Latest / Future of Work Tech with Fexingo: Remote Tools, AI Productivity, and Workplace Software / How AI Is Automating Your Internal Email Drafting
Transcript
- Lucas: So there's a number that's been rattling around my head since I saw it in a McKinsey report last quarter: the average knowledge worker spends roughly two and a half hours a day on email. That's about 30 percent of their entire working week. Luna: And most of that time is not reading — it's drafting replies, crafting the right tone, cc'ing the right people. Lucas: Exactly. And the tone piece is actually the harder part. Getting the substance down is usually straightforward. But figuring out whether your email sounds too curt or too effusive, especially across different audiences — that's where the friction lives. Luna: So this is where generative AI steps in. Not just suggesting the next word, but drafting the whole thing from a few bullet points. Lucas: Right. And there are now a bunch of tools doing exactly that — integrated directly into Gmail, Outlook, even Slack's canvas feature. But I want to zoom in on one specific deployment I came across. A mid-size tech company — about 800 employees — rolled out an internal AI email assistant to their whole workforce back in March. Luna: And what did they find? Lucas: Within six weeks, they measured a 23 percent reduction in time spent on routine internal correspondence. The tool was pulling context from the user's calendar, their recent email threads, and a stored preference profile that included things like formality level and typical sign-off style. Luna: So the model is basically learning your voice. That's interesting because the biggest complaint I hear about ai written text is that it sounds generic — like a LinkedIn post written by committee. Lucas: That's exactly the problem these newer tools are trying to solve. Instead of a single generic tone, you can give it a few examples of your own writing and it fine-tunes a lightweight model on the fly. The company I mentioned let users select from three tone presets — 'direct,' 'warm,' and 'neutral' — and then further adjust with a slider that goes from 'very casual' to 'very formal.' Luna: A slider. That's actually kind of brilliant. But it does raise the privacy question immediately — the model has to read your emails to learn your style. That's a non-trivial ask. Lucas: It is. And the company handled it by making the training opt-in and fully on-device for the first month. They didn't send any email data to a cloud model until users explicitly approved it. That slowed adoption a bit — only about 40 percent opted in initially — but the ones who did reported much higher satisfaction. Luna: So the people who tried it liked it. But the other 60 percent were wary. That's a real adoption barrier. Lucas: It is. And I think that's the central tension with a lot of these generative AI workplace tools right now. The value proposition is clear — you save time, you reduce cognitive load, you stop agonizing over whether your email to the VP of Engineering sounds too passive. But the cost is that you're handing over a pretty intimate record of your professional communication. Luna: And it's not just privacy. There's also the question of authenticity. If I use an AI to write an email to my colleague, is it really me communicating? Or is it a smoothed-out version of me? Lucas: That's the 'not like me' pushback I've heard from several employees at that company. They said the drafts were grammatically perfect and polite, but they didn't sound like the person they are in the hallway. One engineer told me, quote, 'My real emails have typos and run-on sentences. That's how people know it's me.' Luna: Honestly, I kind of love that. There's a certain charm to imperfect human writing. But I also wonder if that's a generational thing — maybe people who grew up with autocorrect and smart compose are more comfortable with the machine smoothing things out. Lucas: It could be. The data from this company showed that adoption was highest among employees under 35, and lowest among those over 50. The younger group was also more likely to use the 'warm' tone preset, while the older group mostly stuck with 'direct' or didn't use the tool at all. Luna: Fascinating. So the tool is not just saving time — it's actually changing the tone of internal communication. And maybe that's a good thing overall, but it's a shift worth paying attention to. Lucas: Yeah, and that's actually what I want to dig into for the rest of the episode — the second-order effects. Because once you start drafting emails with AI, it doesn't just affect the sender. It affects the receiver too. If everyone starts sounding warmer and more polished, does that change the culture? Luna: Or does it flatten it? If everyone uses the same 'warm' preset, do you lose the quirky individual voices that make a team feel human? Lucas: Right. And I think there's a parallel here with how we've seen AI impact other forms of writing — like how everyone's Instagram captions started sounding the same when the algorithm favoured a certain style. But let's stay with the workplace for a moment. Luna: Sure. So beyond the adoption curve, what did the company actually see in terms of outcomes? They saved time, but did the emails actually work better? Lucas: That's a great question. They did measure response rates — and they actually saw a slight increase, about 4 percent, in the likelihood that an ai drafted email got a reply within 24 hours. The hypothesis is that the ai drafted emails were clearer and more action-oriented, so recipients knew exactly what was being asked of them. Luna: That makes sense. A lot of badly written emails bury the ask in the third paragraph. If the AI puts it upfront, you're more likely to get a response. Lucas: Exactly. But there was also a small but measurable increase in the number of follow-up emails needed. So the initial reply came faster, but sometimes it was a request for clarification because the AI had stripped out some nuance that a human would have included. Luna: So you save time on drafting but potentially lose time on back and forth because the AI is too efficient. There's a trade-off. Lucas: There is. And I think that's where the human-in-the-loop becomes critical. The best use case, based on the interviews I did, was for routine, low-stakes emails — scheduling, status updates, quick approvals. For sensitive or emotionally charged messages, almost everyone said they preferred to write from scratch. Luna: Which is smart. You don't want an AI drafting your apology email or your performance feedback. That's where authenticity really matters. Lucas: Right. And I think the tool designers are starting to recognize that. The newer versions let you flag a draft as 'high sensitivity' and it will deliberately write a less polished version — even adding in a few grammatical imperfections to make it sound more human. Luna: Wait, the AI is intentionally adding typos? That's almost postmodern. Lucas: It is. And it shows how much we've learned about the social cues embedded in writing. A perfectly formatted email can feel like it came from a robot, even if a human wrote it. So the AI is now learning to simulate imperfection. Luna: That's a weird loop. We're using machines to sound less machine-like. But I guess if it makes the communication more effective, it's worth it. Lucas: I think so. And honestly, if this kind of thinking about technology and communication is useful to you — if today's conversation gave you something you can actually take back to your team — that's exactly the kind of listener support that keeps this show ad-free and independent. Luna: Yeah, it's a small thing but it makes a real difference. If today was worth a coffee to you, the link is buy me a coffee dot com slash fexingo. Lucas: And we mean that genuinely. No pressure, no perks — just knowing that the show is valuable enough that someone would toss a few bucks into the digital tip jar. It helps us keep doing deep dives like this one. Luna: Totally. So back to the email AI — one thing I haven't asked: did the company see any impact on employee satisfaction? Saving time is one thing, but if people feel like their communication is being mediated by a machine, that could be a net negative. Lucas: They did survey that. And the results were mixed. About 55 percent said they felt less stressed about email. But 20 percent said they felt less connected to their colleagues because they were reading more ai generated messages and fewer raw human ones. Luna: That twenty percent is significant. It suggests that even if the tool is efficient, it might be eroding some of the informal social glue that holds teams together. Lucas: Exactly. And I think that's the part that companies tend to overlook when they roll out productivity tools. They measure time saved, but they don't measure relationship quality. And on a remote or hybrid team, email is often one of the primary ways you get a sense of someone's personality. Luna: So what's the fix? Do we just accept that trade-off, or are there ways to design the tool to preserve more of the human element? Lucas: I think there are ways. The company I looked at is experimenting with a feature that shows the original human-written version alongside the ai polished version, so the sender can blend the two. They call it 'hybrid mode.' The idea is you start with the AI draft, then you manually add back a few of your own quirks. Luna: That feels right. Keep the efficiency, but give the user control over the final voice. It's the same principle as using Grammarly but then going back and adding your own slang. Lucas: Exactly. And the early feedback on hybrid mode has been positive. Users say they feel more ownership over the final message, and recipients report that the emails still feel human. The key, I think, is that the AI should be an accelerator, not a replacement. Luna: Well said. So where do you see this going in the next year? Are we all going to have AI assistants drafting our emails by default? Lucas: I think we'll see a lot more integration, but also a lot more customization. The one-size-fits-all approach is dying. Instead, we'll have models that are fine-tuned on your specific writing style, your team's communication norms, even your company's internal vocabulary. And they'll get better at knowing when to step in and when to stay out. Luna: So the future of email might be less about writing and more about editing. You become the editor-in-chief of your own inbox. Lucas: That's exactly the framing I've been using. You're not the writer anymore — you're the editor. You set the tone, you review the draft, you hit send. And that shift, from writer to editor, might actually be a net positive for productivity, because editing is a higher-leverage skill than drafting. Luna: I like that. It reframes the AI not as a crutch but as a tool that lets you focus on the parts of communication that actually require human judgment. Lucas: Exactly. And I think that's the real takeaway from this episode. The technology is moving fast, but the human role isn't disappearing — it's shifting. And the teams that understand that shift, and design their tools around it, are the ones that will get the most value without losing their culture. Luna: Alright, I'm curious to see how the hybrid mode evolves. If you're listening and you've tried something like this, let us know. Lucas and I would love to hear your experience. Lucas: Absolutely. And if you want to support the show, the link is buy me a coffee dot com slash fexingo. Thanks for listening.