OpenNashPodcasts

American Banker Podcast

What are the ethics of AI?

5 Apr 2018 · 28m · Cathy Bessant, Penny Crosman · 18 quoted facts

▶ Play episodeOriginal source ↗

In brief

Cathy Bessant of Bank of America argues that AI can only be as good as the people, data and processes behind it. Because banks control how money moves, she says they have a special duty to build AI under an ethical framework. That framework covers careful hiring, transparent models and relentless testing, so human bias and bad data don't quietly shape decisions about loans, jobs and crime detection.

The central tension is between AI's promise and its hidden risks. Bessant sees real upside in lending, anti-money laundering, hiring and career development. But she warns that well-meaning people and well-designed algorithms can still combine to produce unfair results, such as slower approvals for harder-to-approve borrowers. She also says leaders must plan workforce retraining so automation doesn't cost entry-level jobs.

Key insights

  1. Machines inherit the limits of their human builders

    Bessant argues that humans program, architect and train machines, so human bias, partial data and untested consequences all carry into AI. She worries the public is led to believe AI is better than people when it is only a subset of what people are.

  2. Transparency in AI models is essential for lending

    She compares today's need for model transparency to the earlier push for transparency in credit scoring. Banks must understand how a model decides so they can explain outcomes to borrowers and regulators.

  3. Fair intentions can still produce unfair outcomes

    Her example is a self-learning approval system that favors easy-to-approve borrowers because they are faster and more likely to accept a deal. Harder cases drift to the back of the queue, even though no one intended bias, which is why she calls for relentless, objective testing.

  4. Better AI should reduce anti-money laundering false positives

    Bessant says false positives distract investigators from catching criminals, so using AI to cut them is worthwhile. She believes banks can explain a lower report count to regulators if they can prove the work catches more real crime.

  5. Leaders must plan for job disruption in advance

    She says that if AI wipes out entry-level jobs, leaders will have only themselves to blame, since such changes are predictable. She points to her own early career, when analysts spread financial statements by hand and are now freed to analyze them, and calls for retraining and work with schools.

Questions

Why does Bank of America's technology chief say banks have a special responsibility with AI?

Cathy Bessant says banks affect how money moves, how it is invested and how companies and jobs are financed. She sees that as a monumental responsibility to get AI right.

Can an AI system be biased even if nobody intends it to be?

Yes. Bessant says bias can come from the people who build and train a system, from incomplete data, or from how a system learns over time. She gives the example of a loan approval queue that gradually pushes tougher cases to the back.

What does it mean that AI is a 'black box' in lending?

A black box is a model whose inner workings are hidden, so no one can see how it reaches a decision. Bessant says banks need transparency so they can explain decisions to borrowers and regulators.

How can AI help catch money laundering?

The host explains that AI can sift through huge numbers of alerts and weed out false positives, which are transactions that look suspicious but aren't fraud. Bessant adds that this frees banks to focus on real criminals and could one day let banks compare data across institutions.

Will AI take away entry-level banking jobs?

Bessant says it shouldn't if business leaders plan ahead, because the changes are predictable. She wants companies to retrain workers, work with schools and design career paths that last.

Facts

Stated by the speakers. Each quote matches the transcript, which shows what was said, not that it is true.

American Banker named Cathy Bessant the most powerful woman in banking, and she oversees more than 100,000 people at Bank of America across 35 countries.

“She was also declared the most powerful woman in banking by American Banker in October. And she also oversees more than 100 ,000 people at Bank of America globally in 35 countries around the world.”

0:00:33 · Penny Crosman · Bank leadership

Cathy Bessant describes artificial intelligence as massive amounts of data combined with massive computing power and a formula that produces an outcome.

“AI is nothing more than massive amounts of data to which massive amounts of compute power are applied with a formula to produce an outcome.”

0:08:07 · Cathy Bessant · AI definition

Cathy Bessant says the banking industry pushed for transparency in credit scoring decades ago and should now demand transparency in AI modeling.

“just as decades ago we pushed for transparency in credit scoring, we have to have transparency in modeling so that we understand how decisions are being made”

0:08:07 · Cathy Bessant · Model transparency

Cathy Bessant warns that lending models using social factors and news as inputs raise questions about whether those inputs have a proven record of predicting credit risk.

“If I have a model that uses social factors, news, other variables, A, what history do we have about their predictiveness, and B, how do we know they're right?”

0:09:46 · Cathy Bessant · Data quality

Cathy Bessant gave an example of a self-learning algorithm that moves easy-to-approve borrowers to the front of a loan queue because speed makes them more likely to accept the deal.

“And if an algorithm over time puts borrowers ahead in the approval queue because they are easy to approve, therefore fast, therefore higher likelihood that the borrower will accept the deal”

0:10:56 · Cathy Bessant · Algorithmic bias

The episode's sponsor, AI Foundry, is described as focused on digital mortgages and helping banks automate lending.

“AI Foundry is focused on making all mortgages digital mortgages and helping banks automate their lending processes.”

0:13:31 · Penny Crosman · Sponsor

Penny Crosman says some banks are testing AI to filter through hundreds of thousands of anti-money laundering alerts, many of which are false positives.

“And I know some banks are testing that, you know, because you could apply AI to filter through these hundreds of thousands of alerts or what are sometimes called false positives”

0:14:02 · Penny Crosman · Anti-money laundering

Cathy Bessant says false positives in anti-money laundering work are a problem because they distract effort from catching criminals.

“So anything that produces a better outcome has to be pursued. And the reason false positives are a problem is they distract activity away from catching bad guys”

0:15:14 · Cathy Bessant · Anti-money laundering

Cathy Bessant says regulators should want the fewest type 2 errors, meaning missed problematic activity, and the most accurate data.

“We want the fewest type 2 errors, which would be missed problematic activity. And we want the most accurate data.”

0:16:37 · Cathy Bessant · Regulation

Cathy Bessant sees an opportunity for AI to let banks view anti-money laundering data from other financial institutions, not just their own.

“Imagine if we weren't looking at our own anti-money laundering data, but there was a way to see what was happening at other financial institutions.”

0:18:31 · Cathy Bessant · Interbank cooperation

Cathy Bessant says she had 12 hours of statistics in college and believes understanding the power and misuse of statistics is important.

“Now, scary for most of my team, I had 12 hours of statistics in college, and I'm a big believer that really understanding the power and the misuse of statistics is important.”

0:19:51 · Cathy Bessant · Statistics

Cathy Bessant joined the bank about 36 years before the interview, when analysts spread corporate financial statements by hand with a pencil on orange paper.

“When I joined the bank, oh, scarily, 36 years ago, we would spread corporate financial statements with a number two pencil on an orange piece of paper”

0:21:40 · Cathy Bessant · Banking history · 1980s

Cathy Bessant says spreading financial statements and calculating ratios, once done by hand, is now an automated data feed.

“Today, that's all done. It's an automated data feed, and it's an automated outcome.”

0:22:20 · Cathy Bessant · Automation

Penny Crosman says branch networks are shrinking and teller jobs are falling because of mobile and online banking.

“we're already seeing to some extent, you know, shrinkage of branch networks, so fewer teller jobs, basically due to the advent of mobile and online banking.”

0:20:24 · Penny Crosman · Banking jobs

Cathy Bessant says women make up 50% of Bank of America's global workforce, 30% of its board and 40% of its global executive management.

“I'm fortunate to be a part of leading a company where 50 % of our workforce globally is women, where 30 % of our board is women, where 40 % of our global management, executive management, is women.”

0:25:16 · Cathy Bessant · Gender diversity

Cathy Bessant says the United States graduates 50% fewer women in STEM fields than it did 30 years earlier.

“Today we graduate 50 % fewer STEM graduates that are women than we did as a country 30 years ago.”

0:25:55 · Cathy Bessant · STEM graduates

Cathy Bessant says Bank of America has focused on diversity and inclusion for 25 years.

“I've been and my company has been focused on diversity and inclusion for 25 years.”

0:26:30 · Cathy Bessant · Diversity

Natasha Lamb has advocated for companies to share their gender diversity and gender pay gap numbers.

“She's been a big advocate of getting companies to share their gender diversity numbers or gender pay gap numbers.”

0:02:02 · Penny Crosman · Pay transparency

Chapters

  1. 0:01:41Why banks have a special duty with AI

    Bessant explains why she calls banking a noble business and why that creates a heavy responsibility for AI.

    “I believe we have a monumental responsibility to get it right.”
  2. 0:03:49Machines can't be better than their makers

    She lays out the core belief that AI reflects human bias, data gaps and training, and calls for an ethical framework built on hiring and testing.

    “machines can only be as good as the processes that train them and the people that train them.”
  3. 0:08:07Transparency and bias in AI lending

    Bessant discusses black-box models, bad inputs and the approval-queue example, where a well-meant system misses its goal.

    “just as decades ago we pushed for transparency in credit scoring, we have to have transparency in modeling”
  4. 0:15:14Using AI against money laundering

    The discussion covers false positives, suspicious activity reports and how to explain fewer reports to regulators.

    “the intent of anti-money laundering and economic sanctions work is to catch bad guys.”
  5. 0:18:31Other promising uses of AI

    Bessant sees opportunity in broader hiring pools, cross-bank data sharing and getting past labels to the underlying facts.

    “Every transaction has a sender and a receiver. And today, for the most part, we can only see one side of that.”
  6. 0:21:40Jobs, retraining and workforce change

    She argues leaders must prepare workers for automation rather than let entry-level jobs vanish.

    “if that happens, shame on us.”
  7. 0:25:55Women, careers and diversity in tech

    The conversation turns to using AI for career pathing, and to the gaps that remain for women in technology and senior roles.

    “even in the tech disciplines, particularly, I would say we have not, I'm sorry for the pun, but we have not cracked the code.”

Transcript

0:00:00
Penny Crosman

Welcome to our podcast. There are many ethical questions being asked around artificial intelligence these days as more and more companies adopt the technology within their company. Can machines that lack a conscience really be trusted to make moral decisions? To what extent can machines do the work of humans? If something is happening inside a black box, should that be used to make decisions that affect human beings? Today we have with us Cathy Bessant, who is Chief Operations and Technology Officer of Bank of America.

0:00:33

She was also declared the most powerful woman in banking by American Banker in October. And she also oversees more than 100 ,000 people at Bank of America globally in 35 countries around the world. So we're very happy to have you here, Cathy. Thank you for coming. Thanks, Penny. I've heard you say that banking is a noble business. Do you think that banks have a special responsibility to try to use artificial intelligence responsibly with regard to their employees and their customers?

0:01:06
Cathy Bessant

I do believe banks have a special responsibility. I actually think we've had special responsibilities since the sands of time. If you even look back to how this country was founded or other markets around the world, how they were founded, how money moves and how people are empowered by money is a very important part of a political structure, a social structure, equality. So I talk about nobility because I believe it's that fundamental to what it means to be a bank.

0:01:41

Now, your question is about AI, and I think because we affect the money, whether it's the movement of money or the investment and the return of money, or it's how capital markets work for companies and jobs, I believe we have a monumental responsibility to get it right.

0:02:02
Penny Crosman

And I think you have also mentioned that we don't want to allow human biases to creep into artificial intelligence. I recently happened to speak with Natasha Lamb about that. She's been a big advocate of getting companies to share their gender diversity numbers or gender pay gap numbers. And one of the things that she's concerned about is if you don't have coders who are women and minorities, then you could have the biases of white men creeping into your algorithms.

0:02:40

But, I mean, that's one aspect of it. I'm sure there are many aspects of the biases we all have. And how do you think we can prevent our human foibles and biases from, you know, creeping into our software? Sure.

0:02:58
Cathy Bessant

Well, I think it starts with a fundamental belief that I have, that our firm has, that machines cannot, by definition, be better than humans. They can't. They are programmed by humans. They are architected by humans. And so every element of how a machine works, regardless if after programming they can work on their own, every element of how a machine works is driven by a human being. And so every risk that exists in the marketplace or in the world around human bias, intentional or unintentional, information, do we have full information and data when we're creating models or partial, our analysis and testing of unintended or intended consequences.

0:03:49

machines can only be as good as the processes that train them and the people that train them. So it is a very big issue. And sometimes I worry that the global public at large and leaders, and in some cases, companies that build algorithms and artificial intelligence, make us believe that it's better than we are. And in fact, it's a subset of who we are.

0:04:17
Penny Crosman

That's a good point. So maybe some of it is hiring the best and brightest people who can, you know, use logic and maybe some of their own ethical guidelines to, I mean, do you see creating sort of an ethical framework for AI software development where you try to build basic principles into software, whether it's being used for marketing or lending or risk or any other area of the bank?

0:04:52
Cathy Bessant

I think we have to have an ethical framework, and it isn't enough for just one sector, corporates or political systems or public sector organizations, not enough for any one sector to have that framework. It has to be a universal that as a globe, we will not accept the concepts around artificial intelligence without an ethical framework underpinning it. You know, I think it starts with hiring. Fundamentally, judgment and bias are conduct issues. issues. And as you know, whether it's the securities industry or anything else, huge focus on conduct.

0:05:42

And underpinning all of that has to be we hire the right people. Second, that we train them in our vision and values. So, you know, when we're hiring, making sure there's diversity, making sure that that diversity is also geographic. It is every element of diversity in addition to some of the really tough spots like in the technology space race and gender. It is then training and then everything about how the development process works. But fundamentally, I'm a big believer, Penny, in the importance of testing.

0:06:19

We can't leave outcomes to chance. We're talking about can someone save for their retirement, right? Does that algorithm produce a result that allows them to live in the way that they hope to that they've worked their whole lives to be able to. So the stakes are super high, and testing relentlessly is a very important part of quality assurance and quality control. Not to sound like an engineer, but testing and quality control are a big part of this.

0:06:53
Penny Crosman

Well, that's a good point. I know some of the companies that are using AI in their lending are using the CFPB's test for disparate impact as they look at all the outcomes. And the CFPB is watching that too in the case of Upstart. And that gives a little bit of comfort that maybe these lending decisions are being made fairly. A lot of banks are leery of using artificial intelligence in lending. I think, A, some of them are not as interested in banking people who have a score, a FICO score of under 750, which is what AI really is good at, figuring out who's credit worthy, even if they don't have a high credit score.

0:07:37

But also there's the concern that, you know, if you're using a black box to make lending decisions, could it make these correlations that are unfair to a segment of the population? You know, could the software start realizing that people in this zip code, for instance, are more credit worthy than people in that zip code? And I mean, that's just a random example. But is that something you worry about? And do you think that could be overcome, that concern?

0:08:07
Cathy Bessant

Well, as a company and as a leader, we and I worry about that significantly. Just a couple of thoughts that your question spurs to mind. One is models, in order to be effective, have to be transparent. AI is nothing more than massive amounts of data to which massive amounts of compute power are applied with a formula to produce an outcome. And what's in that formula, the black box to your term, is incredibly important. and just as decades ago we pushed for transparency in credit scoring, we have to have transparency in modeling so that we understand how decisions are being made and, by the way, not unimportantly so that we can explain them to borrowers or potential borrowers and to the

0:09:00

regulators and other stakeholders. So I think at the heart of this is transparency. as a company when when we're in and as a leader when we're in the business to facilitate commerce and capital and empowering people with their money we want to make every good loan that's out there the there's no question about our goal to get more credit into the marketplace especially in the banking business where we're held to such high capital standards and where we're able to attract so many deposits, how we deploy that capital and deposits is not only good for customers, but really important to shareholders.

0:09:46

So very important to make all the loans and extend all the credit we can. The only question is, how does AI factor into that? And my worry always here is the garbage in, garbage out worry. If I have a model that uses social factors, news, other variables, A, what history do we have about their predictiveness, and B, how do we know they're right? So I think the efficacy of the work has to start with the efficacy of the inputs.

0:10:23

And I think I want I believe we should leverage artificial intelligence as much as we can to expand the range of creditworthiness. But I also worry a lot about the quality of inputs. So the responsibility at every point in the chain is huge. If I could for just a second, there was one other really important point that you made, which is an algorithm that starts out to be great but then learns from itself. And people call that machine learning.

0:10:56

I saw the other day a very great example of the risk here. So say we're underwriting two different clients and we have a premium because we know customers want to work with banks who are quick to give a response or even fintech companies. often what they market is speed. And if an algorithm over time puts borrowers ahead in the approval queue because they are easy to approve, therefore fast, therefore higher likelihood that the borrower will accept the deal, and the tougher to bank citizens take a little bit longer to approve, you have a model that's well intended and maybe completely transparent, but over time misses the target of speed to answer.

0:11:51

And so these are, whether it's easy for your listeners to

0:11:56
Penny Crosman

see that and visualize that, these are real world issues. Absolutely. And I think those are really good points that I wasn't even thinking of, that A, the pulling in disparate data sources could also bring bias in because there are people who don't use Facebook. There are people who don't use Twitter. You know, there are people who aren't interested in some of these channels that are feeding data. And also the people being put to the back of the queue, as you described, are likely to be the people in more disadvantaged socioeconomic strata.

0:12:35

So again, there could be a form of disparate impact, even if it's not technically disparate impact. You know, it could have a negative impact. So those are really good points.

0:12:46
Cathy Bessant

And it's not that the intent is disparate impact or that there's even, in that case, human bias, but the complicated system of reward and recognition for employees, along with what it takes to approve a more complicated transaction, that's a situation where the net sum of multiple well-intended people and machines and well-designed algorithms can in fact produce the wrong result, which is, again, why I come back to just relentless and unbiased, hard-cold, objective testing.

0:13:31
Penny Crosman

Now let's take a quick break to recognize our podcast sponsor. AI Foundry is focused on making all mortgages digital mortgages and helping banks automate their lending processes. The goal is to decrease the time it takes to fund mortgages, increase customer loyalty and retention, and of course, increase enterprise profitability. With so many banks aiming to make their mortgages digital, AI Foundry is a partner worth looking at. So another area that I know banks are thinking about AI is anti-money laundering.

0:14:02

And I know some banks are testing that, you know, because you could apply AI to filter through these hundreds of thousands of alerts or what are sometimes called false positives, things that look like they might be fraud or things that trigger a transaction that's over $1 million, for instance, but that isn't necessarily fraud. And the AI can help do some of that research to figure out what is really fraud and what isn't. But I've heard that banks are worried about reducing the number of suspicious activity reports that they turn in.

0:14:39

So then if this bank using AI is now submitting 5 ,000 reports instead of 10 ,000, but their peers are still submitting 10 ,000 SAR reports a year, that the regulators could look at that and say, oh, you're not making enough effort. Do you have an opinion on that? Do you think that it's worth it, even if it does require some explanation to the regulators that, you know, look, we're narrowing this down and we're giving you more accurate information?

0:15:13

Well, you know me, Penny.

0:15:14
Cathy Bessant

I always have an opinion. So a couple of thoughts. One is the intent of anti-money laundering and economic sanctions work is to catch bad guys. So anything that produces a better outcome has to be pursued. And the reason false positives are a problem is they distract activity away from catching bad guys, just to put it in the most straightforward context. So we have to get better at using data and modeling and artificial intelligence to do that.

0:15:50

There is no question that our regulators do compare activity. And I'm not so sure they always make a judgment about whether more is better or fewer is better, but they do look at it. I have always found that getting to the right outcome, which is catching the most problematic action and failing to miss, not missing action, can be explained effectively to regulators. So if we stay focused on the objective, which is crime detection and prevention, and can prove the efficacy of what we're doing, that argument typically wins.

0:16:37

And I think it's important to get past sometimes the easily popular answer to why a company might not want to invest in artificial intelligence. In my mind, the regulators should be beating the doors down saying, look, we want fewest false positives. We want the fewest type 2 errors, which would be missed problematic activity. And we want the most accurate data. So you should, you bank, you banks, plural, should be using every tool at your disposal. But that takes money and investment and talent like we've been talking about.

0:17:17

So I always try to peel back the onion and say, is the objection really the objection? Because regulators are, if not as zealous as we are about catching bad guys, more zealous and more worried about the risk of missing something because the political pressures around that are too high. So very important to pursue it because information can only be good in this regard.

0:17:46
Penny Crosman

Are there any other areas where you see a lot of promise in AI where it could really help you and your company?

0:17:53
Cathy Bessant

I do. I think that it can, done well, it has the opportunity to take or to ingest objectivity into the hiring process. to so lots of studies that would say blind hiring to demographic factors produces a more diverse outcome. I'm not so sure that I think there's evidence on both sides but I do think that the ability to widen the pool of applicants to work through applicants in a more systemic straightforward way holds real opportunity to increase the pool of talent that's diverse that we look at.

0:18:31

So hiring is a big one. I do think that interbank cooperation, really big opportunity. Imagine if we weren't looking at our own anti-money laundering data, but there was a way to see what was happening at other financial institutions. But that is a massive amount of data and requires a massive amount of analytical capability, but AI provides that opportunity. Every transaction has a sender and a receiver. And today, for the most part, we can only see one side of that.

0:19:10

There's big opportunity, I think, there. I do think any political situation that involves or where the balance has to be politicization versus objectivity is important. So do we care that it's global warming, called global warming? No, but we do have to, I think, to effectively use resources, to think about financing techniques, to think about trade even. We have to understand the facts. And I think AI gives us an opportunity in almost every sense to go past the label and onto the data.

0:19:51

Now, scary for most of my team, I had 12 hours of statistics in college, and I'm a big believer that really understanding the power and the misuse of statistics is important. So the flip side of that is the use of facts and data has to be put into a mix where we, again where we get past the headline five times more likely to and figure out what is it the

0:20:24
Penny Crosman

underlying data is really saying. Another thing that people worry about is AI taking over all our jobs. And, you know, it does seem like in the initial phases, it's going to be the entry level jobs that are the most affected. I mean, we're already seeing to some extent, you know, shrinkage of branch networks, so fewer teller jobs, basically due to the advent of mobile and online banking. But I think you're starting to see it in the back office jobs, where robotics process automation is doing some of the very simple work.

0:21:01

You're starting to see that in several other areas in a company where the most basic tasks can be done theoretically by an AI machine. And, you know, there are people that worry, you know, there are people who feel like banks have always been a good source of entry-level jobs. So, you know, are we going to lose that? And can that be regained in some way? Do you worry about that, about the jobs that, you know, will inevitably be eliminated over time?

0:21:35

and do you think that enough new jobs may be created along the way?

0:21:40
Cathy Bessant

My perspective here is not easy to execute, but pretty simple, I think, to think through, which is, if that happens, shame on us. Shame on us as global corporate leaders or even leaders of small companies for allowing that to happen. It would be the ultimate sin, right, to not learn from the lessons of the past. When I joined the bank, oh, scarily, 36 years ago, we would spread corporate financial statements with a number two pencil on an orange piece of paper and calculate current ratios and debt-to-equity ratios.

0:22:20

Today, that's all done. It's an automated data feed, and it's an automated outcome. But the speed that we've gotten from automating that means that there's more time to find creative solutions for customers, to actually bank more customers overall. Because today I would be, if I was in that job, analyzing the result versus producing the spreadsheet. But go back to the shame on us part. These changes are predictable. And it is accountable to us as a corporate sector and as employers overall to be thinking about the impact of technology and to be preparing and planning and transforming the workforce to be ready for that.

0:23:11

So, you know, the opportunity to retrain our people is not new to us in financial services, and we have to do it, and we will and are. So retraining a big portion of it, working with educational systems to think differently about the kind of entry-level candidates that we're getting and producing is extremely important. The intersection between commerce and education has never probably been more important. And then really thinking now about the skills that are fit for the future and designing career paths once we hire someone to be able to withstand the test of time, these things all sound theoretical, I think.

0:24:00

I think they sound strategic, but they are really practical and can, in fact, be accomplished. It's why I talk a lot about responsible AI. Our company is underpinned by this notion of responsible growth. AI has the potential to fuel growth has to be done responsibly, including workforce transformation.

0:24:23
Penny Crosman

And just to pick up on that point about the career path, you know, there's been a lot of public discussion lately about Me Too and fairness of women in the workplace and the gender pay gap and the rate at which men and women are promoted in corporations. and a big part of a raw pay gap is the fact that there are more men at the top than women. There are a lot of women in the entry level or in the middle layers, but at the top levels of most of the big banks, there are not as many women.

0:24:59

Bank of America probably has more than most. But do you think AI could be used in that development of a career path to help promote more women through the ranks so that more women do make it to those upper tiers?

0:25:16
Cathy Bessant

I think AI offers us the opportunity to understand, predictively, career pathing that tends to result in positions of greater leadership. It should also, by the way, show us people who, through the course of a career development process, are at risk for choosing to leave or being left behind by a promotional structure. You said it. I'm fortunate to be a part of leading a company where 50 % of our workforce globally is women, where 30 % of our board is women, where 40 % of our global management, executive management, is women.

0:25:55

But even in the tech disciplines, particularly, I would say we have not, I'm sorry for the pun, but we have not cracked the code. Starts with graduation rates. Today we graduate 50 % fewer STEM graduates that are women than we did as a country 30 years ago. Biggest tech boom ever, half the proportion of graduates. Okay, problem A. Problem B is when we, and we have great success at this, in fact, our recruiting class out of this year's college graduates, roughly 50 % women.

0:26:30

something like 40 and change women. And yet, as we go through the ranks, one of my direct reports is a woman. So these issues that you're citing are real. And so career development, tracking, using data to understand how a career path could work or who's at risk. And frankly, simply hard work. I've been and my company has been focused on diversity and inclusion for 25 years. These are hardworking, smart people. I like to think I'm a hardworking, smart person, and I haven't cracked the code.

0:27:14

So staying focused on outcome, redoubling the effort, using data differently. and regardless of anything else, the discussion that we're having now compared to a year ago is healthy, appropriate discussion. Anything that contributes to a work environment that is better from a diversity and inclusion perspective is to the good. Sure.

0:27:40
Penny Crosman

Cathy Bessant, thank you so much for coming in today.

0:27:43
Cathy Bessant

Penny, thanks for having me. Great, important topics. If you enjoyed this podcast, you should subscribe to AmericanBanker.com for analysis and insight on the financial services industry that you can't find anywhere else. And just for AB Podcast listeners, we're offering a deep discount. Go to AmericanBanker.com slash AB Podcast for your exclusive discount code. That's AmericanBanker.com slash AB Podcast.