Latest / Future of Work Tech with Fexingo: Remote Tools, AI Productivity, and Workplace Software / How AI Avatars Are Replacing Your Coworkers on Zoom
Transcript
- Lucas: So you get a calendar invite for a weekly team update. You join the Zoom, and there you are — literally, your face, your voice — delivering the update. But you're not actually there. You're at the dentist. The person on screen is an AI avatar trained on your previous recordings. Luna: Wait, that's actually happening? I've seen the demos where you type text and get a talking head, but people are using these in real meetings? Lucas: They are. And the numbers are growing fast. Companies like Synthesia and HeyGen have seen their enterprise customer base double over the past year. Synthesia alone raised over ninety million dollars in 2025. And the use case is shifting from marketing videos toward internal communication — async standups, sales enablement, even client-facing product demos. Luna: It's interesting you mention the dentist thing, because that's exactly the kind of scenario where an avatar could save someone's day. But also — if I'm on the receiving end, do I know it's an avatar? Or is it presented as the real person? Lucas: That's the key ethical question, and it's one the companies are dancing around. Synthesia's terms of service require consent and disclosure, but in practice, if someone sends you a pre-recorded avatar video, they might just say 'here's my update' without explicitly saying 'this is a digital twin.' The technology has gotten good enough that unless you're looking for tells — a slight glitch in the mouth movement, a too-perfect posture — you might not notice. Luna: And by the way — episodes like this are exactly why listener support matters. A couple of dollars a month is genuinely what keeps these going — buy me a coffee dot com slash fexingo, if you've gotten something out of them. Lucas: Yeah, it's a small thing that makes a big difference for us. So, back to avatars — I want to talk about a specific case that I think illustrates both the promise and the risk. There's a mid-sized marketing agency, about two hundred people, that started using avatar-generated weekly updates about six months ago. They cut meeting hours by forty percent in the first quarter. Luna: Forty percent? That's huge. How did they structure it? Lucas: Each team lead records a five-minute avatar update once a week. The avatar is trained on a twenty-minute video of them speaking naturally. Then the lead writes a script, the avatar delivers it, and the team watches asynchronously. No live meeting, no scheduling conflicts, no 'can you repeat that?' The avatar even matches the lead's tone and gestures. Luna: So the leads are effectively cloning themselves to do the boring part of their job. I can see why they'd love that. But what about the team watching? Do they feel like they're being talked at by a robot? Lucas: That's the interesting part. In their internal survey, seventy percent of viewers said they preferred the avatar updates over live meetings because they could watch at two times speed, pause, or skip. They felt more in control of their time. The complaints came from about fifteen percent who said they missed being able to ask follow-up questions immediately. So the agency kept one live Q&A session per month for that. Luna: So it's not a total replacement — it's a hybrid model. The avatar handles the one-way broadcast, humans handle the interaction. That feels like the right balance. Lucas: Exactly. And that's where most companies are landing. But there's a more extreme version emerging. A few sales teams are using real-time avatar interfaces — where the avatar interacts with clients using a large language model under the hood. The client asks a question, the model generates the answer, and the avatar speaks it in the salesperson's voice and face. Luna: That's getting into deepfake territory. Is that legal? Lucas: It's a gray area. If the client knows they're talking to an AI, it's fine. But most companies don't disclose it because they're worried clients would feel deceived. There was a case last year where a startup used HeyGen's real-time API to let a sales rep 'attend' three client calls simultaneously. The rep would listen in and occasionally override the AI. The clients had no idea. When one found out, they threatened to sue for fraud. Luna: Yikes. That's the kind of story that could set the whole industry back. The trust issue is huge. Lucas: It is. And the response from the avatar companies has been mixed. Synthesia has a pretty strict policy — they require all avatars to be created with consent and they watermark non-consensual content. But HeyGen has been more permissive, and their API is easier to embed. So we're seeing a split between the cautious players and the ones racing to capture market share. Luna: What about the quality? How close are we to photorealistic avatars that could fool anyone? Lucas: We're alarmingly close. If you watch a demo from 2024, you can see the uncanny valley — stiff lip sync, weird eye movements. The 2026 versions are dramatically better. Synthesia's latest model, released in March, uses neural radiance fields to capture facial micro-expressions. The result is that the avatar can smile, frown, raise an eyebrow in sync with the script. The only tell left is that the avatar's responses are pre-scripted — it can't improvise naturally. Luna: So if you ask it a question it didn't expect, it might give a generic answer or freeze. That's the limit. Lucas: Exactly. And that's why the real-time conversational avatars are still a work in progress. They use a large language model behind the scenes, but the latency and the naturalness of the delivery aren't there yet. Most of them have a one to two second delay between the question and the response, and the voice intonation sounds synthesized. So for now, the sweet spot is pre-recorded avatar updates. Luna: I can see this becoming a standard tool for distributed teams. But it raises a bigger question: if your digital twin can do your meetings, your standups, your client calls — what does it mean for your job? Are you now just a scriptwriter for your avatar? Lucas: That's the existential angle. Some futurists are already talking about the 'avatar economy' where knowledge workers are valued for their strategic thinking and their avatar handles execution. But there's a darker flip side: employers might start expecting employees to be available 'through' their avatar even when they're off the clock. If your avatar can take a meeting at 10 PM, why wouldn't the company schedule it? Luna: Right. The avatar never sleeps. So the boundary between work and life blurs even more. Lucas: Exactly. I spoke to a labor lawyer who specializes in tech, and she pointed out that under current employment law, if an avatar is working, the real person isn't necessarily being compensated because they're not actively performing work. But the company is still getting value from their likeness and data. That's a legal time bomb. Luna: So we could see a wave of litigation over avatar rights — who owns the digital twin, who gets paid when it works, and what happens when it says something wrong. Lucas: Exactly. There's already a precedent. In 2024, a former employee sued a company for using his avatar in a sales pitch after he left. The court ruled that the avatar was a form of personal data, and the company needed explicit consent for each use. That case is shaping how companies draft their employment contracts now. Some are including clauses that grant the company a perpetual license to the employee's avatar. Luna: Which employees might not realize they're signing away. Lucas: Right. And that's the thing — this technology is moving so fast that the legal and ethical frameworks are struggling to keep up. We're at a point where a tool that saves forty percent of meeting time could also erode the very idea of presence at work. Luna: So what's the takeaway for someone listening who's considering using avatars in their team? Lucas: First, be transparent. Tell people when they're interacting with an avatar. Second, limit avatars to one-way communication for now — use them for updates, not negotiations. Third, get legal review on your employment contracts around avatar rights. And fourth, keep the human connection alive. The avatar can handle the information dump, but the real conversations — the ones where you read the room, where you build trust — those need to stay human. Luna: That's a good rule of thumb. The avatar as assistant, not replacement. Lucas: Exactly. And if we get that balance right, we could actually reduce meeting fatigue and give people more time for deep work. But if we get it wrong, we end up with a workforce of digital puppets and an even bigger trust deficit. Luna: Let's hope teams choose wisely.