Latest / Future of Work Tech with Fexingo: Remote Tools, AI Productivity, and Workplace Software / How AI Copilots Are Quietly Reshaping White-Collar Work
Transcript
- Lucas: So there's this McKinsey report from late last year that I keep coming back to — and it's not the flashy one about how many jobs AI will replace. It's the one nobody talked about, about how people are actually using these tools day to day. Luna: The one that found the average knowledge worker is using an AI copilot for about 30 percent of their daily tasks now? Lucas: Exactly. That number was 8 percent in early 2024. So we've gone from a curiosity to nearly a third of a typical workday in two years. But the really interesting part isn't the headline — it's that most of that usage is completely invisible to managers. Luna: Invisible how? Like, people aren't telling their bosses they're using these tools? Lucas: Right. McKinsey's survey data shows that over 60 percent of employees who use AI copilots regularly haven't disclosed it to their direct supervisor. They just installed the tool, started using it, and never mentioned it. And in many cases, their companies don't even have a policy about it. Luna: That feels like the early days of bring your own device, where everyone was using their personal laptop for work because the company laptop was terrible. Lucas: It's a really good parallel. BYOD was this wave that IT departments initially resisted, then eventually had to embrace because the productivity gains were undeniable. I think we're in the early BYOAI phase — bring your own AI. And the data suggests the gains are real. McKinsey found that the heaviest copilot users report a 22 percent reduction in time spent on routine tasks like drafting emails, summarizing documents, and data entry. Luna: Twenty-two percent is huge. But if managers don't know who's using these tools, how do you even measure that? It's self-reported, right? Lucas: It is. And there's a selection bias problem — the people who choose to use AI tools might already be more efficient. But even so, the scale is striking. McKinsey estimates that if you extrapolate across the US white-collar workforce, the aggregate time saved is equivalent to about 9 million full-time workers' worth of hours per year. That's not nothing. Luna: So if that time is being freed up, what are people actually doing with it? Because the fear is always that we just get more meetings. Lucas: That's the million-dollar question. The study suggests that about half of that saved time goes into higher-value work — things like strategic thinking, client relationship building, creative problem solving. The other half, frankly, goes into personal tasks or just shorter hours. Some workers are using the productivity gain to leave at 5 p.m. instead of 7 p.m. Luna: Which is a different kind of disruption. If some people are getting their work done in six hours and others are still grinding for ten, that creates a two-tier workforce within the same office. Lucas: Exactly. And it's not evenly distributed. The data shows that younger employees — under 35 — and employees at tech-forward companies are adopting copilots at roughly double the rate of older workers or those in traditional industries. So you're getting an AI adoption gap that mirrors the digital divide. Luna: I remember a case from a mid-sized law firm in Chicago that was in the news last month. They rolled out an AI document review tool, but only 40 percent of partners used it. The associates used it almost universally. Lucas: Yeah, that's a perfect example. The partners — who bill at the highest rates — essentially said, 'I trust my own judgment.' The associates saw it as a way to get through 200 documents in an hour instead of two days. And the firm couldn't figure out why the associates' productivity metrics suddenly looked better than the partners'. Luna: So you have this interesting tension: the tools are clearly useful, but they're being adopted in a patchwork way that leadership can't see. What does that mean for how companies should think about AI strategy? Lucas: I think the smartest companies are moving away from top-down mandates — 'everyone must use this tool' — and instead trying to create conditions for organic adoption while maintaining visibility. Some are setting up internal AI marketplaces where employees can try approved tools and share feedback. Others are doing quarterly audits of software installations across devices, just to see what's actually being used. Luna: And then there's the question of data security. If employees are using free consumer-grade AI tools with company data, that's a risk. Lucas: Massive risk. McKinsey flagged that too. Over a third of respondents said they've pasted confidential company information into a public AI tool — sales data, customer lists, internal strategy documents. Most of them didn't think twice about it. That's a compliance nightmare waiting to happen. Luna: So we've got this picture: productivity gains that are real but uneven, a visibility gap that leaves leadership in the dark, and security risks that are growing faster than policies can keep up. What's the takeaway for someone running a team right now? Lucas: Listen before you dictate. The worst thing a manager can do is ban AI tools outright — that just drives adoption underground. Instead, start a conversation. Ask your team: what are you using? What's working? What's not? You'll probably learn that they've already figured out a lot of things you haven't. And then build policy around that reality, not around some idealised version of how work should happen. Luna: That's the opposite of the command and control approach. It's more like gardening than engineering. Lucas: Yeah. And honestly, that might be the biggest shift of all. The future of work isn't about which AI tool wins — it's about whether leaders can adapt to a world where their teams are already ahead of them. Luna: Hey, before we move on — I want to mention something quickly. We keep this show ad-free by design, and that's only possible because of listeners who chip in. If today's episode was useful to you, you can support the show at buy me a coffee dot com slash fexingo. It's a small ask, but it makes a real difference. Lucas: Absolutely. Even a couple of dollars a month helps us keep these conversations going without any sponsors or interruptions. So thank you to everyone who already does. Luna: Alright — so let's get back to that security angle. You mentioned that a third of employees have pasted confidential data into public AI tools. What's the actual exposure here? Lucas: The exposure is that most consumer AI tools — the free versions of ChatGPT, Claude, Gemini — they use inputs for training unless you opt out. So if an employee pastes a customer list into a prompt, that list could become part of the model's training data. That's a data breach, effectively. Luna: And companies are only now waking up to this. I've seen several banks and law firms issue blanket bans on certain tools in the last six months. Lucas: Right, but bans don't stop shadow adoption. What's more effective is providing approved enterprise-grade tools that have data privacy guarantees. Microsoft Copilot for Microsoft 365, for example, runs inside your tenant — data doesn't leave the organization. Same with some of the enterprise tiers from OpenAI and Anthropic. Luna: So the answer isn't 'no AI,' it's 'safe AI.' Lucas: Exactly. And the companies that figure out how to balance safety with flexibility are going to have a real talent advantage. Because the workers who are most comfortable with these tools — they're not going to want to work somewhere that treats AI like a threat. Luna: That's a good closing thought. So to summarize: copilot usage has tripled in two years, most of it is invisible to management, it's creating productivity gains that are real but uneven, and the security risks are serious but solvable with the right tools and culture. Lucas: That's the picture. And the next 12 months will tell us whether companies step up to manage this intentionally, or let it stay underground until something forces their hand. Luna: I suspect we'll see a few high-profile incidents that accelerate corporate policy-making. But for the listener who wants to get ahead of that curve — start the conversation now. Lucas: Yeah. Don't wait for the crisis. Ask your team tomorrow what they're using. You might be surprised by what you learn.