AI, which I absolutely love. It's my favorite thing that I do. And our intent is to make AI, like, bring kind of the latest research and news to women, but in a way that's entertaining and really accessible. And we had a lot— we have a lot of fun. We laugh probably 45% of the time, and then the rest of the time is like really serious discussion about what's going on. And all of these are meant to be entry points to invite women in and to say, you know, this belongs as much to you as it does to any Silicon Valley, you know, tech VC investor.

And I think that's such an important message. We all have to hear that, that AI is for everybody, and you don't have to have have a computer science degree. You certainly don't. You don't have to be of a particular age.

No.

You don't have to be of a particular— you don't have to have any kind of academic background. I think from my experience and my steps into it, it's like you just have to kind of keep going, you know, push past your own fear. Just use the tool, keep using it every single day. And like, I basically am using them hours every single day because I'm myself seeing if I don't do this, I'm going to be so far behind and I won't be as capable to scale and to serve as many women as I want to. So therefore it's— and I think I know on, on another podcast, uh, or another webinar of a great woman entrepreneur, she kind of described this moment as the Kodak moment. Where Kodak, if those of you remember the name, like, you know, that Kodak were the ones who invented film, and, and they were the first ones to kind of— they were on the cusp of the digital camera, and they were there at the very beginning, but they hung on deeply onto, no, we're going to stay with traditional film, we're going to stay with traditional film. And then they got bypassed, and now they— I don't think they hardly exist, only just specialists. To provide old film, physical film, to certain niche segments. And they were the world leaders. And it's like very scary when we see that. So that's why I liked her metaphor of the Kodak moment. We are now in a Kodak moment, meaning that if you hang on to life as we know it, even last year, even 6 months ago, you know, you're going to miss out on everything. And it's like Again, on all of the YouTube videos, channels, podcasts, webinars that I've been on so far, people are liking this moment. And please, I would love your perspective of in time, as this is like the world just getting internet— internet, not even email, but the internet— and all the knowledge being available to everybody, but multiplied by a gazillion. Like it's a whole new level. And therefore, if you thought the world was disrupted by the internet 20, 30, whatever years ago, 40 years ago, just hang on to your seats because we are. And I remember getting a first taste of that maybe, I want to say, 15, at least 15 years ago. I was at the Global Summit of Women before AI was talked about in external circles, circles. And, um, there was IBM Watson who were showing how at the time they were already using the first versions of AI to analyze medical research on breast cancer. And so they were able to basically ingest all of this data, find the patterns, and they were able to more accurately predict the, the mammography scans of women who had early-stage breast cancer than humans, than the doctors, because they had a bigger database to compare to. And I remember that, like, and that was at least 15, if not more, years ago, sitting there gobsmacked going, this is incredible. And then, okay, I left the conference, and then that was it, because it was not widespread. Nobody was talking. But now we have that. in our phone and more. So, so please, from your perspective, Mara, how should we— like, what do you think? How should we approach AI as women and as young girls? I would, I would love girls to be listening to this as well, and we should share it. How should we approach starting AI, going to use it? Why is it important? And then, you know, tell us, bring us more into your world as to how we can change things.

So I'll tell you a little bit more about how my thinking has evolved in this. So my initial concern— so, okay, the research shows that women use generative AI tools around 20% less than men.

Hmm. And this is—

there was a meta-analysis that was done at the end of 2024 by Harvard Business School. But then there was another research that was rerun recently. I can't remember where it was in the US and UK, and it was published in the FT about earlier this year, same 20%. So it recurs, right? There's some— and this is at work specifically. The usage has equalized around kind of more general purpose everyday uses, but for work. And my initial concern was one around the gender pay gap.

Hmm.

Which I am very concerned is gonna widen. If this, if this persists, if this usage gap persists. And, and I think that that then extrapolates out to, right, the projected revenue gains between women-run businesses and male-run businesses, right? There's kind of a corollary, I suspect.

Mm-hmm.

But as I started thinking about it more, And as I understood the transformational potential of this technology, which, you know, we're talking about it as not even the internet, but as electricity.

Yeah.

Right. And like, really, the industrial revolution. This is why the Pope weighed in. You know, the Pope— we have a recent podcast on our podcast called The Pope Has Entered the Chat. Like, even Trump felt like she had to win, right? That was a very big deal.

It's a great point.

And we cannot have women standing on the sidelines at this moment when everything is going to change. So absolutely, yes, at a micro level, women being able to keep up is critically essential. But as important to me is that women feel a sense of purchase and ownership over how our societies and our economies are going to evolve on a systems level, right? Like, we need to be at the table, and no one is inviting us in.

Yeah.

No one is going to invite us in. And so I take a lot of— there's a woman, Shirley Chisholm, in the US. She was the first African-American woman to be elected to Congress, and this was in the '70s. And she ran for office, and she has this very famous expression of saying, if they don't give you a seat of seat at the table, bring your own folding chair.

Yeah.

And that is what we need to do right now. We need to be bringing our folding chairs, and they do not want us there. But that should not deter us because we have all of the expertise that we need. We have our, our domain expertise. We know our fields of influence. And all you need to do is shape an opinion in terms of What's going on in the AI space, use it yourself, see how it might— imagine how it might influence your school system. How might it influence your business? How might it influence your sector? I'm being invited in to speak at a group of women in wastewater management.

Mm-hmm.

And they're asking me to come in to talk about how is AI going to change the wastewater management industry and what do women want that to look like?

Right?

The woman who's bringing me in is very concerned about the disruption that she sees in her sector to the administrative core, which are many of whom are non-college-educated women of color who she's— And she's got to find jobs. Are on the cusp of being disrupted. And it's not to say that you can stem the disruption, but what you can do is bring people along and say, this is how you can augment your work.

Yeah.

Right? Because it's the people who are going to be augmenting their work that are going to have the better prospects of staying involved and staying in the game. But you have to be part of the conversation. So what I talk about is I wrote a piece in Stanford Social Innovation Review, which I would love if you would attach in the show notes. We'll share it in the show notes. And it's called The Paradox, The Gender AI Paradox. And what I talk about is this position of fierce ambivalence where we have to be fiercely passionate about using AI to empower ourselves and others while being— maintain incredibly high expectations for how AI is implemented in our communities and our societies and our companies and how it's governed and how the frontier models are developing the tools and how they're being adopted to make sure that they're as free of bias as they possibly can be in a world that's structurally biased to begin with. Wow.

And I mean, that, that's like, it's a, it's a big ask, I think, in the sense that if we have— if we're talking to women who are taking the first steps in AI and who may feel very intimidated about starting to use it at work and at home, because I really think We need, we need to be using it every single day for everything, like planning. One of the key issues for women in our daily lives is that we carry so much more of the mental load for the family or the people that we care for, whether it is significant others or elders or children, etc. And we're the ones who are doing all the thinking and the planning and so on. And so AI now is an incredible tool to help take a lot of that off. As long as you just said earlier, we have all that knowledge inside of us, our own life experience and knowledge. So it's how we learn how to prompt it and say certain things, give it context before just saying, find me XYZ.

No.

Okay. You are like this. This is what I've done before. I have— I'm going traveling with a— which I did. I was planning my summer vacation. I'm going traveling. My mother, who's 76 and has some mobility issues, I have to factor that in. I have a dog. Da da da da da. And it was giving me all these things and double-checking all the things. This is like we need to be using it every single day in our home lives as well as in our work lives. And I think it's like it's I want to say it's a bit like when I came to access to funding for women entrepreneurs 13 years ago, discovering this big funding gap and saying, "Oh my gosh! We gotta help women realize there is a difference in in how we are seen and and how we access." capital. We need to know it. There is this funding game. We need to know the players, the rules, how to play it, not transform ourselves into something we're not. However, we need to know all those things in order to kind of navigate it. And I think that in the beginning, that's where the beginning of my work was all about demystifying. And I think this is the first step. I'm sure you do it, um, Mara.

Yeah.

It's like, how do we demystify using AI on a daily basis for our personal and work lives? I mean, tell us a bit more about— okay, because you're looking at this as kind of saying we gotta wake up. This is when you say the fierce ambivalence, like it's a huge opportunity and there are risks involved. And I want to go more down the risks part in a minute, but like tell us a bit more from your perspective. If we like, as we demystify this and say, hey, this is a huge opportunity for women, it's an opportunity and the risk. The risk being if we do not get on this train, we, we are going to be, as you say, even more in competition for jobs, right?

But I will say something that I resist. So this idea of women need to wake up, I think the women are awake.

Yeah, true. Thank you for correcting me.

Yes. And I think that they're— this is like, this is Caroline Criado Perez— women see faults in systems that they've had little part in building.

Yes.

Right?

Yes.

But they also rarely shape systems that they don't understand. So both of those things are true.

True.

Right? That's basically my ethos, right? So I want to fully respect because I have those same misgivings, the misgivings that women are coming into this moment with. They are not wrong, right? And there's— so, okay, so what are those, right? I mean, apart from the 20% gap in usage, I really wanted to understand what the underlying source of that 20% gap was, right? And the evidence is fascinating. There are multiple reasons, but we've already named bias. Women see bias in the systems, right? And they get turned off by the biased responses. They're worried about the environmental effect of what's going to happen with the water usage in data centers. They're worried that they're going to be judged as cheating.

Yeah.

Which men aren't. Men are not worried about this. And in fact, there's research that was put out by LinkedIn last year, this year, that showed that, that women get, um, are, are less likely to get praised for using the tools, and men are more likely to get praised for using the tools.

So that from the classroom, that those studies and the research that show that in, in children in classrooms 50, 100, I don't, maybe not, but 50 years ago, whatever, a long time ago, it's still manifesting even at the managerial and employee level.

Wow. And so women, so, and, and there was just research that was put out by, um, by a researcher that used to be at Facebook, and now she has her own firm. Anyway, I can give you the link. But it showed that when the identical resume was presented to resume reviewers, but with a man's name and a woman's name, and it was evident that the resume was made using AI, the man was like, oh, complimented for being innovative and thinking ahead using these tools. And the quote for the woman was she couldn't even write her own resume.

Jesus.

Okay, so the fact that women are reticent is not irrational.

Mm-hmm.

It is— they are paying— they are sensing that something is off, and they're like, I'm not gonna get screwed by this thing, right? So that's where the fearsome ambivalence comes in.

I get that.

I get that there are faults. And I say you have to use it anyway because you need to be in there shaping what it looks like. If you stand on the sidelines of your company, you're like, I'm not going to use it, I'm not going to get involved, then you're leaving power on the table. Other people are going to make those decisions without you.

Mm-hmm.

It's not to— it's not to neglect the fact that those— that women are experiencing this moment and these instruments differently. That is 100% true.

Mm-hmm.

And at the same time, you rarely shape systems that you don't understand. So you have to get in there, you have to be using the tools, you have to use them to empower yourself and others, and then be paying attention to how this is all rolling out around you. I mean, it's a big ask, right? Women already have care responsibilities. They have, you know, they have care responsibilities at work. You've got a million things on your mind, and then you're layering this in on top of it. I hear you. It's rough, but there are ways in. And I think that the tool, the key is kind of to try to break ourselves out of this like panicked anxiety mindset and just, and try to kind of break down the usage, understand the risks, understand how to protect yourself from the risks of usage, break down, break down ways to engage in kind of bite-sized chunks. that work for you and find a buddy. That's been really transformational for me is my podcast co-host, Logan Curry, who is brilliant. She and I very early on started just connecting regularly, like an hour a week, and be like, what are you seeing? What are you reading? What are you trying? What are you testing? And it built kind of this mutual accountability. around using the tools, that has been— it's changed, it's changed the game completely because trying to learn it by yourself exclusively is just not how women want to do it.

Exactly. I get that. I love that. And I think that's a very— that's a great point. So I just want to make sure, circling back, you had mentioned the different, the different things that we need to be aware of. So have we covered them all before we go any further?

No.

Yeah, so I want to make sure we come back to that. I think you've covered 2 or 3, but that's why you mentioned there'd be 5 or more. Yeah, so the things that—

so again, I'm not a, I'm not a doomer, but I'm a real— I guess I'm a realist. I mean, I'm not an optimist and I'm not a doomer. I'm a realist. And I use it every single day for hours on end, but I use it knowing a couple of things. So I'd say On an individual level as well as at a managerial level, probably the most important thing to know is that these tools do hallucinate.

Yeah.

And a hallucination is where the AI tool will confidently make up an answer that's inaccurate. And it does so because of the way that these systems are incentivized. They are incentivized to please the user. And they're incentivized to deliver quick results. And that will sometimes mean delivering something that is not actually true but sounds correct.

So there are some tools that are worse at that than others, and I don't know if you feel comfortable giving your opinion on those because I think—

I'm totally fine.

I think it's important we need to kind of say, okay, if, if we're starting from scratch, like there's the, there's the big mama that we all know that— I'll let you say the words. So who are the ones that we all know, uh, and use that we say, oh well, they're actually very guilty of that hallucination? And then maybe maybe the ones in the middle of the road that a lot of people are using in the work context, and then the ones that are more researchers and more, you know, developers and so on are using. Can you maybe give us a bit of a landscape just to demystify that further for our listeners to say, okay, this is the— this is part of the ambivalence piece that we're talking about and understanding the risks and say, okay, there are tools, they're morphing, there's built-in bias, and it can hallucinate because of all those different things. And because also It wants to please us, but some tools want to please us a little bit less. So can you give us a bit of a lowdown on, on like the ones that we should be— well, well, just your, your, your input on that?

Yes. Um, so they all, they all have hallucination risk. And so this is where I get back to my domain expertise point. And what I said to you yesterday, and I say it all the time, I would never take out someone's appendix using AI. It would probably help me do that. It would probably tell me how to go through the steps. But I'm not going to do that because I don't know how to take out someone's appendix. But if I am using it to research anything on women's economic empowerment, I know immediately when it's making something up. I know immediately because I'll see research that I would have seen that. And so that's why I think you can take some of the fear away from this because you will know.

Yes.

You will know. But, but, but to answer your point specifically on the tools, and I want to say they will all lie to you.

Hmm.

Um, from a, um, from a, um, so my favorite tool is, um, Claude, uh, which is by Anthropic.

Mm-hmm.

Um, I find that one to be the most precise, um, and the least hallucination-prone.

Mm-hmm.

And it gets— so there are different modes that each of these tools can follow. In the thinking mode, which is when you're dialing up the tool to be at its most kind of— its most circumspect, that's where you'll get the fewest hallucinations. So I would say Claude in thinking mode.

Yeah.

ChatGPT in free is not very good. And you will likely— I don't use it in free, but what I want— my understanding is that you will likely get a lot of hallucinations if you're using it in free. ChatGPT, at least originally, there's something called temperature, which is— temperature is like, imagine the most kind of stickler accountant that you've ever worked with in your life. Right? And imagine talking to that person who— a stickler accountant with no sense of humor and, and, and like, just is about the facts. So that's setting the temp temperature at 0. And then imagine like hanging out with your most artistic friend who's like about to join Cirque du Soleil. That's like a, a temperature of 100, right?

Mhmm.

ChatGPT would set itself up as sort of more middle of the road between those two. And you would, you would ask, why would you want the artistic side? It's because that's where some of the creativity comes in that gives you some really good responses when you're doing things like brainstorming, right? Where you're not actually looking for facts, but you're looking more for ideas that are outside of the box.

Mm-hmm.

Because that's pulling from different data sources and kind of mashing it up today in a way that's really exciting as sort of like a, Like an interdisciplinary standpoint, right?

Mm-hmm.

So ChatGPT, from my understanding, in free can be very— can hallucinate a lot. In its thinking modes, which are in the paid level, I think that's now SOL. The model is called SOL, I think, for its seeking modes. It's better. What you want to do is when you're doing something that's fact-oriented, you want to tell it to never hallucinate and show you the evidence.

Yeah.

So link to existing sources that you then click through to make sure that they're correct. And you always need to be doing the double-checking in the actual source document. I've had my biggest hallucination problems with Gemini in Google, which is something that a lot of people use if they're in the Google Suite. I find it to be very quick and very inaccurate, so I barely use it even though I have access to it. And then, and then the other thing that people use a lot is Copilot, which I don't know very well, but I understand that it's getting much better because they're, I think, training their own models now, which should be quite good. I think people have found that to be just the tool to be a little bit flat.

And then there's also Perplexity. Perplexity. There's also Perplexity, Max. So if you're a researcher, if you're going further and you really want to have everything documented and to check out the sources of everything and to be able to go through it and to prepare very detailed, more academic. Perplexity Pro and Perplexity Max, I believe, are— well, I know my colleagues in the W20 are really using that a lot.

So, but you must pay for it because in the free mode it's not very good. Yeah, no, it will still provide resources. I have gotten hallucinations with Perplexity in free. So again, the free tools. So this is getting to the second risk that I want to mention. Is that the free tools, um, with the free tools, first of all, it's a better— it's a, it's a poorer quality product. But also, it, it, as we say in social media, with the free tools, you are the product, not the consumer. Yeah. So you— people need to know that because—

Explain that a bit better because for some people who may not know, understand what you mean by that.

So the way that they— these models, um, their, their business model and the way that they perfect their instruments is through gaining as much data as humanly possible. And the trade that you're making in, um, in say ChatGPT for free is that you have no ability to turn off You are giving it all of the— all of your data.

Yeah.

For the purpose of improving the way that their tool functions. You cannot turn the privacy settings on in ChatGPT for free. It doesn't exist. In Claude, it does. Yeah, yeah, yeah. Perplexity is different because Perplexity is actually built off of other large language models, so I'm not exactly sure about its data setting function. But anyway, the point is I do not put anything into my AI tools that I wouldn't be happy seeing in a newspaper one day because I do not— I do not trust them with my data.

Yeah.

Even Claude, which I think is the most ethical and is the most trustworthy, I'm very circumspect. In Claude in free, you can turn on your data privacy settings. In the free version, which is why I think that's my number one recommended go-to in terms of a tool to use. But for every single tool that you're using, you must go in and you must turn on the privacy settings.

And you must— and I would suggest, like, I know I've moved on to Cloud Cowork now for a while. I've been paying for it, and I've now even moved to the maximum of the individual level because I want the power. I'm now in it every day, and I'm sitting, and sometimes it's like like, ah, because I'm making it work so much, it's like, ah, it's like steam coming out of the ears. And I'm like, no, no, no, you keep working, buddy. And so, yeah, it, it, yeah, it's worth it. I just want to say to people, it's worth it. Okay, so when we were— okay, have we covered then the other risk factors? Because I want to, I want to circle back to that, um, because there was a point you made in our conversation as well that I want to come to. Yeah.

So then I want to say on an organizational level, Yeah. One of the biggest risks that enterprise managers are facing is what's called shadow usage, where you've set a policy where you're saying that there's either no tools are available. Believe me, I've seen that in organizations. Or where you've mandated a particular tool that where you're workers don't like that tool as much, and where they are using their own private version of the tools and feeding your company data potentially into those private versions. So then that data is traveling outside of your company. And so I think the biggest way to manage a risk, which still a very high percentage of enterprises do not have an AI policy, Yeah, a shocking level do not have an AI policy. And so in that circumstance, if your employees are putting private data into their own— putting company data into their own private tools, it's really your fault. You need to be communicating what— how you want your employees to be using the tools. And there are low-hanging fruit ways that they can use even their own private settings that doesn't involve your proprietary data. But you need to make that distinction clear of when you can use— when you must use enterprise tools, when they can use their own tools if they want it, and when no data must be hitting these systems in any account because they are— the data is just way too sensitive.

Mm-hmm.

But setting up policies and communicating those out, and even maybe it's not even a policy, maybe it's a guidance memo that you can be updating regularly as these things change, is critically important not only to protect yourself, but also, but also in order to make sure that there's usage parity within your organization. Because the research has shown that when there's When there's not transparency, when there's an opaque message, then men will use it and women won't. So it's up to you as enterprise leaders to be modeling the type of behavior that you want to see throughout the organization and making that very clear to your workers in terms of how you want them to be using it.

And I think you, you also, um, You said something as well about a lot of data and privacy risk companies are carrying, um, like frontier models that are shipped 60% ready. So they're, they're, they're, they're just using AI, they're shipping it, it's not fully ready, this probably doesn't have all the privacy baked in. And you had a really cool phrase. What was your phrase that you, that you used? Do you remember about the FDA? You said something about— Well, so yeah, there's—

we have no Food and Drug Administration clearance of any of these tools, right? So can you imagine like some medicine being shipped when it's, you know, 60% ready and then we're unwitting beta testers of whether or not the medicine is safe without being 100% sure? There's no— for the US models that are being shipped, there's no agency Hmm. That's ensuring the safety of these tools for the companies that are ingesting them. And so all of the risk that's coming from the tools shipping at 60% readiness are just flowing downstream into the organizations themselves and being imbibed directly. So it's up to enterprise leaders to be as vigilant as possible about what the risks are. And now, again, I'm I'm not a doomer. I, I love using these tools. They're absolutely fantastic. They can transform— they've transformed my work. I get more done in the course of a day. It is wild how much I can get done now, and I absolutely love it. At the same time, I have my own guardrails set that I'm very comfortable are protecting me and protecting my reputation, right? So I think that's one thing to be very, very concerned about. The other thing, getting back to the hallucinations, is There are real risks that companies have had to manage when their employees have put out hallucinated materials.

In their brand names.

Yeah, under their brand names.

And then it gets tested through a machine and say, oh, that isn't actually— and none of it's true, or it's not correct.

And there was—

you've got all the legal stuff behind it.

There was an article last summer where the Chicago, um, Sun-Times Published a list of the top 20 summer reads and half of them were fabricated.

Oh my God.

Can you imagine? Wow.

Yeah. So there's a huge reputational risk. Yeah.

Reputational risk. So then this gets out of like the realm of kind of policy to, to like what process changes need to happen within your organization to ensure that the copy that's developed From a marketing standpoint, you can absolutely use AI for marketing purposes. Yeah, marketing copy, there's no question. But as— but assuming that what it puts out is correct is just not a fair assumption. You need to go back and you need to be doing— and this is where I get— yes, it gets to the domain expertise, but also gets the critical nature of the human involved in the process.

Yes, and I think that's super important. Thank you for saying that, because I think we need to remember These are machines working. Yes, we've been programming the machines, but they're machines talking to machines talking to machines. And if you take the human out, that's the problem. So that's also when we're using the tools. And for example, like, I have the guardrails you're talking about. It's like when you go in your Claude Cowork or whichever tool you're using with the paid versions with more guardrails, you get to choose and say you can see all these files but you cannot edit and you cannot send any emails. You put the guard right here. Yes, you can see all my emails, but you cannot send anything. You can't access. No way can you send, no way can you edit. You cannot be changing any files on my, on my desktop without my permission. And that's to the point that I'm walking around the park and I'm getting Claude going, I need your permission on this, I need your permission on this. And I'm like, oh, you know. So it's like, okay, I'm, I'm also learning where— how far can I go with this? And like, yeah. But okay, so now I want us to zoom out. Well, have we covered— we've covered a lot of the risks there because we're coming to the end of this episode, and I want to make sure that we cover an important thing, which is more zooming out and just AI and funding, um, and how at the moment the world is— the majority of investing is going to fund AI all over the world. That's where the investors are putting their money. So if you are a woman entrepreneur and you are not using AI, you can— doesn't mean you have to have it as your core business. But if you're not using it, you are certainly going to miss out because the investors will say, well, why not? You're not serious if you want to scale, because an investor wants to invest in a company that is scalable beyond yourself, right? They want to get a return on their investment. So if you're not using AI, they will not take you seriously. So I wanted to get your take on, on, on this last part around What advice would you give to women entrepreneurs when it comes to getting on the AI train? Let's just call it a rocket. It's not even a train, it's a space rocket at this stage.

Well, so I think that that's absolutely right. I think it's going to end up being a diligence question. It's going to end up being a question that you're asked in diligence.

Yeah, it is. Yeah.

And, and I think that the smart investors will like an entrepreneur that says, I'm using it like this, this, and this. And the guardrails that I've put in place are this, this, and this. And the policies that I have in place for my employees are this, this, and this. I don't think the answer is, oh, I'm using it for everything all the time, and I'm building a one-person company and all the rest of my employees are agents.

I think that's so—

it's so foolish. The agentic layer is very, very interesting. I use it for a lot of things. I think that there are going to be a lot of, a lot of companies that are going to blow up and burn because their agents start promising all sorts of things that are not realistic, that are completely hallucinated. Like, anyway, so yes, I think, I think so. A few. So absolutely think it's going to be a non-negotiable diligence question. I also think it's not just like get on the rocket, but it's also like What are the safety controls that are built around my rocket to make sure that I don't blow up when I leave the security layer or whatever? And you can use AI to prepare for all of these questions.

Yes.

One of my favorite ways to use it is as a thought partner.

Yes.

This was— I did a training with Awaken Hub earlier this year, and there was an— and we did an exercise where we trained the entrepreneurs in the AI assuming the role of an investor that was asking them due diligence questions.

Mm-hmm.

And one of the entrepreneurs came up to me afterwards and she said, I have a, I have a pitch next week. And it asked me questions that I wasn't, weren't in, weren't even in my line of sight.

On my radar. Yeah.

Even in my radar. And now I'm better prepared for the pitch.

Yeah.

This was a, a 1-hour session.

Yeah.

Um, so I'd say if, if you're trying to struggle how to use it, start by using it to sharpen your own thinking.

Yeah.

Right? You can start by uploading, um, in a, in a version that you have the privacy settings turned on, upload your strategy, upload your annual plan, upload that background of an investor that you're looking to pitch, and then say, hey AI, I want you to play the role of the investor. Now ask me 15 questions about my financing structure. Ask me 10 questions about my operational plan. It will drive you through the paces in a way that will prepare you more than you could have imagined.

Hmm.

If you're someone that's been using it for a while, what I want you to do is I want you to go in and I want you to ask, hey AI, reflect back to me the ways in which I've succeeded in the last year that I may not even be aware of.

That's a great question.

And say, be fact-based. Don't suck up to me because I'm smart and I'll know the difference. But tell me just fact-based, what have I achieved that I'm not giving myself credit for? This— I did this recently and I was shocked. I was so— it made me feel so proud of myself because I think particularly for women, we hold ourselves at such a high standard.

Yeah.

When we achieve things, we're like, oh yeah, anyone could have done it. But you see it in the aggregate, and you're like, oh my God, look at that!

Which which can be a great tool also if you are employed and you are preparing your your annual performance performance review. You can be feeding it in and saying, okay, these all the projects I did. Da da da da da. If you're continuously working on the model, it gets gets to know you very well, um, in your work context. But you can ask it to, to, to, to prepare, okay, the list of all your achievements, etc., etc., you learned and whatever. And, and it's— I've had chats with AI on personal things, like, and I was like, wow, it did— like, it says, like, you did this, this, and this, and you've made this progress, and you forget. It's exactly that, because it remembers. It's a machine and it remembers. It doesn't care It really doesn't care about it whatsoever, but it remembers, and we forget. So because there's so many different things going on, so that— I agree with you— can be very cool. Okay, so if you could change one thing, Mara, about how the AI era is being built so that women aren't leapfrogged, what would it be?

Well, I'm bringing this up especially because we didn't touch on it quite as much as I would have liked to, and that's my own fault, is for the systems to be fair and unbiased. Because I think that has a supply and a demand effect. Yeah.

Right?

It puts women off of using the tools. I ask, and this is also, I say this as a warning for enterprise leaders, I asked it to prepare a picture for me of 3 business executives looking worried, and it created an image of basically triplet white men where I— then I make the joke like it's more likely that I would be in a meeting with triplet white men investors than a single woman or a person of color according to AI. So I wish that Was remedied. And I do think that the tools are getting better at this. There is a way to train them over time where I think that they have multiple incentives in order to get better at this. But then the— I think it's something that— and then, you know, in the algorithmic— in tools that are then used in order to identify who's qualified for what position, right, in the HR tools, who should be promoted, there's so much systemic bias that's built into those evaluations. And I also raise this here because that is a risk for organizations that they must manage. Just because you bought a tool, right, if that tool you're using it now to sift through resumes, you are still liable for the— well, I don't want to be a lawyer, but my understanding is that you are still responsible For if there are biased results that come through that tool.

Yeah. And I've seen— it's so interesting. I was on a learning webinar, again, on Code Cowork last week, and the organizers created a skill to do a raffle. So with the people who were on, who were in the chat, who had signed up and were actually live on the webinar with us, it had created a skill and it was raffling, I don't know, a free ticket to XYZ. And I just noted, and I'd been on a couple of the webinars, every single time they were all men that came up. Now, okay, there were some women who were in the audience, but every single time. And I said, you got to change your skill, guys. You gotta, you gotta train it that at least it's gender equal when it's looking at the, you know, when it's looking at the sample. If there's only 10 women in it, we'll take it 10, 10 men, 10 women, and do the raffle that Like, you know, and I was calling it every time I'm on a call with them. I call it out. And this is part of this changing the system. It's calling it out. And because men just don't see it, they don't see it.

This is getting back to Caroline Criado-Perez. She says this is a function of, of, of, um, this is— I don't think it's deliberate. I think it's a function of not seeing. Yeah, it's a function of not seeing and not understanding. But then we need to be at those tables calling it out and saying, you're missing something and it's creating an organizational risk for you.

So as we close up, I would like to ask you one quick question. What gives you hope right now when it comes to AI?

I will tell you what gives me hope. What gives me hope is that— I'll tell you a quick story. And I know we're running out of time. But Time magazine profiled a woman this year. Her name is Sandra Drummond. She's a nurse in North Carolina. And she saw how AI was being implemented. AI tools were being implemented in their hospital system. She saw that there were risks that were coming along with them. So she organized with her nurses union that going forward, nurses are going to have an increased say in terms of which AI tools are adopted and how they're implemented.

Well done.

Wow. And this is not only good for the nurses union, it is good It's good for the hospital system.

Mm-hmm.

So I think increasingly what's going to happen is that this co-created, co-deployment, which on the one hand can be seen as slowing up the process, I think there will be a number of case studies that demonstrate that to go slow is to go fast.

Yeah.

And that participatory nature of having people who are closest to the risks involved in how these implementations happen, is actually going to turn out to be best practice, and that's going to give women more and more opportunities to have a say in terms of how the future is written.

Great example, great example, Mara. So where can people find you at First Prompt and on Women's Planning, etc.? So tell, tell everybody here.

Yes, please, please. So come find me at Mara at First Prompt. www.firstprompt.net and Women's Blaming AI, which is www.womensblamingai.com. And we are on Spotify and on Apple iTunes. And we would love to hear from you. We love our audience listeners, and please spread the word about both.

Wonderful. Well, thank you so much, Mara. This has been It's been a whopper conversation. This is one of our absolutely— I think it's our longest episode ever. And I think for a reason, because we're at the cusp of major systemic and societal change, and AI is at the center of it. And so this discussion, we've covered several bases, and that's why it's so rich. Thank you so much. Thank you so much, Mara. I'm going to give you what I give to some of my guests, an open invitation to come back. We will come back and revisit this maybe next year, get your take on what has changed since, since this conversation. And thank you so much for everything you're doing to help change the system. It's been such a pleasure having you with us.

It's been such a pleasure being on, and thank you for all of the work that you do.