Latest / The Diversity Career Podcast with Fexingo: Inclusion, Equity, and Underrepresented Professionals / How One Company Matched Every Employee to a Mentorship
Transcript
- Lucas: There's a well-known problem with mentorship programs in large companies. They're almost always optional, they rely on junior employees to find their own mentors, and the people who need mentoring most are often the least likely to ask. Luna: Right — it becomes yet another task for someone who's already stretched thin, or who doesn't feel comfortable cold-emailing a senior leader. Lucas: Exactly. So a couple of years ago, one of the big global consulting firms decided to flip the model. Instead of saying 'go find a mentor,' they assigned one to every single employee — using an algorithm. Luna: An algorithm to match people for mentorship? That's either brilliant or a recipe for awkward coffee chats. Lucas: Both, potentially. But the results are worth looking at. The firm — I'll call it 'Apex Consulting' because they haven't publicised the program widely — matched roughly 12,000 employees across five countries. They used a system that considered each person's stated career goals, their team, their tenure, and their self-identified demographic background. Luna: And they did this for everyone? Not just high-potentials or underrepresented groups? Lucas: Everyone. Every single employee got a mentor match within their first 90 days at the firm. The thinking was: mentorship shouldn't be a perk for the well-connected. It should be infrastructure. Luna: That's a big philosophical shift. And I imagine it costs a lot to administer. Lucas: It does. But the firm calculated that turnover costs were higher. And this connects to something we talk about a lot on this show — if we've helped you think through a career move or a strategy at your own company, we'd love it if you considered supporting the show. We deliberately don't run ads. We think that keeps the conversation honest. If that's valuable to you, you can support us at buy me a coffee dot com slash fexingo. No pressure, just an option. Luna: Yeah, and it genuinely helps us keep doing episodes like this one. So, back to Apex — how did the algorithm actually work? Lucas: The first version was pretty simple. It looked at job function, seniority level, and location. Then it cross-referenced those with the mentor's availability and the mentee's stated goals — things like 'I want to improve my client presentation skills' or 'I'm interested in moving into a leadership role.' Luna: So it wasn't just random — it was preference-based. Lucas: Correct. But here's where it got interesting. In the first year, the data showed that women of color in the program were 40 percent less likely to be matched with a mentor who shared their racial or gender identity. The algorithm was optimising for functional fit, but it was replicating the existing demographic structure of the firm. Luna: So the algorithm mirrored the company's own lack of diversity at senior levels. Lucas: Exactly. A senior white male partner was available and had the right skills, so the system kept pairing junior women of color with him. Not terrible — but not ideal for the mentorship outcomes they wanted. Luna: What did they change? Lucas: They added a second filter: a preference-matching option. They asked every employee to optionally indicate whether they'd like a mentor with a similar background, or a different one. And they weighted that preference in the algorithm. Luna: So if a junior Black woman said she wanted a mentor who shared her experience, the algorithm prioritised that over pure functional match. Lucas: Right. And they also expanded the mentor pool. They brought in mid-level managers and even some senior individual contributors, not just partners. That gave the algorithm more diverse nodes to work with. Luna: And the results after that change? Lucas: Within two years, promotion rates for underrepresented groups in the program increased by 25 percent compared to a control group of employees who hadn't been matched yet. Retention among those with a mentor also improved — 18 percent lower voluntary turnover. Luna: Those are significant numbers. But I'm curious about the people who didn't want a mentor at all. Lucas: Good question. They made it opt-out, not opt-in. About 12 percent opted out. But the firm tracked them anyway, and interestingly, those who opted out still had slightly higher promotion rates than the control group — possibly because just being asked made them think more deliberately about their career. Luna: That's a nice Hawthorne effect. So the real lesson here isn't just 'use an algorithm.' It's 'use an algorithm that's transparent about its own biases.' Lucas: Precisely. The algorithm is a tool. If you don't audit it, it'll reproduce whatever inequities exist in your data. Apex had to build in a preference filter and actively recruit a more diverse mentor pool to make the system work. Luna: And they had to be willing to change the algorithm when the data showed a problem. That takes humility — most companies would just pat themselves on the back for having a mentorship program at all. Lucas: Right. And Apex didn't stop there. They also started publishing anonymised data on match quality — things like how many meetings each pair actually had, and whether mentees felt the relationship was useful. That data fed back into the algorithm. Luna: So it became a learning system, not a static list. Lucas: Exactly. And one more thing — they capped the number of mentees per mentor at three. That prevented the most popular senior leaders from being overwhelmed, and it forced the algorithm to distribute mentoring across more people. Luna: That's smart. Because if you only have a few 'good' mentors, you're just creating a bottleneck. Lucas: Right. And it also meant that mid-level managers got a chance to develop their mentoring skills. A few of them later became partners, and they credited the mentoring experience with giving them visibility they wouldn't have had otherwise. Luna: So the program ended up benefiting the mentors too, not just the mentees. Lucas: Exactly. It's a two-way street. Apex found that mentors who participated were 30 percent more likely to stay at the firm over the next three years. It gave them a sense of purpose and connection. Luna: That's a huge retention lever, especially for mid-career people who might otherwise feel stuck. Lucas: Absolutely. So what can a smaller company take from this? You don't need a massive algorithm. But you do need to make mentorship systematic, not optional. And you need to measure outcomes by demographic group. Luna: Which is something most companies don't do. They track number of matches, but not whether those matches actually lead to promotions. Lucas: Right. Apex tracked promotion rates, retention, and even qualitative feedback. They found that pairs who met at least twice a month had significantly better outcomes. So they built reminders and scheduling tools into the platform. Luna: That's the kind of operational detail that actually moves the needle. Not just 'have a mentor,' but 'have a mentor who you meet with every two weeks and who has been trained to be an effective sponsor.' Lucas: Exactly. And they trained mentors, too. The algorithm was just the matching layer. The real work was in teaching mentors how to advocate for their mentees — how to recommend them for projects, how to introduce them to key people. Luna: That's the difference between a mentor and a sponsor. A mentor gives advice; a sponsor opens doors. Lucas: Right. Apex explicitly trained mentors to act as sponsors. And they tracked whether mentees received visible assignments after six months in the program. That was one of their key metrics. Luna: So they weren't just measuring activity — they were measuring outcomes. Lucas: Exactly. And the data showed that women of color who had a sponsor-mentor were promoted at nearly the same rate as white men, closing a gap that had existed for years. Luna: That's the kind of result that makes the investment worth it. I wonder how many companies are actually doing this. Lucas: Not many. But the ones that are — and we've covered a few on this show — tend to see real returns. The key is treating mentorship as strategic, not as a nice to have HR checkbox. Luna: And being willing to fix the algorithm when it gets it wrong. Lucas: Right. Because if you don't, you're just automating inequity. Apex's story is a good reminder that technology can help — but only if you're honest about the data it produces. Luna: And only if you're willing to change the system, not just blame the tool.