Connect with us

Tech

Transcript: AI and Work — Can I send a chatbot to that meeting?

Published

on

This is an audio transcript of the Working It podcast episode: AI and Work — Can I send a chatbot to that meeting?

Laila
Hi there. I’m Iliana. I’m an innovation executive, global citizen and public speaker. Well, actually, I am Laila, Iliana’s digital human twin. I allow Iliana to be in two places at once. And I’m making her life easier.

Isabel Berwick
Hello and welcome to Working It from the Financial Times. I’m Isabel Berwick. Over the last year or so, we’ve been hearing about AI nonstop. But a lot of the articles, podcasts and books on AI are a bit short on specifics. We keep hearing it’ll change the way we work, but not what that change will actually look like. So over the next three weeks, I’ll be bringing you real world stories about artificial intelligence in the workplace and speaking to people who are using AI in their daily working lives. You’ll hear about the jobs that AI might replace, when we can expect that to happen, and why it may change the way we think about work forever. In this episode. Do you ever wish there were two of you doing your job?

Iliana Oris Valiente
I’ve essentially created an AI powered digital human twin of myself. She looks like me. She sounds like me. She’s been trained on the last five years worth of case studies of every single project that my teams have run in Canada. In addition to being trained to multiple pieces of Accenture thought leadership and research reports, we’ve also fed her my various personality assessment results. And the idea is that she becomes an extension of me, powered by AI and giving me superpowers.

Isabel Berwick
That’s Iliana Oris Valiente. She’s a managing director at Accenture Canada, and she’s also the company’s head of innovation. I spoke to her from her offices in Toronto.

We’re here to talk about your digital twin, Laila. Could you tell me a bit about her? What’s the problem she’s solving? Is it that there’s only one of you?

Iliana Oris Valiente
I’m a bottleneck. The objective of building Laila is to A, allow me to be in multiple places at once. B, allow me to shift some of the lower value work that I do off to the AI that then frees up more of my time, that I can then reinvest in activities that are the highest and best use of a human.

Isabel Berwick
Can you talk a little bit more about how you made her look like you? Because I think you went in for, is it body scans . . .

Iliana Oris Valiente
I had to submit really awkward close up photos of my face from all angles. I’ve had animation designers marking up my hair. We also did these 3D body scans. So volumetric video of someone walking around me as I stood still. And then you get to the voice. I have spent upwards of six hours in podcast recording studios, recording everything from reading children’s stories to reading obituaries. So Laila could understand that nuance in tone. So that was the voice process. And then we, of course, trained her on various personality results of mine, in addition to all of the Accenture thought leadership and the Accenture projects that my teams have worked on.

Isabel Berwick
Did you come up with the idea for Laila? How did she come about?

Iliana Oris Valiente
We have a team at Accenture that’s been thinking about this concept of digital humans for the past few years, and with the advent of Gen AI, think of it as the accelerant that suddenly was able to take the work we had been previously doing, combining it with the magic of LLMs to unlock a whole host of new capabilities. So, for example, we recently used Laila at the South by Southwest conference. It’s a major event. I go to lots of conferences, I enjoy them, what I don’t enjoy, and I think your listeners will resonate with this as well, is the act of going through the conference website and figuring out what should be on your personalised agenda. Which sessions should you attend? Read the descriptions, read the speaker bios and decide is it in? Is it out? With Laila, she can be trained on all of those details. So you can just have a conversation with her, and suddenly a three-minute conversation where you explain who you are, what you’re interested in, what you’d like to explore, what you don’t want to hear more of. And tadaah, she can make you a customised itinerary I would otherwise be spending an hour or an hour and a half if I were to really meaningfully prepare to make the most out of every minute I’m spending at an event like that. A human can’t scale. Meanwhile, with AI, you can suddenly do this at scale, and it’s easy.

Isabel Berwick
So would I be right in saying that she can sort of filter and solve things like scheduling problems, but she wouldn’t be able to go to the events herself, or would she?

Iliana Oris Valiente
She’s not able to go to events herself yet, but she can and absolutely will start to replace my attendance in certain meetings. For example, every week, at least several times a week, somebody approaches me, either an internal person at Accenture or a client, and the conversation goes along the lines of Iliana, we’re facing problem XYZ and such and such industry. Have you done anything along these lines before? Is there anything that we should be thinking about and considering? And I sit there and I go through the Rolodex of all of our projects and all of our experience, and I say, this kind of reminds me of this example over here and that example in, in an adjacent industry. And you should maybe connect to these three people and go read this one research report. Now, if I’m not available that day, once again, I’m sick. I’m otherwise committed. That conversation can’t move forward, and nor do I necessarily want to be having that conversation, because by the time you’ve had it 15 times, it’s really not that interesting. Meanwhile, Laila doesn’t forget she has this infinite memory bank, and she can be trained on those associations and to be able to point people to the research that they need. And by the time they’re ready to have a conversation with Mimi, the real human, we can hit the ground running and have this incredibly productive conversation. And who doesn’t want to have a more productive meeting?

Isabel Berwick
If your colleagues are getting your avatar essentially instead of you, does that annoy them? Do you get pushback?

Iliana Oris Valiente
I don’t think it’s going to get any level of pushback because I’m not replacing myself. I am augmenting myself. And if someone had the choice of waiting two weeks to get time on my calendar, versus talking to Laila today, I think they’d rather take the option of talk to Laila today, get her pointers, prepare. And by the time you have that meeting with me, we’re all that much more effective. And then, instead of it being a rushed conversation, we can actually sit down over lunch and have a proper discussion.

Isabel Berwick
Yes. And is it pretty clear to you and all your colleagues which meetings and events Laila goes to, and which you go to? Has that evolved or has it been clear from the start? I’m interested in on behalf of listeners who are just trying to get their heads around what a digital assistant might be able to do.

Iliana Oris Valiente
It’s evolving every day. I’ll give you an example. Part of my role is that I oversee so many different domains and technologies, and I’m constantly context switching. And if I didn’t have to memorise all of the statistics from all of our research reports and instead be able to call Laila up as my secondary co-presenter if I’m at a conference, wouldn’t that be compelling?

Isabel Berwick
Is there anything that she’s better at than you are?

Iliana Oris Valiente
She is better at endless recall. If you asked Laila the nuances of a particular project from six years ago, she could recall that better than I would be able to recall that.

Isabel Berwick
So how far off is this as a future where every executive has a digital twin or a helper?

Iliana Oris Valiente
Well, based on the research that Accenture has done, three quarters of knowledge workers are projected to use co-pilots every day by 2026.

Isabel Berwick
So not everyone will have a digital twin, but everyone will be using co-pilot and other kinds of augmented help. And just the last question, I presume Laila is getting smoother as she gets older, essentially. You know, how important is that a humanising of Laila or is there a limit to that, do you think?

Iliana Oris Valiente
We are in an era where technology is becoming a lot more human by design. Laila is just an example of that broader trend. And it’s true with every passing week there are improvements made so she can now recognise hand gestures. So if you’re in the middle of talking to her and she’s responding and you want her to stop, you simply raise your hand and you gesture to pause and she will wrap up her remarks. You can wave at her, and she will use that as an opportunity to start engaging in a conversation. So this humanisation of technology is going to open up so much more interactivity, because it no longer requires a complex degree in computer science in order to be able to interact and engage with these technologies.

Isabel Berwick
Perhaps we’ll have Laila on next time. But Iliana, thank you so much for joining us.

Iliana Oris Valiente
Thank you so much for having me. And maybe Laila will join you.

[MUSIC PLAYING]

Isabel Berwick
Iliana makes a digital twin sound pretty useful, and she’s bullish on how soon we’ll all be using them. But I wanted a second opinion on some of what she said. To get it I sat down with Madhumita Murgia, the FT’s artificial intelligence editor, and I asked her how new is the idea of a digital twin?

Madhumita Murgia
Well, you know, this idea of the digital twin has been around for many years since I’ve been covering it. But I think because of what’s happened with ChatGPT, which is the chat bot that you can now sort of converse with sort of twins and avatars, it feels a lot more kind of close. And so I’m not surprised, but I think the fact that she’s using it in her daily work is quite significant.

Isabel Berwick
Yeah. So Iliana reckons that three quarters of knowledge workers will be using a digital assistant by 2026. That seems soon. What does that sound like to you?

Madhumita Murgia
Well, that’s certainly what the big tech companies would have us believe. The last couple of weeks, we’ve seen advances and announcements from both Google and OpenAI and Microsoft as well. Actually, there are versions of this, you know, agents, as they call them, essentially digital assistants that can help you navigate your email or, you know, your meetings and things like that. And so suddenly a lot of money is being invested into it and is riding on it being successful. But I think from a sort of more behavioural perspective, it’s going to take a lot more behavioural change trust for us to have these twins or assistants or whatever do our jobs for us. Because they’re not fully accurate. You know, they’re riddled with errors. This is inherent, actually, to how generative AI systems behave. They make things up. And so, you know, in some situations that doesn’t matter very much. But actually sometimes that could be pretty crucial. So I feel like for us to navigate that line and say, is it OK for me to have a digital twin take notes in a crucial meeting that I’ve got to act on? That’s a big jump from a kind of human perspective. So I’m not convinced that we’re all just going to sort of divest responsibility to a digital twin because we get something out of having human interaction, right?

Isabel Berwick
She said that colleagues don’t feel upset that she sends the digital twin to a meeting. Surely that brings a two-tier meeting system where some meetings are worth a person, and then you realise your meeting’s only worth a digital twin. How’s that going to make you feel?

Madhumita Murgia
Yeah, I thought that was a really interesting question, actually, because the general view from the tech world is, of course, this is going to be better. It’s an improvement on what we have today. But as you say, just previously she’s talked about how does that work that she’d rather spend her time doing and, and sort of lower value work that she would, you know, assign to her digital twin. So automatically you’re saying that the places you’re sending your assistant to are places that are less important, less crucial. And this is true of all AI systems in other domains as well, right? Where you might have, you know, diagnostic systems where it’s good enough to do it with AI, but you only if you pay more, you might get a human doctor instead. And so you’re creating these two-class systems of sort of automated versus human. And, you know, I find it difficult to imagine that it will be that widespread because of the value. And, you know, that sort of respect we give one another. When you say, I’m coming to this meeting, I’m going to share my ideas with you and instead sending sort of a bot. It’s hard to imagine what that would do to a workplace.

Isabel Berwick
So now I’ve got a vision of a load of bots having a meeting with each other, making stuff up. That’s just really . . .

Madhumita Murgia
Yeah, they’ll just talk to each other. We’ll just stop having any humans at all.

Isabel Berwick
So Iliana’s twin looks very much like her. You know, she went to great lengths. She did body scans. The hair is correct. Is that common? I’m slightly freaked out by the idea of having a twin that looked like me. Maybe an even better version of me.

Madhumita Murgia
Yeah, it’s interesting that this is also a visual representation of her, right? Because what the Microsofts, Googles and OpenAI’s of the world are saying is just you’d have an AI assistant, whereas having an avatar that speaks like you and you know, it’s far more invasive in a way. What’s interesting is how people respond to that. People will respond to it as you, and they expect that what the AI says is what you’re saying and therefore trusted in the same way or treated in the same way. But it’s not, you know, it might be trained on her interactions and things, but ultimately it’s a predictive engine. So I think there are things that could be really useful for, like note taking, you know, something where it’s kind of digesting information, summarising it. You know, she gave an example of, at an event, it could pull up statistics. So she’s speaking out loud, you know, quite helpful for that kind of thing. But if it’s about relationship building or getting something done where you need to kind of bring somebody on side or motivate somebody to do something. You know, I think people often anthropomorphise these things without realising it, and we forget that it is ultimately just technology working inside. It’s not Iliana.

Isabel Berwick
Has anyone been thinking about the unintended consequences of using AI at work? You know, might we lose the knack for doing some simple things? Or does it matter if we do?

Madhumita Murgia
I haven’t seen anything myself, and I’d be really interested to read, as you say, about how it’s going to change our ability to do simple things. So, you know, something that struck me when Iliana was talking about I have to say the same things again and again about from previous projects. When people come and say, you know, what are your learnings about something? And so it would be easier for her digital twin to do. It did make me think, you know, if she stops doing that, what happens to her ability to collate that information, to pattern spot, to come up with new ideas? Because so much of, you know, my job, which is thinking back to what I’ve written before and connecting the dots, helps me to come up with new things or find gaps in that, you know? But if I wasn’t doing that, do I lose, you know, those muscles to join the dots or find the unexpected thing because I’m not explaining it to other people. So I think, you know, I would love to do more and read more on how it’s going to change our own ability to do what we do best at work.

Isabel Berwick
I’m guessing. But I’ve seen some stats on this that, you know, lots of people, they have this kind of procedural AI introduced at work. They might try it once, but then it drops off. What would it take for the AI to really catch on among the late adopters?

Madhumita Murgia
Well, yeah, I think this is the this is the sort of, key question, right, for any new technology as it enters a workplace. What’s the killer app as, as they call it in Silicon Valley. And, I think the bet now is it’s going to be digital assistants. That’s going to be what makes us as consumers really engage with this. Because, you know, now if I think about doing my expenses or wanting to figure out when I can enrol for private healthcare, what’s the window? Let me not miss that window. If I’m supposed to send a memo to somebody about something, all of these things, I can ask the help of an assistant. And if it’s easy enough to use which in this case, it should literally just be, I can write to it in plain English and say, do this. Can you bring me this? I think that’s what, you know, could cause a real shift in how we interact with the web and, and therefore with each other in a workplace, too. But, you know, I think it will take another year before we can see how, you know, how much they get integrated into our working lives. But we can see it’s happening already.

Isabel Berwick
That’s not long, though. A year is it.

Madhumita Murgia
You know, Microsoft has already introduced these things through its suite of productivity tools which touch billions. Similarly, you know, we’ll see this with Google. These are big consumer brands. So I think it is just a matter of people getting used to how it can be most helpful to them. And I think we’ll gradually adapt to, as we did with sort of the web and the mobile kind of app revolution as well.

Isabel Berwick
Thank you Madhu.

Madhumita Murgia
Thank you.

Isabel Berwick
Madhu’s views on AI assistants were perhaps more sober than Iliana’s. But even if we don’t all have full on digital twins in a few years, it does seem likely the AI helpers will take on some of our more mundane and boring work. And I, for one, am excited about that. Iliana said her digital twin allows her to focus her time on more human tasks. Madhu noted that really crucial meetings will still require real people, so I’m not exactly rushing to announce the end of work as we know it.

On next week’s Working It, I’ll be speaking to an entrepreneur who claims to have invented the world’s first AI job interviewer.

AI interviewer
Hey, I’m your AI interviewer how are you?

I’m good. How are you?

Great. Let’s jump into it.

Isabel Berwick
The job application process is a mess. Could artificial intelligence help? Join us next week to find out.

This episode of Working It was produced by Mischa Frankl-Duval and Tamara Komornick and mixed by Simon Panayi. The executive producer is Manuela Saragosa and Cheryl Brumley is the FT’s global head of audio. Thanks for listening.

Continue Reading