Latest / Work It Like A Mum / AI at Work – Threat, Tool or Opportunity for Women?

Transcript

AI at Work – Threat, Tool or Opportunity for Women? · Work It Like A Mum

• 0:00
  Hey, I'm Elizabeth Willis, and I'm obsessed with helping as many women as possible achieve their boldest dreams after kids and helping you to navigate this messy and magical season of life.
• 0:14
  I'm a working mum with over 17 years of recruitment experience, and I'm the founder of the Investing in Women Job Board and Community.
• 0:22
  In this show, I'm honoured to be chatting with remarkable women, redefining our working world across all areas of business.
• 0:30
  They'll share their secrets on how they've achieved extraordinary success after children, their boundaries and balance, the challenges they faced, and how they've overcome them to find their own versions of success.
• 0:43
  Shy away from the real talk.
• 0:45
  No way! Money, struggles, growth, loss, boundaries and balance.
• 0:49
  We cover it all.
• 0:50
  Think of this as coffee with your mates mixed with an inspiring TED talk, sprinkled with the career advice you wish you'd really had at school.
• 0:58
  So grab a cup of coffee or a glass of wine, make sure you're cozy, and get ready to get inspired and chase your boldest dreams or just survive Mondays.
• 1:07
  This is the Work It Like a Mum podcast.
• 1:12
  This episode is brought to you by Investing in Women.
• 1:16
  Investing in Women is a job board and recruitment agency helping you find your dream part-time or flexible job with the UK's most family-friendly and forward-thinking employers.
• 1:27
  Their site can help you find a professional and rewarding job that works for you.
• 1:33
  They're proud to partner with the UK's most family-friendly employers across a range of professional industries.
• 1:40
  Ready to find your perfect job?
• 1:42
  Search their website at investinginwomen.co.uk to find your next part-time or flexible job opportunity.
• 1:50
  Now back to the show.
• 1:56
  Hello and welcome to the fourth session of our International Women's Day uh summit, Give to Gain.
• 2:03
  This is a very exciting session today on a very, very hot topic on AI.
• 2:09
  So we're going to be covering AI at work.
• 2:11
  Is it a threat, tool, or opportunity for women?
• 2:14
  And we're going to be talking about how AI is already changing roles that women dominate, where bias shows up in AI tools and decision making, how women can engage with AI in a way that protects and strengthens their value at work, what employers need to think about now to ensure AI adaptation doesn't widen the gap.
• 2:34
  So AI isn't coming, it's already here, and it's shaping women's roles, workloads, and career paths.
• 2:40
  And today we have got an incredible panel with us.
• 2:44
  And I'm going to just go around really quickly and let everybody introduce themselves.
• 2:49
  Heather, I'll come to you first.
• 2:52
  Oh, lovely.
• 2:52
  Thank you, Claire.
• 2:53
  So my name's Heather Black.
• 2:54
  I'm the CEO and founder of Supermums.
• 2:57
  Over the last 10 years, we've been helping women relaunch their careers in technology.
• 3:01
  I'm very proud of where they where they've what they've accomplished and everything they're achieving now in their new careers.
• 3:07
  Perfect.
• 3:08
  Thank you, Heather.
• 3:09
  And to you, Lasha.
• 3:11
  Yeah, thanks, Claire.
• 3:12
  I'm Lasha Kelly.
• 3:13
  I'm an AI transformation partner with We Are Agentic.
• 3:17
  We're an end-to-end AI transformation agency, and we help businesses, teams, and also specifically women as well get confident and upskill with AI and feel prepared for this future that is here.
• 3:31
  Perfect.
• 3:32
  Thank you, Lasha.
• 3:33
  And Andrew?
• 3:34
  Hi, I'm Andrew Bailey.
• 3:36
  I'm CEO of Rathbun Results.
• 3:37
  Um, my background's Formula One, and I used to work for one of the world's leading AI labs.
• 3:41
  Um, but at Rathbone, we basically check, we sort of pair Gen Z and Gen X and millennials to create really great AI solutions for people from adoption right through to actually building models.
• 3:53
  So a sort of cradle-to-grade service.
• 3:57
  Perfect.
• 3:57
  Thank you, Andrew.
• 3:58
  And Vanessa?
• 4:01
  Hi, hi everyone.
• 4:02
  I'm Vanessa for I'm an economist with a background in statistics and data science at Auxera.
• 4:09
  Uh Auxera, we are an economics and finance consultancy.
• 4:13
  So we we use economics, finance, and data science to advise companies and policymakers in terms of kind of complex challenges that they may face and how to build kind of strong strategies in the long-term horizon.
• 4:28
  Perfect.
• 4:28
  Thank you, Vanessa.
• 4:29
  And for everybody logging on now, I can see people logging on.
• 4:33
  Drop us a comment.
• 4:34
  Is this your first session that you've joined?
• 4:37
  Second session, where are you logging on from?
• 4:40
  Feel free to ask questions in the comments as we go.
• 4:44
  And hopefully by the end by the time we've got through all the questions, we'll get around to answering some of your questions that you may have.
• 4:52
  So let's get started.
• 4:54
  So, Heather, I'm coming to you first.
• 4:56
  So, from the women that you work with directly, how is AI already showing up in their working lives, even when it isn't labelled as AI?
• 5:06
  Yeah, it's a great question, isn't it?
• 5:08
  I think we don't necessarily know AI is embedded in a lot of things that we're using every day unless we know the nuts and bolts of it behind.
• 5:15
  But the reality is that you know we have apps on our phone that we use for meal planning and for weight management and um health, you know, coaching.
• 5:25
  There's lots of apps on our phone that will use AI and embedded in that.
• 5:28
  And so the likelihood is we're already using it, but not necessarily aware of it.
• 5:32
  A lot of people are using things like Chat GPT to do research, analysis, and sort of craft things, and then also when we're going on websites every day, we're now getting more used to seeing AI agents talking to us and then being present and engaging with us before meeting a person.
• 5:50
  So more and more we're seeing AI within our lives, and I think the reality is you know, how does that then show up in at work for us?
• 6:00
  So certainly we can be finding that we're using more tools at work that embed AI in them, which can make our jobs 10 times faster, and you know, that's why I support a lot of mums to learn is you know, how do you use those tools practically at work?
• 6:17
  So companies might have them, they might not embrace them yet, but they're there.
• 6:21
  And then the other side of AI, which I'm sure um some of the other panelists can lean into, but it's about building AI products and actually being behind and leading that.
• 6:33
  Um, and that's also where we do support mums to relaunch, also.
• 6:36
  So you could be a user of these products or you can be the designer of these products, and I think that's where what's quite exciting right now is it's an open playing field.
• 6:44
  Absolutely.
• 6:46
  And do you think that women are feeling empowered by it or anxious about being left behind?
• 6:53
  I think the problem is that AI is a huge term, right?
• 6:57
  That encompasses so many different things and so many use cases.
• 7:00
  So we're actually talking about AI isn't helpful.
• 7:03
  And personally, I've been in tech and helping women relaunch a career in tech, and I felt overwhelmed by AI, and I only really embraced what did it mean for me last October, and it was October half term, and I needed to get these training manuals done for my team because we had new hires coming in and we needed to refresh everything, and this AI tool popped up to help me recreate training manuals, and it would save me so much time.
• 7:31
  I was like, oh my gosh.
• 7:32
  And so all of a sudden, I was like, okay, I can use AI for this.
• 7:36
  This is a tool I need to learn, and then the penny dropped for me in terms of right, you know, I need to find really practical use cases of AI that are going to help me in my workplace.
• 7:46
  And now I'm using a range of AI tools which I've had to spend time learning, but it's speeding up how I do my job in the workplace, and I think that's when you make it practical, um, and you then lean in and learn those tools, which takes time.
• 8:01
  I mean, I went through six different AI presentation tools until I found the one that I liked, and I was ready to throw my computer out the window, and I'd be like, right, no, be patient.
• 8:10
  I'll try another one, I'm not gonna give up.
• 8:13
  And I found this great one, which we now use in the workplace, and you know, learning those practical tools is what you need to do.
• 8:21
  And so I've taken that back to my super mum's, and we've now got like a boot camp where I teach them these 10 practical tools to help them design presentations, do their research at work, you know, create those project management tasks automatically rather manually, like very practical ways of using AI, and then learning about how agents can be built on what they can do.
• 8:44
  And I think when you get to the point of going, okay, I need a tool, I need to use AI in the context it's gonna help me, it's gonna help me be better at my job, do things quicker, get more time back, then we feel empowered with it.
• 8:58
  But without feeling empowered on the use case, it's gonna be really beneficial to me.
• 9:03
  The whole thing is a little bit like where do I stop and start?
• 9:07
  I've really sort of bubbled it down to be very specific use cases because I've got to help my community understand it and embrace it.
• 9:14
  So that's the approach that I've taken, Claire.
• 9:16
  Amazing, thank you.
• 9:18
  Um, so Lasha, coming to you, um, where does bias most commonly show show up in AI systems, particularly in workplace tools like recruitment, performance, and productivity platforms?
• 9:32
  Yeah, thank you, Claire.
• 9:33
  Um, it's a it's a great question and one that definitely deserves exploring and understanding.
• 9:38
  And I want to start with a moment that has stuck with me.
• 9:41
  Um, a woman I know, a brilliant experience, 40 years in her industry.
• 9:47
  She was screened by an AI recruitment tool, and before a human ever saw her CV, and she was highly qualified.
• 9:57
  And she had two big career breaks, um, breaks in her career.
• 10:01
  One was to fight cancer, one was to care for her mother with dementia.
• 10:06
  And the algorithm didn't like those career breaks in those gaps.
• 10:10
  So it penalized her for that.
• 10:12
  That is bias in AI, and it is everywhere and it's showing up in places that we haven't expected or looked for yet.
• 10:20
  I think it's highly visible in recruitment.
• 10:23
  We see this, um, but also tools are trained based on hiring data.
• 10:27
  Um, they're learning to replicate based on past patterns.
• 10:31
  And those past patterns, they're not neutral.
• 10:35
  They reflect a workforce that shaped has been shaped by structural inequalities.
• 10:40
  And in AI, it's it's not discriminating, it's just trying to optimize.
• 10:46
  So we see that in terms of recruitment bias.
• 10:49
  Um, we also see in performance tools.
• 10:51
  So that's another area, and that's one that I'm watching closely because it's also not as visible.
• 10:56
  So when AI scores output, when it flags underperformance, when it looks for or measures productivity, um, that it's a template that's often based on someone that's always available, always linear working.
• 11:11
  It's not the person that is managing school pickups and presentations to the board.
• 11:17
  It's not the person that took a sideways step in their career to expand their breath.
• 11:23
  Um, so that template is gendered and it's baked into these systems that people are trusting.
• 11:29
  And then also the most common is language bias.
• 11:32
  Um, so much of the data that has trained AI systems has been produced by a very small sliver of humanity.
• 11:41
  And you can ask AI to show me what a powerful leader looks like or a high-performing executive.
• 11:49
  And what you'll see is a very specific image that it reaches for.
• 11:54
  Um, so at We Are Agentic, we see that 70% of AI implementation challenges are people and processes.
• 12:03
  Um, they're not the tools.
• 12:04
  Um, and bias is the most human of those challenges.
• 12:09
  The good part of this is that solution is also human.
• 12:13
  Um, so one of the things that I want to acknowledge is this great organization that's hosting us here today, so investing in women, um investing in women, and they exist precisely because this gap is real.
• 12:25
  They're actively helping women who have had career breaks find their way back into work.
• 12:30
  So flexible job listings, um, career coaching, CV support, this is the human infrastructure that we need that's doing the work that undoes this bias.
• 12:41
  Um, so if you're if you're searching for a career, you're trying to find your way back, I think that is an excellent place to start.
• 12:48
  Um, also, another piece of advice I would provide is before your organization deploys any AI tool in a people context is to ask three questions.
• 13:00
  Um, what data was this trained on?
• 13:03
  Has it been audited for demographic bias?
• 13:06
  And what is the appeals process if someone believes they've been unfairly assessed?
• 13:12
  Um, those are those questions are free.
• 13:14
  They take five minutes, and asking them is an act of leadership, especially in this AI progressing world.
• 13:25
  It's interesting.
• 13:26
  We we see it every day with um candidates applying for our um you know positions and their CVs going through the systems and them being you know rejected for their CV not ticking a certain thing that you know an AI tool is looking for.
• 13:44
  It's huge, isn't it?
• 13:46
  Um so would you say that bias is usually intentional or simply inherited from existing systems?
• 13:53
  Inherited, absolutely.
• 13:55
  AI is built based on the data, and this is historical data in historical systems.
• 14:01
  And I think the more that we see women shaping and developing, training AI systems, being involved in it, we'll see that pivot.
• 14:08
  But it's absolutely inherited.
• 14:10
  AI has no consciousness, it has no emotions, it doesn't lie, it doesn't deceive, it just analyzes the probability of the next answer.
• 14:19
  Um, so that's really what we're up against.
• 14:22
  And to me, my biggest advice and guidance to people is discernment and diligence.
• 14:27
  AI is not going away, but our way of working in a whole part of AI fluency is understanding how to discern the outputs.
• 14:36
  That's still human critical judgment.
• 14:39
  It's absolutely essential in this today and in this future.
• 14:42
  And then doing that due diligence.
• 14:44
  And a lot of things, especially for women where you see caregiving, career breaks, we take them.
• 14:50
  Um, I've taken several, and I think this is a place where we still have to have human judgment in workplaces to support that assessment and not just rely on AI.
• 14:59
  That is my biggest fear is that we give up our decision making to AI because it has inherited bias.
• 15:06
  I've covered three areas.
• 15:07
  There's so many other areas where bias exists within these systems and their outputs.
• 15:12
  So, number one thing is get really diligent and discern what it's showing you.
• 16:08
  Thank you, Lasha.
• 16:09
  So, Andrew, from a leadership perspective, where do organizations most underestimate the impact AI could have on women's roles?
• 16:18
  Well, I think I think AI is brilliant news for women.
• 16:20
  And I and I say that because we've just run a survey ahead of this to find out what you know what the impact is.
• 16:25
  And I'll share some of that right now.
• 16:27
  So I think the first thing is a lot of a lot of women have a disproportionate amount of the of looking after families and looking after the home and perhaps parents and so on.
• 16:36
  And so it's a it's a huge time saver.
• 16:39
  So we're we're actively deploying it in you know in our admin functions, which are historically in most companies where you find the majority of women.
• 16:47
  I think then that brings a danger because obviously there's a danger to the roles that are going to be automated first.
• 16:54
  Having said that, um, I think it's a great catch-up agent.
• 16:58
  One of the things in our survey was that actually women tend to work flexibly, tend to have career breaks, they need accelerators, they're often out of work for a while.
• 17:06
  It's a great accelerator to catch up, um, which I think has been really interesting.
• 17:11
  The other thing I think building what Lasha says, it's just I think I just encourage women to lean in because funny in a funny way.
• 17:17
  My my daughter's 26 and she's a she's a classicist, she she uh speaks a number of languages, she's not a computer science scientist at all, but she's training large language models.
• 17:27
  And I think the uh the people at risk, the the traits which the sort of future holds and and and values are going to be creativity, emotional intelligence, critical reasoning, as well as a general awareness of technology and a want to learn.
• 17:42
  So I think some of these aren't wholly female traits, but they are traits we generally associate with women.
• 17:48
  So I think actually it could really favour women in the workplace, and I think it might lead to a sort of a fairer workplace because it won't be around miles on the clock or direct experience of tech.
• 17:58
  And I think actually we're moving to a position where knowledge of tech is largely irrelevant, it's going to be about critical reasoning, ethics, control of models, running models.
• 18:07
  Um, the thing you just talked about about data discernment.
• 18:10
  And the word discernment I think is a critical one.
• 18:12
  We ran a big conference on defence um at Windsor Castle as part of a sort of a top 13 AI group, and we had a very, very um convincing Norwegian diplomat.
• 18:22
  And what she said was AI, future of AI and defence, for instance, is all about discernment, it's all about critical reasoning, and people being in charge of these models have got good value sets and can discern well.
• 18:32
  So I think um to wrap that up, I think it will have quite an impact on jobs which historically have been done by women, but actually it could be a real accelerator for women in ways that I think people are quite appreciating.
• 18:44
  Absolutely.
• 18:45
  And so, what questions do you think that leaders should be asking right now before rolling out AI tools more widely?
• 18:52
  So the first question we we we we generally get an engagement, and I guess lashes the same.
• 18:57
  We get basically, oh, my CEO says we need AI.
• 19:00
  Can you give us a hand?
• 19:01
  And the first question they say, why do you need AI?
• 19:04
  So it needs to have a meaningful outcome.
• 19:06
  So the first place we go is we we sort of do as we it's it's like any other tech transformation.
• 19:11
  Why do you need it?
• 19:12
  Where do you need it?
• 19:13
  When do you need it?
• 19:14
  Who's gonna do it and how you're gonna you're going to do it?
• 19:16
  And actually, most of our transformations actually start with um actually come through our people practice.
• 19:22
  Although we're we're we're very heavily into engineering and a lot of the manufacturing side areas, um, we often start with a people assessment effectively, because it's really about uh AI literacy, uh uh breaking down barriers, people understanding what it can do and actually looking at where you use it.
• 19:41
  What often happens with us is we go in to do AI and we go, well, you haven't got digital done, your data's not in a good enough state, your systems aren't wired together effectively.
• 19:50
  So we go and end up going back into the core systems of a company like the ERP and CRM.
• 19:55
  And again, I think there's a tremendous opportunity there because we've got quite a good gender mix in our engineering team, and we've got some really brilliant uh ex-Rolls-Royce engineers in the team, and they're getting to share their knowledge of time served with manufacturing with a much younger generation of engineers who are programming the models who will not have had the time experience.
• 20:16
  So, actually, in some respects, it started.
• 20:18
  I think it brings those who I would say over 50 and have done the done the whole engineering degree experience in manufacturing, it's actually starting to blend their skills or expertise with some of the younger engineers we have.
• 20:31
  So I think there's an opportunity here to bring together not just the genders but also across the age spectrum as well.
• 20:37
  Yeah, absolutely.
• 20:39
  Um, Vanessa, so so from an economic and regularity perspective, what are the broader risks of if AI adoption accelerates with that inclusion lens?
• 20:50
  Thanks, Claire.
• 20:51
  Um, so from an economic perspective and and regulatory as well, um, I think the biggest risk is that we would amplify uh the existing problems rather than uh you know solving them.
• 21:03
  I think um Lasha and Andrew as well mentioned, you know, talked about our system being trained on historical data.
• 21:11
  And so um, you know, we we know you know if this data reflects uh structural inequalities in in not just the labor markets, but in other areas like you know, access to credit, uh healthcare, um, you know, and and uh and so on.
• 21:26
  So then and we do know that these uh these issues do exist uh in a lot of aspects in in society.
• 21:33
  And so if AI systems are trained on the data that reflects these um structural inequalities, then the models can encode and scale these patterns even further.
• 21:43
  And so these biases, then you know, um, it could be you know used to be more localized, um, so they can now be more uh systematic and embedded into kind of the decision making at an organization through AI adoption.
• 21:58
  Um, you know, um Lash and others have kind of talked about some examples, but uh you know, also a well-known example uh quite a few years ago now was the recruitment tool developed by Amazon.
• 22:12
  And so it uh was uh allegedly trained on CVs submitted over a 10 year period and largely came from men because you know um it's mainly tech roles.
• 22:23
  Um and so that the systems uh effectively learned that male candidates would be more uh preferable, and then it be kind of started uh to downgrade CVs that included the word woman.
• 22:34
  And so Amazon uh team allegedly uh attempted to correct uh these issues, but then it was still not reliable and the project uh was then abandoned.
• 22:45
  And some, you know, uh there are kind of many other examples, but you know, a more recent.
• 22:50
  One is kind of the researchers from School of London School of Economics and Political Science found that AI tools used by English councils to summarize case notes tend to downplay women's physical and mental health issues.
• 23:08
  And so the concern here would be that actually the language generated from these AI models actually could influence the care assessments and ultimately affect the service outcomes for women.
• 23:21
  And so these examples mean that biases in AI, you know, actually have really big practical and economic implications.
• 23:30
  So it could, you know, could meet a risk of worsening inequality.
• 23:34
  So if certain groups are exposed more to kind of job displacement, they underrepresented in higher productivity, high wage jobs.
• 23:45
  That means that the gains from AI would not be evenly distributed.
• 23:48
  And that's not just a social concern, but it's actually an efficiency concern.
• 23:54
  Because if capable individuals are actually filtered out or undervalued because of bias system, we actually not allocating talent in a most productive way for our society.
• 24:08
  You can also look at more market-based examples and risk there, you know, risk of systematic mispricing.
• 24:16
  So if AI models in in areas like red underwriting or insurance rely on more bias proxies, you know companies could be misestimating risk.
• 24:28
  And that means a distortion to the market and would kind of make markets not working so well.
• 24:34
  And then there's also kind of the trust issue here.
• 24:38
  So I work in highly regulated sectors, so digital markets and financial services.
• 24:44
  And here kind of trust is quite important for adoption.
• 24:49
  And so if consumers perceive that AI-driven decisions are kind of opaque and unfair, you know, people will complain and may not be using the service.
• 25:01
  Regulators may start kind of intervening more aggressively as well.
• 25:05
  And so that's not just kind of um it's not just kind of a value or a social issue, but actually is kind of a condition for sustainable adoption of AI.
• 25:17
  Great, thank you.
• 25:18
  Do you think we'll see more regularity scrutiny around bias and fairness in AI, Vanessa?
• 25:24
  Yeah, um, so that you know, um we we are already seeing um increased uh scrutiny and and uh around bias and fairness.
• 25:34
  So you know, many um existing regulatory frameworks are technology technology neutral.
• 25:40
  So you know, we we have kind of anti-discrimination law, consumer protection law, in financial services, there's kind of financial conduct regulation that already kind of applying to areas around um that would be quite relevant for AI systems.
• 25:57
  Um so regulators uh don't necessarily need um entirely new rule books to intervene where they can see kind of outcomes are being unfair.
• 26:07
  Um what's changing is um the intensity of supervision and expectation around governance.
• 26:14
  Um, so regulators are asking, you know, can companies explain their models?
• 26:18
  Um have they tested for negative impacts or unfair outcomes?
• 26:22
  Uh, is there clear uh accountability?
• 26:25
  Um, there are kind of you know uh differences in approaches.
• 26:29
  So in Europe, there's the EU AI Act um that introduces a risk-based framework with kind of quite explicit obligations for higher risk um systems.
• 26:40
  Um, the UK um kind of has adopted a different approach.
• 26:44
  Uh, so the UK has not uh passed a cross-sector AI statute.
• 27:54
  Um so so in the UK, kind of the system is relying on high-level principles and existing regulation to and existing regulators as well to oversee AI within their remit.
• 28:07
  And so even with this approach, kind of the direction of travel is uh your companies are expected to consider fairness and inclusion um from the outset and not uh retrospectively.
• 28:19
  So they need to kind of monitor consumer outcomes um and and and and kind of address any issues as they uh go along in kind of sewing um their customers.
• 28:30
  Perfect.
• 28:31
  Thank you, Vanessa.
• 28:32
  Heather, coming to you.
• 28:34
  If women engage early and confidently with AI, what would they gain?
• 28:41
  So yeah, I think one of the big benefits is if you're in the job market right now and looking for work, if you've got AI tools on your CV and you've been using them, even if it's at home and applying and using the free trials out there, that and you can communicate that in interview.
• 29:03
  If you can say that you can do your job quicker than somebody that doesn't use AI tools, that's gonna land really well with an employer.
• 29:11
  So if you can design and develop presentations 10 times quicker than somebody that just uses PowerPoint and you're using a tool like Gamma instead, you know, and you can show your efficiency, that's gonna sound you in a much more light better light, I would say, now when you're going for job interviews than somebody that doesn't have those tools or understanding of those tools.
• 29:33
  If you start to learn these tools and you are blown away by what they can do, which certainly I think you know it does naturally happen, you might then become a bit of a leader in AI.
• 29:46
  So you might start helping companies roll out AI tools amongst their teams, for example, and that's one of the roles that we help Super Mums get into.
• 29:55
  The other angle to this is you might get involved in building AI tools and learning how to build agents, and again, we help mums get more technically involved in those angles, and so you've got different paths that you could go down with AI.
• 30:10
  You know, the first is I'll just use it in my normal job role that I do to be more efficient, the second is I could become a leader and roll help companies roll out AI and become an AI consultant, and the third is you know to get really in the grassroots and build AI products and agents.
• 30:26
  So I think there's loads of opportunities.
• 30:27
  LinkedIn ranked AI jobs as the number one um growth, uh growth job role for the last sort of year.
• 30:35
  So it's very much where jobs are.
• 30:38
  The reality is only 20% of those jobs are being filled by women, which is why it's my mission to make sure women are aware of these great career opportunities because they are flexible, they are well paid.
• 30:49
  And it's why I started Super Mums 10 years ago, because I was like, more women need to know about these jobs, you know, don't be scared of them.
• 30:56
  Like it's neat, it's an even playing field right now.
• 30:59
  We can learn these tools, it's new to everybody, and actually, a lot of the AI tools, and certainly helping companies roll out AI tools, you don't need to be technical, you don't need code, and actually, AI's done away with people needing code, um, you know, these things can be built.
• 31:14
  So there's a lot to be gained in the job market from learning AI, really practically like it's a very, very exciting place.
• 31:21
  And I think the reality is if we don't get on board with learning it, you're gonna get left behind because we're getting in that industry revolution right now where everything's evolving and changing.
• 31:31
  And the and the best thing of it all is it's great for us because it saves time for us, it means we get more time getting more stuff done and doing the things that we like to do rather than some of these boring administrative tasks, but also it can give us more time back with the family as well, like in home and work life.
• 31:48
  So that's you know, that's why I do what I do every day.
• 31:51
  I just think there's some great opportunities there to be had.
• 31:55
  Thank you.
• 31:56
  And so, Andrew, from an employer perspective, what do organizations gain when they introduce AI thoughtfully rather than reactively?
• 32:05
  Well, that's what I've got what Heather said.
• 32:06
  So I guess my personal life PhD is how do humans work with technology?
• 32:11
  And and and my job's interesting.
• 32:13
  I start off in shipbuilding and now I'm you know, and I ended up working for one of the leading AI firms in the world.
• 32:17
  And so it just shows the adaptability of a fairly generic skill set.
• 32:21
  Um, I'm not a coder, I I did at university, but I think it's really important what Heather just said.
• 32:27
  I think this is a brand new job market in which skills I mentioned before of critical reasoning, language, communication will really excel and really matter.
• 32:34
  And it's a whole this idea of developing ethnic uh ethical frameworks.
• 32:38
  Moving on to the question of the employer perspective, I think you gain people's buy-in and then it sticks.
• 32:45
  So this is all about augmenting humans.
• 32:47
  We have this we have this phrase where we say um AI won't replace people, but people who use AI will replace those who don't.
• 32:55
  And so the tool sets are designed and set up to be really easily learned.
• 32:59
  In fact, they help you learn.
• 33:00
  So there's nothing to be fear by leaning on stepping in.
• 33:03
  I would just people say how to get started.
• 33:04
  I just say, well, just gotta get something that's properly licensed and dive in.
• 33:09
  Um, and so I think the point is that if you introduce it thoughtfully, you start with people, you start with getting rid of the fears.
• 33:15
  What's the impact of change and what's the very clear use case?
• 33:18
  Because most of the AI projects that fail is because they haven't really got a good reason to do it, although applying it on the wrong things.
• 33:26
  Um, you know, we see it a lot where people are trying to replace junior jobs, whereas we're going, well, actually, this AI is the tacit knowledge of your whole firm of your IP.
• 33:35
  So we'd be making sure that your seniors are uh own it as well.
• 33:38
  So I think from a leadership perspective, uh you can't delegate this, you've got to step in and own it from a leadership level, not just as for NSA said the whole governance framework.
• 33:47
  You know, AI isn't it's an active agent, it's not GDPR, it's not static, it's doing all these things.
• 33:53
  And I always think it's a bit like a a toddler on your tech stack.
• 33:56
  If you don't own it and control it, you know, you'll have bad outcomes, as we heard in the case of the Amazon example.
• 34:03
  But I think what you gain is you're just gonna gain massive productivity.
• 34:06
  And also, and I've got this there's a lovely phrase that comes out of Steve Peters out of the chimp paradox, which is happy teams win.
• 34:14
  And this is all about learning faster than everybody else.
• 34:17
  And we know that happy teams win.
• 34:19
  So if you adopt it carefully and considerately and in a consent-based way, then you're gonna really, really you're gonna your your firms will lift off.
• 34:27
  And I think if you do it that way rather than as a sort of a way to save job, you know, lose lose jobs or whatever, you do it a way to increase productivity and to grow your market share and all these good things, then it will be and and also to give time back to employees because we know we in the Western world work ridiculous hours.
• 34:44
  Um the great benefit of AI is it takes away a lot of the things which humans can do but aren't great fun, allowing us to uh devote our talents to more useful useful things.
• 34:56
  Great, thank you, Lasha.
• 34:59
  For women who aren't technical but who want to stay relevant, what should they be doing right now?
• 35:06
  Yeah, I I want to tell you about a woman who joined our AI ready AI Ready Woman program a few months ago.
• 35:13
  So she came in completely convinced that she was way too far behind and not technical enough and late to the party.
• 35:21
  And by the end of the day, she built her own AI assistant and it was tailored exactly to her role.
• 35:29
  Um, the thing that I will say is it wasn't just upskilling and learning how to work with the technology, it's also the mindset shift, um, the way that her relationship changed and how she works with the technology.
• 35:41
  And I think that's exactly what we're here for today, and that's exactly what I want every single woman to hear.
• 35:47
  You do not have to be a developer.
• 35:49
  Um, you need to learn how to work with AI.
• 35:53
  And fortunately, that is a deeply human skill that you already have the foundations for.
• 36:00
  So I will say every single workshop I've run, every cohort, every team that I've worked with, the people that progressed the fastest, they weren't the most technical.
• 36:10
  What they did is they knew how to think clearly.
• 36:13
  They knew how to articulate a problem precisely.
• 36:17
  Um, they knew the context, the nuances, they knew how to evaluate in their expertise whether that was a good output.
• 36:26
  Um, and this is where you can be exceptional.
• 36:29
  So I will give you three things that I say you can start doing today, right now, um, after this session, um, that will help.
• 36:37
  The first is to start.
• 36:39
  Um, and just like many have shared, super moms, pick Heather, pick a task that drains you and try it with an AI tool.
• 36:48
  You will learn more in 30 minutes of practicing and working with it than three hours of reading about it.
• 36:55
  The second thing is learn to prompt well.
• 36:59
  Um, the output that AI produces is directly determined by how clearly you ask it.
• 37:07
  Um, we use a framework, all of these are acronyms we call ours CGFTE, but they all have similar components to it.
• 37:15
  Ours is you give it context, you give it a goal, you tell it what format you want it to output, what is the tone, and give it some clear examples.
• 37:25
  Um, and what you will see is that this is a foundational piece that of AI fluency, and it will completely determine the difference between a generic output of AI that is left you very unsatisfied, or a genuinely useful output.
• 37:42
  Um the third piece I will say is get strategically curious about your field.
• 37:49
  Um where is AI already automating?
• 37:53
  Where is human judgment irreplaceable?
• 37:56
  And where does your where is your depth most valuable?
• 38:01
  Those are the places where you start building, and that's exactly what our world needs.
• 38:05
  We need more women who are shaping how AI develops.
• 38:09
  We need your perspectives, your experience, the way you think.
• 38:13
  Um, this is exactly what our technology needs more of.
• 38:16
  And I will say to you, you are not behind.
• 38:18
  You are exactly on time.
• 38:20
  I train people every single day who've never even used AI.
• 38:23
  I work with businesses who don't even know where to begin.
• 38:26
  So I will tell you if you start today, you are already ahead.
• 38:31
  Um, and you will get there quickly.
• 38:34
  Wonderful.
• 38:35
  Thank you.
• 38:36
  Vanessa, what safeguards or principles should organizations put in place to ensure that AI doesn't widen existing gender gaps?
• 38:45
  Uh yeah, so so when we talk about safeguards, uh, I think kind of there may be two areas um to kind of focus on.
• 38:52
  So first is on the data um aspect.
• 38:55
  And you know, we've we've talked quite a bit about um kind of the data, the input data that AI systems uh rely on.
• 39:02
  And so organizations um would need to interrogate their data sets, you know, who is represented, who's missing, kind of what historical patterns are embedded in that data.
• 39:14
  And so that means you know, conducting these uh bias and representativeness assessment before you know using this data in the model.
• 39:24
  Um and that's also mean kind of not just looking at the averages, but also you know, looking at the data across different groups, including you know, gender here.
• 39:34
  Um and so when the model kind of performs well um overall on average, um, and but it may kind of underperform in in one group in terms of accuracy, um, you know, that model may not be so meaningful to use.
• 39:49
  And uh in principle, you know, AI models are kind of like any other models.
• 39:54
  And you know, uh in my job, I I've been kind of you know advising organizations on in terms of how to interpret the data, uh, you know, not just looking at averages and looking at kind of the distribution of data across different groups, you know, for a long time.
• 40:10
  And so this is just kind of an extension of that.
• 40:13
  Um, and and people who work on data and models have this saying, uh, gabish in, gabish out.
• 40:19
  And so that means, you know, if the data and the input data is uh of bad quality, so if it uh reflects you know built-in biases, that just means it would be kind of magnified further through AI models.
• 40:31
  Um but another thing I want to emphasize is it's not just testing input data, but also kind of testing uh the outcome data uh of these models.
• 40:42
  Uh, because you know, AI system, we're not talking about you know, one model, one single equation, or you know, a few lines of codes is a very complex system and sometimes it's a system of systems.
• 40:55
  And so um, and and they can kind of rely and bring in combined data from many different sources.
• 41:01
  And so these models is it's quite difficult um to uh to you know predict you know the interactions of these data, these data inputs uh from different sources and and monitor how it um how they interact uh with each other.
• 41:16
  So the best way to monitor and minimize biases is actually to check um the the output uh of these of these um models.
• 41:25
  So you you see potential biases um in the way customers are treated uh at the end of that process.
• 41:32
  So whether that's pricing, hiring decision, or you know, medical treatments, for example.
• 41:38
  Um so that's that's the point on data.
• 41:41
  Um second is um you know governance and accountability.
• 41:45
  And so you know, um the use of AI and how you know the safeguarding of AI should not sit only with the data science team.
• 41:53
  Um so there needs to be accountability at more senior level, you know, risk committee or kind of board oversight.
• 42:01
  Um, and there needs to be kind of um defined ownership for fairness outcomes, so someone responsible.
• 42:09
  And you know, there's also kind of increasing needs for kind of human oversight on kind of many different steps of the AI decisions as well.
• 42:19
  Um, how practical and how feasible it is in the future that is still to be seen.
• 42:24
  But you know, at the moment we see a lot of kind of companies that are still kind of implementing that uh system um uh for now.
• 42:32
  And and you know, we uh in terms of regulatory um perspective, we we increasingly seeing kind of expectation around model documentation, uh auditability, and you know, that human oversight.
• 42:46
  Um but you know, beyond kind of regulatory compliance, um, you know, companies uh should also kind of ask the question, you know, can we explain um this decision?
• 42:55
  Is there kind of a clear route to escalate if you know we we observe that outcomes look a bit skewed in some ways?
• 43:03
  And and you know, is there uh different ways to kind of independently challenge uh what was built in in terms of the model development process?
• 43:12
  And um the last point is kind of in terms of diverse teams and um you know to link in with what we discussed so far is you know, if the people who designing, testing, and approving these systems uh you know have similar backgrounds and perspectives, uh you you can you can expect uh blind spot quite quite easily.
• 43:31
  Um but inclusion uh from kind of the different dimensions in a workforce uh in itself is is a safeguard.
• 43:39
  Thank you.
• 43:40
  I'm conscious that we've got five minutes left.
• 43:42
  If anybody does have any questions, then please pop them in.
• 43:46
  And if we do get time, we will try and get to them at the end.
• 43:50
  Um, I'm just gonna go around each of you now and ask you all the same um question.
• 43:55
  Coming to you first, Heather.
• 43:57
  If organizations get AI wrong for women, what will the long cost be?
• 44:03
  Long-term cost be?
• 44:05
  There's so many different angles to that, isn't there?
• 44:07
  Um, I think in terms of for women, um, you know, you want to have a diverse tech team.
• 44:16
  So a lot of companies approach supermoms, as I'm sure they do with investing in women for you know a range of jobs, but in the tech sector, because there's such a shortage of women, they come to us because they want to make sure they are accessing candidates to diversify their candidate pool.
• 44:34
  So, um, leaning on what certainly Vanessa you were saying earlier, if companies don't have a team that is reflective of you know our general demographic, and you don't have a diverse tech team building your products and implementing them, there is a risk.
• 44:49
  So I think you know, the one thing I've put in the the ring is making sure that if you're building an AI team, a tech team in your company, make sure you've got that diverse pool of candidates coming in because that will reflect how it rolls out and reduce the bias.
• 45:05
  Thank you, Andrew.
• 45:08
  Uh I mean, why would you want to miss out on half the talent pool that's available to you?
• 45:12
  I mean, it to do anything else will be complete madness.
• 45:15
  I think the other thing is that that talent pool is now more available to you because there's a whole bunch of working mothers and people in caring professions who can't may not want you to work part-time, who could now be as efficient as full-time workers.
• 45:26
  Yeah.
• 45:26
  So you don't have to go, uh, you know, you can fill an awful lot of your roles now fractionally, because you it just gets such people's become so much more efficient.
• 45:34
  So, I mean, what's not to like about any of that, really?
• 45:38
  Perfect.
• 45:39
  Lasha?
• 45:41
  Yeah, I'm gonna reinforce what Heather and what Andrew said, but if the people building, auditing, and deploying AI systems, if they're not genuinely diverse, then we will keep encoding the past into the future and we'll keep calling it intelligence.
• 45:59
  And I think that cost is not just talent loss or legal exposure.
• 46:04
  It's that we will build AI that doesn't work for half of the world.
• 46:10
  And I think for organizations and the economy, that's that's going to be the impact.
• 46:15
  So we have to be really conscious and considerate of how we're building these systems, who we're built, who's building it, how we're auditing it, how we're having that discernment and that diligence built into these processes, and it's even who's making the decisions as well.
• 46:29
  Um, so we need to continue to reinforce that.
• 46:32
  That's our accountability as humans, because we are accountable for AI.
• 46:37
  Um, we can't say it was the AI that did it.
• 46:41
  So I think it's really important.
• 46:44
  Yeah, perfect.
• 46:45
  And Vanessa.
• 46:47
  Yeah, so I mean, um, I'm just gonna repeat every what everyone has said.
• 46:51
  Um, yeah, so I mean, you know, huge uh legal, regulatory risk, uh reputational costs if you know people, you know, employee employees and customers uh believe that you know AI systems are skilled and you know adoption would be you know would be slowed down.
• 47:07
  Uh, but also you know, from the economic perspective, it's you know a misallocation of talent and and capital uh when kind of you uh you kind of modeling the market uh incorrectly, inaccurately.
• 47:20
  And so it's not just uh being unfair, but it's you know inefficient uh for society.
• 47:26
  Perfect.
• 47:27
  Thank you.
• 47:27
  I think we might run out of time.
• 47:30
  So thank you so much to the the panel.
• 47:34
  You've been incredible.
• 47:35
  I'm sure any employers, organizations, and also candidates watching have found that really useful.
• 47:41
  Everyone will get a link um to the replay to their inbox, along with the our a link to our survey results that we have just completed, a Women at Work survey, which is where we've uh surveyed women across the UK on the reality of uh the workplace for women right now.
• 48:00
  Um so thank you very much, everyone.
• 48:02
  And um, yeah, thank you for um coming on and sharing your uh expertise and insights.
• 48:11
  Thank you.
• 48:11
  Thank you, everyone, thank you, Claire.
• 48:13
  Thank you, bye.
• 48:15
  Thank you for listening to another episode of the Work It Like a Mum podcast.
• 48:20
  If you enjoyed this episode, please rate, review, and subscribe.
• 48:24
  And don't forget to share the link with a friend.
• 48:27
  If you're on LinkedIn, please send me a connection request at Elizabeth Willet and let me know your thoughts on this week's episode.
• 48:33
  You can also follow my recruitment site, Investing in Women, on LinkedIn, Facebook, and Instagram.
• 48:39
  Until next time, keep on chasing your biggest dreams.