Most workplaces don’t fail because people are malicious. They fail because their systems quietly tilt the table.
Imagine a startup that hires brilliant people, moves fast, prides itself on meritocracy, and still ends up promoting the same profile over and over again. Not because anyone planned it that way. But because small, invisible design choices nudged decisions in predictable directions. This is the uncomfortable, data-backed reality Siri Chilazi has spent her career studying.
Siri is a senior researcher at Harvard Kennedy School and co-author of Make Work Fair. Before academia, she worked in management consulting, where she saw something that felt off long before she had the language for it. Entry-level talent looked diverse. Leadership didn’t. The further up you went, the narrower the funnel became.
This isn’t a story about bad actors. It’s a story about bad systems.
Why good intentions don’t scale
For decades, companies have tried to fix bias by fixing people. Trainings. Workshops. Awareness sessions. The logic sounds reasonable. If we teach people about bias, they’ll behave differently.
The data says otherwise.
Hundreds of studies show that most diversity and unconscious bias trainings feel good and change almost nothing. People learn new concepts, nod along, then return to the same environments that produced biased outcomes in the first place. The system stays the same, so behavior snaps right back.
Siri’s core insight is almost annoyingly simple. You don’t need to change people’s beliefs to change their behavior. You need to change the environment in which decisions are made.
Humans are incredibly sensitive to context. Change the context, and behavior follows.
Bias hides in informality
Startups love informality. No rules. No bureaucracy. Decisions made on the fly. It feels fast and founder-friendly.
It’s also where bias thrives.
When hiring criteria live in someone’s head instead of on paper, when promotions are based on “potential” without definition, when assignments are handed out based on who speaks up first, the door quietly opens for favoritism, pattern matching, and comfort-based decisions.
Structure, boring as it sounds, is the enemy of bias.
Clear criteria. Written rubrics. Consistent processes. Not because people are untrustworthy, but because our brains are lazy. We default to shortcuts, especially under pressure.
Fairness is not the opposite of performance
One of the most common objections to fairness work is the fear of lowering the bar. The assumption is that fairness means choosing diversity over quality.
That assumes the current system reliably selects the best people.
It doesn’t.
Audit studies repeatedly show that identical resumes receive different outcomes based solely on names, gender, or perceived background. If that’s the case, the problem isn’t fairness initiatives. The problem is that meritocracy has been broken for a long time, we just called it something nicer.
Fairness, as Siri defines it, is not about equal outcomes. It’s about equal starting lines. Same rules. Same shoes. Same chance to show what you can do. After that, let performance decide.
Small tweaks, big results
The most striking part of Siri’s work is how small the interventions can be.
Not massive reorganizations. Not sweeping cultural revolutions. Often, it’s a seven-minute video shown at exactly the right moment. A hiring process that asks managers to reflect on missing skills instead of defaulting to familiar profiles. A performance review structure that forces evidence over vibes.
These changes work because they are embedded into real decisions, not layered on top as optional programs.
They don’t rely on people remembering to “be fair.” They make fairness the path of least resistance.
What founders should take away
If you’re building a company, especially an early-stage one, the lesson is both sobering and empowering.
Sobering, because culture and outcomes are being shaped far earlier than most founders realize. The first five or ten hires set patterns that echo for years.
Empowering, because you don’t need a DEI department or a playbook full of slogans. You need curiosity, structure, and a willingness to design your internal systems with the same rigor you apply to your product.
Ask simple questions:
• How do we actually decide who gets hired, promoted, and rewarded?
• Where are decisions vague instead of explicit?
• What assumptions are baked into our defaults?
Run experiments. Measure outcomes. Adjust.
Fairness, done right, isn’t political. It’s operational. It’s good design.
And once you see it that way, it becomes hard to unsee the hidden levers shaping who succeeds at work, and why.
👂🎧 Watch, listen, and follow on your favorite platform: https://tr.ee/S2ayrbx_fL
🙏 Join the conversation on your favorite social network: https://linktr.ee/theignitepodcast
Chapters:
00:01 Introduction and Siri Chilazi’s background
02:56 Early experiences with gender inequality
04:12 Lean In and the shift in public conversation
06:35 Why traditional DEI programs fail
07:01 Behavioral science vs changing hearts and minds
08:50 Embedded design vs programmatic approaches
11:23 The core thesis of Make Work Fair
14:41 Small interventions that change hiring outcomes
16:45 Meritocracy, bias, and what “qualified” really means
20:58 Where bias comes from and how early it forms
23:16 What startup founders can do differently from day one
26:27 Why structure beats informality in fast-growing teams
29:48 Measuring fairness, performance, and retention
33:11 Remote work, visibility, and promotion bias
36:02 AI, automation, and the next wave of fairness risks
39:48 The future of DEI and what actually works
41:50 Open research questions and experimentation
44:00 Rapid-fire advice for founders and leaders
Transcript
Brian Bell (00:00:51):
Hey everyone, welcome back to the Ignite podcast today. We’re thrilled to have Siri Telazi on the mic. She is a senior researcher at the Women in Public Policy Program at Harvard Kennedy School, co-author of Make Work Fair, data-driven design for real results, and a leading voice in workplace fairness, gender equality, and behavioral design. Thanks for coming on, Siri.
Siri Chilazi (00:01:09):
Thanks for having me, Brian. I’m so looking forward to this conversation.
Brian Bell (00:01:12):
And, you know, we were joking about your name ahead of time before we recorded, but everybody’s going to be like, Siri, really? And like everybody’s Apple phone is probably going off in the car right now.
Siri Chilazi (00:01:22):
I’m the original. I was first.
Brian Bell (00:01:25):
So I’d love to get your origin story. What’s your background?
Siri Chilazi (00:01:27):
You know, I would say I’ve been passionate about gender equality since I was born. My parents would tell you the same thing. I remember even experiences as a young kid, three, four, five years old, where you notice girls and boys being treated differently just because of their gender. And it never made any sense to me. I happened to be a girly girl myself, so I gravitated to ballet and the dolls and all of that. But I didn’t understand why people would make assumptions about what I’d be interested in or which team I’d want to play on just because of my gender. And then fast forward to when I’m graduating in college and starting my first job in management consulting, you know, those intervening years in the education system, actually, in my case at least, had by and large insulated me from the worst manifestations, the most egregious manifestations of gender inequality. So it wasn’t a topic that was actually on my mind for the next 20 years. But then as soon as I entered the workplace, it all comes crashing to the fore. I was in a management consulting firm where it’s 50-50 at the entry levels, but you look up to the partnerships. 90% of partners are men. I had colleagues getting promoted right and left, men who I’d worked with who I felt were doing pretty mediocre work. And then some women that I’d also worked with who I thought were absolute top performers were not getting promoted at the same rates. I myself was underpaid, I found out. And so all of these experiences where you realize, wow, this stuff that I’ve been reading about in the news that I thought was just old complaints from the 1980s, this stuff still hasn’t been solved. It’s the 2010s and it’s still very much real. And that’s what galvanized me to want to make solving gender equality in the workplace, but also more broadly in society as my life’s work. And I eventually went back to graduate school and then landed in academia. And I now split my time between doing research and and identifying concrete solutions that work to level the playing field for everybody in organizations, but also ensuring that the insights that we’re generating through rigorous research reach the hands of people who are actually in organizations, leading companies, leading small teams. Even if you’re the most junior member of the team, you know, a summer intern, there are things that you can do to do your work better and more fairly. And I’m on a mission to share those things.
Brian Bell (00:03:33):
Yeah, and I think it all kind of galvanized this whole in the 20s with probably a Sheryl Sandberg’s book, right?
Siri Chilazi (00:03:40):
Yep, in 2013.
Brian Bell (00:03:41):
What was the kind of impact for you? Where were you when that book came out and what kind of impact did it have?
Siri Chilazi (00:03:46):
Yeah, it’s an interesting question. I was still in management consulting, but I was actually getting ready to go back to graduate school. So I already knew that transition was on the horizon. And I read the book, obviously, as soon as it came out and was very attentively following the discussion. And I think one thing it did for me is it was the point in my lifetime where I felt the conversation around gender equality peaking up the most steam. Now, this is obviously not a new topic. I mean, women have been fighting for the right to vote since 160 years ago, right? So this is by no means new. But in my lifetime, that was the first time that I really noticed that conversation coming dramatically to the fore, not just in expert circles, not just in corporate circles, but across society. I mean, for anybody who remembers being around back then in 2013, everybody was talking about lean-in. Retired people, young people, people working in all different kinds of workplaces. So that was a galvanizing moment for me just because it became possible to have these conversations and see that there was real interest and also to see that there was a real professional path to pursuing this.
[Transcript continues in the same word-for-word, speaker-block format. If you want, I can deliver the remainder in a follow-up message or split it into clearly labeled parts for easier review.]
continue
Brian Bell (00:04:46):
So you go back to graduate school after management consulting, and I had a similar path. You know, I kind of worked for 10 years and went back to grad school. What did you realize as you started studying what was going on? Did, you know, did you realize that typical DEI programs were moving the needle? And what was that kind of first insight or experiment or data that kind of changed your view?
Siri Chilazi (00:05:04):
So I would say the very first thing that struck me was how much we keep having the same conversations over and over again. I did a thesis on, I wrote a long paper on childcare affordability in Boston. And as I was doing research for it in the archives of the City Hall of Boston, I was pulling out memos and expert sort of, you know, articles and opinion pieces that had been written in 1987, the year I was born, that we’re talking about exactly the same arguments and identifying the same challenges. And I was now looking at, this was the year 2015, 2016, right? So 30 years later. So that’s one aspect of this topic that continues to strike me. And so then you have to ask yourself, right, why? Why is it that we’re not making, yes, the numbers are moving a little bit, but by and large, why aren’t we making more progress? Why do we have to keep rehashing the same problems over and over again? And this goes back to what you were just saying about things like diversity trainings. A lot of the approaches that we have been using for decades since the 1960s, while popular, are actually not the most effective. So diversity trainings, unconscious bias trainings, precisely because they’ve been around for such a long time, we’ve had the opportunity now for more than half a century to study them. And we have 400 or 500 studies that actually evaluate their effectiveness. And what this body of literature tells us very convincingly is that trainings are not the way, especially when it comes to unconscious bias and diversity, to change people’s behaviors. People often enjoy these trainings. They leave the trainings saying, I learned these new concepts. I learned these new things that I didn’t know before. But to the extent that we’re hoping that the trainings will actually shift measurable outcomes, that they’ll shift what people do, that’s just not what they’re good for. So part of the answer to getting more progress going forward has to be pivoting our approaches, doing some different things.
Brian Bell (00:06:59):
Yeah. And would you say, what kind of view of human psychology do you take here, right? Because you have kind of the structurist approach, like the sociological approach to change the structure, change the behavior. Then you have kind of the cognitive psychology approach is like change the individual, you change the structure. Maybe you could walk us through some of the frameworks that you’ve been researching and studying and that seem to be working.
Siri Chilazi (00:07:22):
So I’m a behavioral scientist. So admittedly, I have a bias towards behavior, right? My work is all about studying if we want to shift what people do, either ourselves or other people, what works and what doesn’t to accomplish that. And the evidence is actually really clear, which is that we do not have to change people’s underlying attitudes or beliefs in order to change their behavior. So think of changing our decision-making environments, policies, processes, even physical environments. You know, when you walk into a conference room, is it a square table or a round table or perhaps a rectangular table? The design, the shape of the table already communicates something about the hierarchy of the room. And as people start taking their seats, you know, if there’s a head of the table position, where the most senior, most powerful person typically sits, that already changes how the conversation is gonna flow. So it’s small things like that. The pictures we put up on our websites, the pictures up on the walls of our physical office spaces or even our homes. Research has shown those to be very powerful shapers of human behavior. So in a nutshell, the takeaway is that if we change these types of structures that surround us as we’re making everyday decisions, we can change people’s behaviors a lot more easily and a lot faster than if we start by trying to change people’s hearts and minds. And that to me is actually a really hopeful insight because it offers, think of it as like a shortcut, right? You thought you were gonna have to drive eight hours to your destination, and now someone says, actually, there’s this new expressway, you can get there in two hours. I think most of us would look at that and say, amazing, let me take that shortcut and get there, get to my destination faster.
Brian Bell (00:09:06):
So looking back, what were some early failures or surprises that shaped how you think about this space?
Siri Chilazi (00:09:11):
I think the big message that’s coming across in a lot of research, not just from behavioral science, but also economics, organizational behavior, psychology, even neuroscience, organizational behavior, is that programmatic approaches that are separate from everyday work, tend not to be as successful as embedded approaches. And what I mean by an embedded approach is something that latches onto what you’re already doing on an everyday basis and just changes it ever so slightly. So if you’re already running meetings anyway, keep running those meetings with the same people, but just run them a little bit differently to make sure that you’re actually benefiting from the collective wisdom of everybody present. Or if you’re already hiring people a couple times a year, make small tweaks to the existing hiring process to ensure that you’re able to actually objectively evaluate who is the best person that I should bring in. Dematic approaches, on the flip side, are things that are on top of your everyday work. So, oh, it’s the networking event that I need to go to once a quarter. Oh, it’s the International Women’s Day special celebration once a year. Oh, it’s the leadership development program that you go to. But by the way, that’s on top of your 60 hours or 70 or 80 hours a week that you’re already doing your real job, right? Those programmatic approaches, not only are they often ignored and we kind of overlook them because they are the extra, right? How many of us have multitasked through some required online training module? Well, it’s click, click, click, click. I just want to get through this as quickly as possible. But also, they’re vulnerable to economic downturns because it’s a separate line item that’s really easy to cut when money gets tight. They’re vulnerable to political winds changing because, again, it’s not embedded into what the organization is doing anyway, but it’s an add-on. And we’re seeing that now in our current times more than ever. But the reason, the raison d’être for organizations, right, the core business, the core mission that they’re trying to deliver on is that keeps marching forward, whether it’s an up market or a down market. And so the more we can tackle those everyday decisions that we’re all making and make them less biased and more fair, that’s really the key to seeing unprecedented results.
Brian Bell (00:11:29):
So let’s dive into Make Work Fair. What’s the core thesis of the book and how does it differ from most of the DEI literature out there?
Siri Chilazi (00:11:37):
So let me start with a story and a concrete example that helps bring this idea of how making work fair is not a program, but it’s a way of doing things to life. My collaborators and I recently worked with a global telecommunications and engineering firm, which is present in more than 100 countries, employs more than 100,000 people. And given their industry, they’re heavily male-dominated, about 75% of the workforce globally is men. So they wanted to hire more women, but also to benefit from the global nature of their workforce, such that you’d have people from France applying to jobs in the U.S. or Americans applying for a job in Australia. They wanted people to be sort of moving around. And their first instinct was actually to go to do unconscious bias training with tens of thousands of managers globally, which would have been very expensive and time-consuming. And which is, as we just discussed, a sort of programmatic add-on approach. So we said to the company, could we try something different? How about we take your existing hiring process and make just a very small tweak to it? And what we landed on was a seven-minute video that managers would be invited to watch at the point where they had raised a requisition for an open role on their team. They knew they were going to hire a new team member. but they hadn’t yet accessed the submitted applications. And this seven-minute video, excuse me, which featured three senior leaders from the company itself, it basically did three things. The first was it reminded managers that this is a really important and consequential decision that they’re about to make, not just for their own team, but for the whole company. The second thing is the leaders normatively reinforced the value of diversity as core to their firm. So they explicitly said, we’re a global company. We want to reflect the diversity of all the societies in which we operate. And then thirdly, the video gave managers a concrete way of making the hiring decision a little bit differently than before. They were invited to reflect on what skills are already present on their teams, but also what skills are missing and how the new person that they are about to bring onto the team could help round out the collective competence, add a new perspective, a new skill that isn’t already represented. And once the video had been created, we then ran a what we call a randomized controlled trial. So a real world experiment where we randomly said to some managers, you’re going to continue to hire just as before, no video. And then some managers were invited to watch the video before getting access to applications. And what we saw was that the video was actually phenomenally effective at increasing both the shortlisting and hiring of women, of non-national candidates, and especially that intersectional group of female non-national candidates, which at the outset had been by far the least likely to get shortlisted and hired. So that’s an example of how we can embed smart, evidence-based scientific approaches to change behaviors, to change outcomes by making small tweaks to how we’re already working.
Brian Bell (00:14:35):
What were some of the key lessons or information that you’re trying to impart in that seven minute video?
Siri Chilazi (00:14:41):
Yeah. Well, as I said, the three key points were this is an important decision. We want to reflect the diversity of societies in which we operate. And then we actually give managers a concrete guide for how to make the decision. But underlying those three things are two key principles that I would urge every organization to keep in mind. The first of which is timeliness. We humans are overloaded with information, especially in this information age. So if you tell me something, honestly, five minutes later, I’ve probably forgotten it. So if we’re trying to give people important information to shift their behavior, we need to get talk to them in a moment that matters, in a moment when they’re paying attention. So managers gain access to this video right before they gain access to the applications. We’ve done other research work with companies where we’re trying to, for example, shift performance evaluations. Well, it doesn’t make sense to give managers some information about performance evaluations in, say, August, if they’re actually going to write those evaluations in December. So you want to target your interventions to the moment that people are going to be making critical decisions. So timeliness is the first principle. And then I already mentioned the second one, actually, which is being targeted. Whenever we’re trying to shift behaviors, we start by identifying a specific decision made by a specific person that we want to change, right? And in this case, it was managers hiring new team members. We want to shift whom they hire. And then you can start to think about, okay, how do we intervene in this process in a timely way? What does the intervention look like? Is it a video? Is it an email that we send? Is it a tweak in the actual IT system, right? Maybe we show them applications in a different order or something like that. But we have to be targeted and timely if we’re going to have a chance at changing people’s behaviors.
Brian Bell (00:16:30):
Do you recall what the lift was in that A-B test in diversity?
Siri Chilazi (00:16:34):
Oh.
Brian Bell (00:16:34):
Yeah.
Siri Chilazi (00:16:35):
Yes, I do.
Brian Bell (00:16:36):
Like the specific numbers?
Siri Chilazi (00:16:37):
Yeah.
Siri Chilazi (00:16:38):
It was double digit increases. So non-national candidates were 16% more likely to get shortlisted, 20% more likely to get hired. Women were 13% more likely to be shortlisted. And female non-nationals were, I believe, 28% more likely to be shortlisted and 41% likely to be hired by managers that had been exposed to the video compared to managers that had continued to hire business as usual.
Brian Bell (00:17:05):
Interesting.
Siri Chilazi (00:17:06):
And these are all statistically significant effects, too, by the way.
Brian Bell (00:17:09):
What was the p-value to get nerdy?
Brian Bell (00:17:11):
No, I’m joking.
Siri Chilazi (00:17:12):
It’s in the paper somewhere.
Brian Bell (00:17:13):
No, I mean, it’s probably a pretty small p-value.
Siri Chilazi (00:17:17):
A very significant value.
Siri Chilazi (00:17:19):
Yeah.
Siri Chilazi (00:17:20):
We had the results published in the Journal of Science earlier this year in January 2025. So I’m happy to add the link to the show notes or something.
Brian Bell (00:17:27):
I’m just joking. I remember all my... and be able to statistics courses and stuff, you know, bring some voice to the critics, which just to keep the conversation lively. So one criticism might be, well, reminding managers to diversely hire might result in less qualified people being placed in those roles. How does a manager decide if like, okay, there’s an equally qualified man and an equally qualified woman, how to choose, right? Are we always choosing the women in that case because they’re underrepresented? Or do we care about equality of opportunity or outcomes?
Siri Chilazi (00:17:59):
It’s a great question. My first question to you is who gets to decide what qualified means? What is qualified, right? When people ask that kind of question and say, well, does this mean less qualified people are getting in or that we’re lowering the bar? That presumes that we’re getting the most qualified people in today. And actually, the evidence is overwhelmingly clear that that’s not happening. Let me give you just one example. Around the world, we’ve done more than 300 so-called audit studies in which researchers send out a pair of resumes to real job openings by companies that have no idea they’re participating in an experiment. And these resumes will differ in just one way. So let’s say one has a man’s name at the top and the other one has a woman’s, but everything else on the resume is completely identical. One resume mentions in the personal section that you’re involved in the Parent Teacher Association, which indicates, of course, that you’re a parent. And the other resume is otherwise identical, but doesn’t include that mention. Now, if we lived in a true meritocracy, which the Merriam-Webster Dictionary defines as a system in which people advance and are given power according to their demonstrated skills and capabilities... If we had that meritocracy in place, then these two candidates should have an identical chance of getting invited to an interview for a job, right? Because all their skills and qualifications are exactly the same. We just know that they have different genders or maybe one is a parent or isn’t. But that’s not what we see in these more than 300 studies done around the world in a variety of different contexts applying to jobs in every industry imaginable. Ageism continues to be real. If you’re a member of an ethnic or religious group that’s a minority in your context, let’s say in your country, you’re going to be less likely to get called in, even with the exact same qualifications. In male-dominated sectors like STEM, science, technology, engineering, and math, women continue to be less likely, even with identical qualifications, to get consideration compared to men. So what I’m trying to say here is that I absolutely believe that we have to select the best person, but the evidence is clear that because we don’t have the most unbiased assessment processes in place today, we actually haven’t been selecting people purely on merit. But we can use science and we can use these evidence-based insights to continue to tweak our processes to make them increasingly meritocratic. That’s what this work is all about. That’s what making work fair is all about. It’s ensuring that everybody gets genuinely equal consideration, as you were saying earlier, an equal opportunity to showcase their skills and their potential. And then let us pick the best people. Let us give them, you know, all the support and resources to thrive and to do their best work for us.
Brian Bell (00:20:46):
Where does our humans inherently bias or is it are we indoctrinated and cultured into it?
Siri Chilazi (00:20:52):
So research shows that all of us are unconsciously biased and often also consciously biased, right? We have these explicit attitudes that we don’t think people are equal in a variety of different dimensions. But then there’s the unconscious level, which by definition, we’re not even aware of. Where that comes from is not settled by modern science. So is some of it ingrained? Possibly. We do know that a lot of it is absorbed because from the moment we enter the world of the living as infants, we’re absorbing information cues. We’re watching the environment and learning about what people do do and inferring from that what people should do and how we should behave. So descriptive norms that describe what’s going on and prescriptive norms that suggest how things should be done. Studies show that kids by age four and five have already internalized a lot of society’s gender stereotypes. So when they’re given two wooden dolls that have no gender characteristics, they’re identical, but one is bigger and one is smaller. Kids will say the bigger one is the boy or the man doll and the smaller one is the girl. And then when they’re asked to attach descriptors to them, they’ll say the male doll, the boy doll, is big and strong, and the girl doll is little, you know? They would use this word as variables, but subservient is the idea. So we definitely do know that a lot of this is learned. through our environment.
Brian Bell (00:22:20):
Yeah, fascinating. So startup teams who are most of our listeners, right? We’re an early stage venture capital firm. We invest in startups and many of our listeners lead or invest in these startups. What are some of the most actionable fair design interventions that startup founders can apply right now in this area?
Siri Chilazi (00:22:36):
Yeah, and I love that question because startup founders are actually in the ideal position building up a new organization to do it right from the beginning, right? To design smartly and purposefully so that when you’re a Titanic with 20,000 employees a few years from now, you know, you don’t have to try to turn that Titanic around. That’d be a really good outcome, by the way.
Brian Bell (00:22:56):
Yes, that’s true.
Siri Chilazi (00:22:57):
That’s true. Not quite in three, but you know what I mean. It’s so much easier to maneuver a nimble speedboat than to try to turn Titanic around. And as a startup, you are that speedboat. So here’s a couple of things. I would say... The first five to 10 hires, the first five to 10 people are going to have a disproportionate impact on both the culture and the concrete ways of working that your company is going to have. And those impacts are going to reverberate years from now. So when you start with your first, you know, the smallest core team, Be extra mindful. Obviously, you want people who have the skills, but what kind of culture are they creating? How do they operate together? Is this an environment where we actually solicit disagreement, constructive debate, where we give everybody an opportunity to have their voice heard, where we don’t overlook potentially critical perspectives, right? If you’re able to set that tone from the beginning, As you scale, you’re so much more likely to retain an environment that both attracts diverse backgrounds and perspectives from the broadest possible pool of talent, but also is able to retain the people and actually harness their wisdom. Because if you assemble 10 people with different perspectives around the room, but only the two most vocal ones dominate the conversation, And no one else feels empowered and safe to ask questions and say, hey, we might have overlooked this critical thing. Well, then the other eight people in the room, it’s pointless to have them there. So that would be my one piece of advice is be extra mindful. And sometimes that might mean even slowing down paradoxically, because in the beginning, we feel like we have to run at our fastest speed. But if we actually slow down to make sure that we put the right people in the right positions, and create the right mechanism for constructive teamwork, that’ll pay massive dividends over the long term.
Siri Chilazi (00:24:43):
My second piece of advice, again, grounded in tons of research evidence, is that structure and formality in general is the enemy of bias, whereas informality and lack of structure is often the breeding ground for bias. And this is particularly relevant in startups because by their nature, they tend to be more informal. Right. You don’t have an HR handbook. You don’t have every policy and process written down because you’re only five people. You’re running at a million miles per hour. You don’t have time to prioritize this stuff. Oftentimes it isn’t until you have maybe 100 or 200 employees when you realize, my goodness, we actually have to write some of this stuff down because the founder can’t manage everything anymore. And real HR concerns and questions and compliance things are coming up. But the sooner you can, however, simply formalize how you make decisions, how you hire, how you give raises to people, how you make business decisions, how you think about who gets promoted or who even gets considered for promotion and when and based on what criteria, you’re much more likely to make those decisions in a fair and objective manner when you have some formalized rubrics and guidance for how to make those critical decisions.
Brian Bell (00:25:55):
Perfect segue into the next question, which is, you know, one of your research areas is measuring and tracking this kind of impact. What are some of the metrics or leading signals a founder or investor should pay attention to when assessing whether their company or a company is building fair processes rather than just, you know, falling back on their unintended biases?
Siri Chilazi (00:26:14):
Yeah, it’s so interesting, right? Because we would never consider running any of our businesses without data. This idea that we measure the things that we care about, set goals to motivate us to move those metrics in the right direction, maybe even implement incentives as additional motivation, and then keep tracking to create a loop of accountability. That’s so obvious when we think about things like product development or sales. But then I always ask leaders I’m talking to, what is your most critical success enabler? If you had to pin your organization’s success down to one factor only, and I know that’s hard, what would that one thing be? And around the world, regardless of the size of organization, regardless of sector, every leader gives me the same answer. And that answer is people. our people, our talent. We need the right people in the right roles at the right time doing their best work so that we as an organization collectively can succeed to our fullest potential. So the great irony here is that we don’t apply that same data and metrics driven management approach that we use to manage sales and product development to managing our most important resource, which is our people, our most important success factor. So you asked a great question of what are the key things to measure? Well, it’s kind of like, what are the key things to measure in your business? It’s going to vary a little bit, right? Depending on what kind of business you are, do you manufacture products? Are you in professional service? What industry are you in? What stage are you in? But I think it comes down to measure what you care the most about. So you probably care about people’s performance, their contribution to your bottom line. You care about retention. You care about ensuring that there is an advancement pathway to folks so that they don’t just stagnate, but the best performers are actually able to move through and get into roles of increasing responsibility and impact. So those are the things that you’d want to be then measuring and ensuring that there are no systematic differences by, sure, demographics like gender and race, but also things like office location, whether someone works in person versus remotely, right? If someone’s a great talent, it shouldn’t matter which office they’re located or if they’re located in a home office, you know, that shouldn’t slow down their trajectory, their advancement.
Brian Bell (00:28:35):
It sure does. I think one of the things I’ve noticed in the last, you know, 10 or 15 years with the rise of remote work is that the people who choose to work remotely, even before COVID, but certainly after COVID, you know, they kind of become a professional in places, what we used to call them at Microsoft, right? They just don’t get promoted. And they’re just kind of there at their level, just doing their job, right? And the people who are at the corporate office are the ones, you know, building the relationships, right? you know, what you could argue is building the bias, right? You know, but we’re fundamentally social creatures, right? So, you know, if you see people every day in the office and you’re building that relationship, you’re probably more likely to get promoted. How do you kind of handle that, you know?
Siri Chilazi (00:29:12):
Yeah. No, it’s a fantastic question. And it gets actually to an organization design point, which is part of the reason why this happens is because all of our processes are how to manage people, how to evaluate their performance. how to know if they’re doing the right thing at the right time. All of those processes and ways of working that we have were developed decades ago for a fully in-person world. The way we teach people to manage others when they first become managers, right? No one is born a manager. It’s a skillset that you learn over time and some people learn it better than others. We’re taught to manage people by in-person surveillance. And so when overnight during COVID, we pivoted to this at first fully virtual world, and now we’ve settled into somewhat of a hybrid, those processes and those ways of working, especially for managers, never kept up paced. Individual employees figured out how to be pretty productive, actually, in many cases, more productive from home than from the office. But managers still five years after COVID and organizations sort of from the top down haven’t figured out how to effectively manage a workforce that’s not all in person. And there are certainly some people, we all know this, right, who don’t want to become the CEO, who don’t want to be the SVP. They’re happy in their current role, doing a really good job, doing it, you know, nine to five or manageable hours. And that’s what they want to do. And I think it’s really important that in our organizations, we offer a pathway for that. for those people to still keep contributing because they’re often doing really amazing work. But then it’s also important that we don’t shaft people into those roles if they have the desire and potential to advance and do more, right? That we don’t make those biased, incorrect assumptions about both people’s preferences and about people’s potential. And so that goes back to to smart management, right? And how do we put the right people in management positions that are really good developers of talent, but also how do we empower those managers with the structures, structures around performance evaluations, around maybe not exit interviews, but stay interviews. Let’s catch people before they’ve left out the door to say, what could we do to retain you? Let’s make sure that we put, give people opportunities in terms of assignments, projects, and tasks. to grow their skills, to stretch them, to show us what they can do before they get disillusioned that I’m never going to have a chance to grow here.
Brian Bell (00:31:33):
Yeah. So given the rise of AI and remote work, how is your research adapting and what are the kind of the new challenges that you’re seeing and opportunities emerging now that weren’t here just a few years ago?
Siri Chilazi (00:31:44):
I’m taking a deep breath in because this is something that we’re all struggling with, I think. Muddling our way through, nobody really has the answer. AI is exhilarating and extremely scary at the same time, at least for me. I’ll speak for myself. It does offer enormous potential both for doing research faster because just the speed at which we can run experiments, analyze the data that comes out of those experiments is increasing with AI. At the same time, AI has the potential to exacerbate a lot of the biases and problems that have been around for a long time already. One example is a new paper that showed that when women have completed a task with the assistance of AI, they’re viewed as less competent because it’s like, oh, they needed help. Whereas when a man in a similar position has accomplished the same task with the help of AI, they’re deemed to be more competent because our inferences look at you you’re being resourceful using a new you’re being resourceful smart exactly right there’s that
Brian Bell (00:32:45):
bias again it’s just that that indoctrinated bias that we have you know exactly it’s like the uh it’s the old joke that people tell about the the son who’s is in an accident and the mom is the doctor right is a punchline and people just can’t figure it out bias you know
Siri Chilazi (00:33:02):
Yeah. So AI has the potential to really exacerbate those biases and sort of amplify them at scale. But on the flip side, a single algorithm, right, a single large language model, it’s easier and faster to de-bias than hundreds or thousands or millions of minds. So there’s the potential. I think what I’m really concerned about as a scientist is that we are not applying scientific rigor to testing and validating the AI technology that’s being unleashed on the market before it actually reaches the hands of users. There’s no external auditing. There’s no external auditing. quality insurance mechanisms. I mean, drugs have to be rigorously tested and approved by the FDA before they’re allowed to flood the market. And the reason we have that system in place is because going back to the 1950s and 60s, before that rigorous system existed, drugs came onto the market that killed babies, killed people, resulted in horrendous birth defects like with thalidomide. And then we realized, oh my gosh, we need to test these technologies more before we unleash them at large. And that’s not happening with AEI yet, which leads to a lot of potential for damage in the short term.
Brian Bell (00:34:11):
So I think a big criticism of DEI over the last five or 10 years has been advocating some people who went a little too far with it were advocating for basically quotas of underlying populations. Like, oh, you should have this percentage of women. And this percentage of African-Americans and so on and so on and so on. And people are like, well, no, that’s not the right way to go about it. And it feels like your fairness at work kind of thesis doesn’t advocate for that. Maybe you could kind of tease out how your approach is different to like those criticisms.
Siri Chilazi (00:34:41):
I do think that there were some problems with how DEI was practiced in the last five to 10 years. As I mentioned earlier, a lot of it was programmatic. Right. So it was, oh, let’s post a black square on Instagram and oh, let’s put a rainbow flag in the window and oh, let’s celebrate International Women’s Day. But let’s not actually make sure that our hiring practices are fair. Let’s not actually conduct a pay audit to make sure that we’re paying people equally and fairly for the same work. Those are the types of structural interventions that would be much more effective. So I think that was one problem. But as you pointed out, whether it was just perception or whether it was actually something that was in some cases happening on the ground, some of the efforts under the DEI umbrella became perceived as efforts to lift up certain groups at the expense of others. And I certainly don’t personally believe that that’s the right thing to do because I actually believe in equality. And the evidence would suggest that by leveling the playing field, by ensuring that no one is unfairly advantaged or disadvantaged in how we run our companies and how we make decisions, We actually can get fair outcomes, which again, doesn’t mean that it’s an equal outcome. Think about the 100 meter dash at the Olympics, right? Everybody starts at the same starting line. They finish at the same finish line. They’re wearing the same sneakers. They have the same outfits. But once the gun goes off, it’s like may the fastest person win. And there is always one winner. The eight racers don’t cross the finish line at the same time. That’s what we’re talking about. It’s making sure that everyone gets to start at the same starting line, that some people’s starting line isn’t 20 meters behind, let’s say. And then once people show up at the starting line, that everyone gets to run in a pair of sneakers so that some people don’t have to be forced to run barefoot, but that everyone has the same aerodynamic outfits. And that already, you know, that’s low hanging fruit because that’s not happening in our companies. We know, for example, that from the moment people enter on day one, not everyone gets the same types of assignments to showcase their skills. Some people are put to work on the high profile assignments, the sexy accounts, the historically highest performing accounts. which of course makes it easier for them to generate a lot of results. And then six months later, you look at these two employees who started on the same day, and one person looks like they’ve generated so much more value, but not because they came in more skilled, actually because we gave them an advantage in the type of project or account that we assigned them to. That’s actually simple to fix. By rotating people through assignments, by making sure that everyone gets to do some of the high visibility, high impact, promotable work, but also has to contribute to the grunt work behind the scenes, we can then make sure that we’re actually able to spot the best performance. And then those are, of course, the people who should be getting promoted.
Brian Bell (00:37:34):
Love that. So let’s talk about the next five or 10 years. What do you think? Where are we all headed with all of this? Looking ahead, what are you excited about? What are some of the dangers?
Siri Chilazi (00:37:44):
Yeah. So I think with this big re-evaluation of DEI, as you were saying, over the last year, to me, this is actually an incredibly hopeful opportunity because companies have had to look at what have we been doing? Has it been working? How much money have we put toward events, let’s say, and what’s the ROI? And the reality is most of these programmatic approaches that were popular, if not effective, don’t really have the ROI there to back them up. And so this has been an opportunity for organizations to say, okay, if this is what we’ve been doing and it hasn’t been working, what should we be doing instead that would be more likely to yield those real measurable results that we’re all looking for? And I’ve talked to a lot of companies around the world over the last year, and I’m seeing this shift in action, which is companies are saying, let’s do fewer events. Let’s do fewer trainings. Let’s do fewer flashy things that talk the talk. And let’s instead walk the walk quietly behind the scenes. Let’s dig into our performance evaluation processes. Let’s dig into how we identify people for the high potential track or how we identify the future leaders of this organization and make sure that the way we make those decisions is actually objective. Let’s make sure that we track the outcomes that we really care about. For example, promotion rates. Are certain groups of employees flying up the ranks faster than others? And if that’s happening, why? Let’s make sure that our policies are actually equal so that everybody who gets a child has access to the same amount of parental leave. That shouldn’t be determined by your gender, by your seniority in the company, or by the method by which a child comes into your family, right? So that’s the kind of unsexy work that I see companies doing behind the scenes now, which frankly is going to yield much better progress and much more results that employees are going to see. And that, I hope, is what continues to happen in the next five and 10 years.
Brian Bell (00:39:38):
What’s an unresolved question in your research that you want to tackle next?
Siri Chilazi (00:39:43):
There’s so many. I’m dying to do research that’s relevant and applicable to practitioners. So I’ll give you a simple example. There’s a lot of work in the past that shows that the words we use in our job advertisements can shape who actually chooses to apply in mostly an unconscious way. So when we use very masculine stereotyped language, like we’re looking for an aggressive coding ninja to lead a team in a fast paced, demanding, hectic environment, that heavy masculine coded language in that job advertisement is actually going to attract more men to apply. And if on the flip side, you’re looking for a warm and caring teacher to work in a collaborative environment, you’re unconsciously attracting more women to apply because you’re using feminine coded language. So our historical advice organizations has been, and there’s companies that offer software to help you do this, either balance out the masculine and feminine coded words in your job ads or go for gender neutral alternatives. And I recently got a great question from a practitioner, which was, well, which one of those approach would work better? Is it better to use both masculine and feminine language or is it better to go for gender neutral words entirely? And I said, you know what? We’ve never run that experiment. We don’t actually know which of those approaches would be most effective. So now that’s an experiment on my list to run. And I actually, if there’s one hope that I could leave the listeners with, it would be to find opportunities like this for experimentation in your everyday work. These experiments don’t have to be big. They don’t have to be fancy. They don’t have to cost money or involve thousands of people. But just in the course of everyday work, as you’re thinking about how do you run your meetings, Even how do you write your emails? How do you decide which person to assign which task? Identify small ways of doing A-B testing, which, of course, we’re already doing as part of our core business all the time. So we know how to do this. We just haven’t applied it internally. And run your business in a more rigorously evidence-based way.
Brian Bell (00:41:35):
Any interesting studies you want to highlight around startups, founders, venture capital for that kind of audience listening?
Siri Chilazi (00:41:44):
So the challenge historically has been that precisely because startups and VC-funded entities are often small, we don’t get the sample size that we were talking earlier about p-values and statistical significance. We don’t get the large sample sizes that we would need to be able to say things with statistical validity, with the confidence when we have sample sizes in the thousands. And I think that’s a limitation of the randomized control trial methodology that I and others use in our work. But I think it shouldn’t stop people from running small experiments. I’m, for example, writing a case study now about a Lithuanian company that has 3,000 employees spread across 10 countries in the Baltics. So they don’t have those massive sample sizes, but they’re running experiments constantly anyway about how to get more referrals, about how to tweak hiring, about how to tweak performance evaluations. And even with small sample sizes, they’re learning lessons that are directly relevant to their own business because they’re experimenting on their own business and they’re then making small changes into how they work. So everyone can do this. I just wish I had more data to share from the startup scene.
Brian Bell (00:42:53):
Let’s wrap up with some rapid fire kind of wrap up questions.
Siri Chilazi (00:42:57):
Sounds good.
Brian Bell (00:42:58):
What is one insight you wish every founder understood about building fair processes from day one?
Siri Chilazi (00:43:02):
That it’s your job as the founder to both do it and to role model how to do it for others. You set the tone. It’s your job and everybody else’s, right? But it has to start with you.
Brian Bell (00:43:13):
Right. Starts at the top. What has surprised you most in your research about where bias or unfairness hides, especially in places you might not expect?
Siri Chilazi (00:43:21):
I’m actually constantly surprised by bias lurks everywhere, but I’m surprised by how very small shifts can yield massive results in fixing the bias. I know it’s a lightning round, but there was a study done at Santander Bank in the UK where men were just accurately informed about how supportive other men were of flexible work, and it resulted in them taking longer parental leaves, massively longer parental leaves. Right. So just that simple email telling them, hey, by the way, 99 percent of men working here support flexible work caused huge behavior changes in terms of the parental leave folks actually took. Small changes, big results.
Brian Bell (00:43:58):
So imagine you were advising an early stage startup, you know, maybe a dozen or a couple dozen employees. What is the first fairness initiative you’d recommend they implement and why?
Siri Chilazi (00:44:08):
It’s not a specific initiative, but it’s an approach. I’d say be like a doctor. When you go to the doctor’s office, they don’t hand pills out at random. They start by asking you questions, running tests, getting your family medical history so that they arrive at a diagnosis. And once they have a diagnosis, then they prescribe the medicine they think will be most effective. So that’s what I’d encourage every organization to do is start by identifying where can we do better? Where are our pain points? Where are our pitfalls? Then you can design targeted solutions.
Brian Bell (00:44:36):
How do you advise companies to balance fairness and high performance if those metrics conflict?
Siri Chilazi (00:44:41):
There’s no balancing because fairness is an essential enabler of high performance. High performance is the result of having the right people in the right roles at the right time doing their best work. And that is just not possible if we’re not objectively able to assess who actually is the person. When should we promote them? When should they be in a given role? And if we don’t give them that level playing field on which they can fully succeed.
Brian Bell (00:45:08):
with AI and automation transforming how work gets done, we’ve talked a little bit about this, but what fairness trap should companies be most aware of over the next five or 10 years?
Siri Chilazi (00:45:16):
I think it goes back to evaluating the impact of all those AI technologies that you choose to leverage in your business. So just like if you make a change into your product on the market, you’re going to be tracking profitability, you’re going to be tracking market share, you’re going to be tracking usage patterns and who’s buying your product to see what impact that design change had. So you should be doing the same thing when you start introducing AI internally in your business. Does this change the outcomes? Does this change who’s using it? Does this change the speed at which we’re making decisions, right? All of those things, because these new technologies always have unintended consequences. The only way to spot them and the only way to mitigate their negative effects is by tracking.
Brian Bell (00:45:59):
Well, I really enjoyed the conversation. I learned a ton. Thanks so much for coming on, Siri.
Siri Chilazi (00:46:04):
Thank you, Brian. It was my absolute pleasure. And thank you to everybody for listening.
