Citizen Developers: When Users Automate Their Own Jobs Away

0:00This show is brought to you by the Global Talent Co, a marketing leader's best friend in these times of budget cuts and efficient growth. We help marketing leaders find, hire, vet, and manage amazing marketing talent for 50 to 70 % less than their U.S. and European counterparts. To book a free consultation, visit globaltalent.co. Hey, everybody. Thank you for tuning in. My name is Andres Figueira, and I'll be your host for the day as I'm currently located in Buenos Aires, Argentina. This show is brought to you by The Global Talent Company, a company that is connecting top remote professionals from Argentina, South Africa, and all around the world with leading tech companies worldwide.

0:40So if your team is looking to hire all around world-class South American or South African talent, just head on over to theglobaltalent.co and see how we can help. And also, without further ado, today I am speaking with Adit Jain, the co-founder and CEO of Lina AI, a company that with its own proprietary large language model named WorkLM, is empowering enterprises worldwide to redefine how employees get work done, delivering a transformative impact on productivity and efficiency. So how's it going, Adit? It's a pleasure to have you here. Hey, man, thanks for having me. I know it's been amazing.

1:17No, I can only tell. I was really looking into your company yesterday and the days before the episode, and Lina AI seems pretty crazy. You guys have your own large language model, and it helps to serve more companies than just the regular user, right? Yeah, that's correct. So the quick introduction is, think of us like an accelerant to the agent AI journey for the enterprise. Typical use cases for us in large enterprise would be, you know, reduction in IT, HR finance, help us tickets. And then essentially the overall aim is to reduce the back office in large enterprises, right? So with Lina, a typical enterprise three-year journey to get rid of 10 to 12 % of costs.

2:02Wow. You know what I just said, entire costs, right? So it's like eliminating the back office using Lean I at the end of the day. Yeah. Wow. And just because I'm hearing that it's your own proprietary large language model, and how long did that take to set up, to teach it, to even tutor it? I don't even know how you create one of those. Oh, absolutely. So WorkLM is actually a family of models. And it also includes fine-tuned, you know, GPT 4.1, MixDot 1822B. We're actually fine-tuning in GPT 5 right now, Gemini 2.5 Pro. So we keep fine tuning all closed and open source models. See, we are not a model company.

2:39While we have models that are fine tuned for enterprise use cases, we are not a model company. And we don't intend to be because, you know, I strongly believe that the value is going to get accumulated in the application layer and not in the model layer. And then you can see that happening with token cost, both input and output, just crashing with every launch that these big labs make. Okay, wow. And what was the first inception of the idea or what was that aha moment that you had that you said, hey, I think I'm going to start making what Lina AI is today? Absolutely. Lina AI actually has a crazy backstory.

3:13So the three founders we've known in the last 15 years, we were together in the same undergrad, in the same dorm room in India, an institute called IIT Delhi. It's a top tech school in India. That's where our journey kind began we completed undergrad in 2015 and then we had job offers but we said hey let's try and build something of our own so we started a company called chatteron which is a competitor to ibm watson like it's a platform where you can build your own chatbots without writing code right interestingly that went and scaled from zero to over 30 000 chatbots created over a span of two years but only 20 of those 30 ,000 ever paid us what oh my god talk about bad product market fit right like 20 customers out of 30 ,000 paying you is a bad product market fit we realized we were too horizontal trying to be everything for everyone while being nothing for nobody and that's when we looked at those 20 customers in detail and realized that 16 of them were using us for internal IT HR finance chatbots now at that time we had zero professional experiences so we like okay what the hell is happening?

4:19And then when we went deeper, we realized as organizations grow, there's so many different systems, people, applications, departments. Finding the right information and getting your job done is a huge time suck. So that's the birth of Lina in late 2017, the vision of building a Jarvis or Ciri for the enterprise. That's how it came to Lina. Wow, that is so cool. It all started with chatbots and you noticed how they were actually being used rather than the whole variety of implementations that you could come up. That's crazy. And have you noticed or have you seen when you implement LENA in certain businesses or wherever that LENA is needed, when employees first interact with an AI work assistant, do you feel that they have a surprise?

5:02Do they get along well with it? Do they have the classical fear of, hey, this is going to take my job? I don't know. You know how people get a bit crazy sometimes around AI. But how has your experience been with releasing that product to the employees per se? Our vision is to remove the back office from companies, right? Okay. But we start with specific, like within that, we start with specific use cases in an enterprise, like customer like Coca-Cola, Sony, Puma, Vodafone, Nestle, P &G, Estee Lauder, some really large Fortune 500. They start with specific use cases around share finance, health test reduction, right?

5:35So these are basically, hey, I want to change my benefits. I want to change an address. I want a new laptop. I want access to Adobe Acrobat Pro or things like that. right so for these use cases you know it's a good axle rent from an enterprise perspective to be able to open the minds up of the people right so they absolutely love it because this is stuff that nobody really likes doing right like if you think about it andres nobody really like i don't like raising a ticket or going to somebody for asking a small thing disturbing them and then having to wait them for them to call back etc if i can get it done myself i love it right like if i can just speak to it with voice or I can chat with it.

6:13I just absolutely love that. Right. So that's exactly, you know, how people perceive it. We get crazy adoption, both on voice and on chat. And people love talking in their own languages. Think about like global organizations, which are in like 20 countries, 100 countries, et cetera. So they have to support languages and all the time they might not have, you know, IT teams or HR teams available all the time. So that's basically where this becomes so important that having an AI assistant, having an AI agent for all their problems is extremely well received by people. And then it also unlocks their thinking on what AI can be doing for them and how they can do it.

6:51So I think it's actually an amazing use case to land on because nobody really likes waiting to get answers in the enterprise. And on the other side too, if you think about it from an IT, HR or finance perspective, These are the type of tickets or questions that everybody hates. Nobody did undergrad, went to school to do troubleshooting, like level one troubleshooting for somebody else, right? At the end of the day. So that's what we believe in and we get great traction when we launch in customers. Wow. Yeah. And now that you mention it, I do agree with that in the fact that people at the beginning, they could be a little bit afraid of AI and what it can do and no one's going to take my job.

7:28But once they understand that it's a tool and that you can leverage it to make your job easier, that you can do other things that are more important that need to be prioritized, then the gears start turning and people are, OK, what can I use this for? How can I save my time with this? And it becomes more of an ally than an enemy as one could actually see it. I would also like to say that, you know, people keep saying that the AI is not going to take job and it's going to be an ally and going to make you productive. Yes, but it's also going to take jobs. Like I like to call a spade a spade so that you can actually prepare, right?

8:00Of course. It's the responsibility of large companies and people themselves to know what's going to happen and to prepare for that. So here's my take, right? Like I think in three years, people are underestimating what AI will do. And they're overestimating what AI can do this year. And they're underestimating what AI is going to do in three years. I think a lot of IC job roles that are running five to 10 business processes, 80 % of the day, they're going to go away. yeah right and that's just the truth like now let me tell you what the truth is and then you know as either a person who's doing that job right now or as an enterprise let's discuss what is possible to do for them three years later right maybe and some of these roles might not even take three years it might just be 12 months away or six months away so i like to call a spade a spade so you can actually try to solve the problem for for humanity at large instead of just hiding behind a curtain and say, hey, no, it's not going to take jobs.

8:52It is going to, it is already. And then for us as leaders to come together and see, you know, what's best for humanity next. Of course. Now, and I totally agree because it will take jobs, but I've been also thinking about it a lot. And I feel like the jobs it's going to take the most are those that the employee or the role can't even adapt to using AI. And then somebody else could take a job just because they do know how to speak to one of these large language models or whatever like that. But I do think that the more monotonous jobs and the ones that are more of like the back office that you guys are solving and that type of like standard ticketing that no one wants to answer.

9:29That was so funny to me when you just mentioned it, like nobody wants to take those tickets. It's so cool that Lena can actually help out with all that. And OK, I have another question for you. And this is one that I'm not sure if you guys have measurement to understand a company's adoption rate for Lena. Or have you noticed that there are any red flags that tell you a company isn't ready for AI? Is that something that you've seen? Absolutely. And it's all the time. So there are actually multiple readiness, you know, indicators for an enterprise. So for example, one would be, you know, when you interact with them, they're just not ready to have this conversation that, hey, yeah, the future is AI first, right?

10:07Like I go into organizations, I say, hey, you know, 10 to 12 % cost in three years. That's what you can reduce with lean AI. Like all of your back office gone and they're just not ready to hear that. And they're not ready to, you know, accept it and figure out what next, right? Like, Hey, it's great for the profit margin. So that's amazing. It's great for business. So that's of course true, but like, let's accept that it's going to displace jobs and figure out, Hey, how can we enable our people? How can we train them? Rockskill them? How can we upskill them to do more? Right. So they're just not ready.

10:38They're still ready to go under a rock and say, Hey, it's not going to take jobs. Right. We're not bold enough to go and have this conversation head on with our people. Because, see, that's the biggest indicator that there's something that's going to go wrong. Because here's how I think about it, right? If you don't have that conversation head on, if you don't educate your own team and your own people that, hey, the way AI is coming, there are going to be specific job roles and skill sets that will not be required. There are going to be new ones that form. New ones will get formed. New job roles will be there, right?

11:07But you've got to start preparing. Everybody's got to start taking this seriously, including the organizations and the people themselves. They have to be taking it seriously. If you are running in a job role like customer support, IT support, HR support, recruitment, a lot of these job roles are not going to remain the way they are today, right? Might be new ones that form. There might be new ones that form. But the eight things you do in a day-to-day, none of those, probably like seven out of those eight things are not going to be remaining in the next two to three years. So if you're not ready to have that hard conversation with your people and people themselves are not ready to kind of accept that in your organization, then that's a big indicator of whether it's going to be successful or not.

11:47And then right now I can even say I can even use that as an indicator to tell you whether that company is going to be successful or not in the next five years. Oh, my God. I love that. And it's totally true. It's totally true. and they need to be able to walk the walk if they say they want to use AI and they got to be able to speak honestly and transparently to their employees saying, as you mentioned, it will move and displace jobs, but it's going to create new ones. It's like, whatever was the customer support manager for this client, you're going to have to tutor the AI to make it talk to them and treat them as they are used to because it's not like just randomly speaking to an AI.

12:24They got to learn as well how to do their stuff. So I do believe there's like a pivot in how the jobs will be handled, But AI is going to help out with so much. Oh, my goodness. This is such a noble cause as well, because it's actually bringing companies that might stay, as you mentioned, under a rock into the light of the modern age that it's going to help them even more. And no. And have you noticed if simple tasks that people think are really easy, the AI has trouble handling them? Do you have any situations where a simple monotonous thing is actually really complex for the AI to handle? Yeah, sometimes, you know, there are edge cases.

13:02Every technology has edge cases, right? And things that don't work or work really, really well. The same is with the AI as well. I can give you an example of something that's pretty interesting. It was the case of the earlier models. The updated models have solved that pretty well. But if you would ask, okay, what time is it right now, right? They were just like, what day is it tomorrow? They were just like, there was something crazy going on and they'd always be like four or five days behind or five months behind. So something was off with the models earlier, but we'd solved that in WorkLM even earlier, but now the base models have improved and they now have access to dates, et cetera.

13:36But a lot of those issues are essentially related to being able to feed the data into the models. They know, okay, what's current? Because a bunch of these models are trained on data, which is like, let's say January 2025 cutoff dates, right? From a knowledge perspective. So they are not current. Themselves, they are not current, right? So then you have to feed them all the relevant information or current affairs in RAG. retrieval organization so that they can understand, okay, what's the truth of today versus what they already know, because they've not been trained till today. Oh, okay. Wow. So they haven't been trained until today.

14:11That's crazy. So we know everything else, but not the present moment. Okay. Wow. I had not seen it that way. That's really interesting. And what's the most unexpected way that you've seen employees for companies use AI in their daily work has there been like this weird thing i don't know employees analyzing their nutrition the lunches they do in a group or whatever i don't even know no our end users which are working for large companies they use leaner for a lot of different things right and i'll share a couple of ones which are like more productivity related and then maybe one funny one as well so a couple of them were like one qa actually used leaner to completely automate their job away right so we integrate with browser use.

14:54You can like, Lina's agents can actually, you know, just like a human interact with the web, right? So log in SAP and do things, log in into Salesforce and do things. So this two way person actually created an agent on top of Lina that would completely test the web application that he was responsible to test. Wow. Okay. That's the way to leverage AI. We didn't even know that our product could be used that way. That was amazing. And then that guy was promoted, by the way, in the customer. They work for a large CPG and that guy was promoted. So, you know, ingenuity right now and innovation is going to pay higher than ever before, like ever in history.

15:33Right. So we always have started to encourage that. Like we want employees of large companies that have access to Leaner to actually go and create their own agents as well. So that's number one. The second example would be, you know, the person who's actually created an agent. So their job was to do reconciliation between attendance and roster. So imagine this, if you're working in the aviation industry and you're ground staff, right? So you have a roster. You have to come from 9 a.m. to 9 p.m. And then your attendance is, let's say, shows that you came in at 12 p.m. and left at 12 a.m. So now this is a problem because you've got to be paid in seven days, every seven days.

16:11and now i have to see what's here because the 9 p.m to 12 a.m actually goes at a higher cost because it's overtime pay right so that's like that's three hours which is more expensive than the 9 a.m to 9 p.m so that 9 p.m to 12 a.m is more expensive now that company there's a 40 000 employee company which has like we provide ground services ground staff to large airlines and these guys they basically have 120 people team that just does this reconciliation every single day because payroll has to go out every seven days. So in this specific case, what they will do is they will go to the manager and say, hey, did you call this person?

16:47Did you ask this person to come in late and then go late? So if the answer is yes, then you got to pay them higher at a higher rate, right? So imagine you are one of those 120 people and what you do is you get a record 40 ,000 people and you keep going through their attendance and their roster and keep just messaging them or their manager and figuring these things out to get ready for payroll. And that's your job all day, every day. Sounds so hectic. That is a mind-boggling phrase. Now they use Lina agents and AI colleagues to completely automate this entire job away. And I'm telling you, while it was heartbreaking for the people, of course, because they did not have the role, the company handled it really well.

17:28And all the while that they were implementing Lina, they actually upskilled and cross-skilled or cross-trained the people to do other jobs in the organizations and move 80 % of them into different departments. right and that's the power so that's what i'm saying whenever i go into an organization and i say that this is going to happen and you say oh no no let's not talk about it you know we're not comfortable i know that company and the right way to do it is understand ai understand that this is coming and hit it head on educate your people and of course investing them invest in them to upskill and crosskill them as well so that's the two examples that stuck with me really well and then funny one.

18:04So somebody really asked our ENA agent for a trip of a, you know, an itinerary for a trip for a dog to Mars. So we did not answer that question because we don't answer beyond the enterprise. We keep it to enterprise. So we did not answer the question, but that one thing always, I don't know what they're thinking, you know, an itinerary of a trip for a dog to Mars. Okay. I love that. That's just like testing how far the AI can go. Like, can you get creative? Let's see. It's crazy. An itinerary for a dog in Mars. That is something that I'm not going to forget. Yeah. Oh my God. But talking honestly with the other experiences that you've had, this guy that used an agent to totally do his job for him, that is what I feel is it really adds value to AI and what you mentioned about being like, what's it called?

18:53A curious and being able to recreate and do this stuff with your job. I feel those people that you think are lazy, but they get AI to work for them. there's something there because that is a different type of smart. You're just trying to do something else instead of having to do your regular job. No, absolutely. I don't know if you heard Bill Gates. He said, I like to hire the lazy people because they figured out the easiest way to get the job done. It's true. They'll find the fastest way to do it because they just don't want to do it. So that'll probably work. Okay. Adit, if you could integrate Lina AI with three specific business applications that you feel would deliver the biggest impact?

19:30Do you have a set that have on the top of your head or do you just want Lina to integrate with everything that has to do with the workplace? So for large enterprise applications, everything. Because like right now we integrate with Workday, SAP, ServiceNow, Salesforce, Oracle, name it. We have an integration with all large enterprise applications. We're also partners with almost all of them, all the big ones we are partners with, right? Like SAP, Workday, we're partners with them, right? So at the end of the day, if you think about the future of work, and if you think about work, and if you think about like eliminating the back office, right, you'd realize that every single job role in there, or every single IC job, individual contributor job role in there, and the work that they do in a day is not in a single application.

20:11So they work across multiple applications, right? Like the work might start from, let's take an AR analyst, account receivable analyst, right? In the finance teams, their job starts by downloading the invoices that are passed due from SAP in the morning at 8am, So they download a list of 15, what's the data positive? Then they go into Salesforce to find the customer's phone number, and then they call them. And then they update notes in another system, like in high radius or something. So as you can see, they're using multiple applications to do their job. And that's why integrating across all large enterprise applications is such a big important thing for our AI colleagues that we build at Hina.

20:48and we have thousand plus enterprise applications available today already and we are constantly building more and more and more all the time we also allow our customers and partners so we have partnerships with si's like ibm and many others and we allow our customers and partners to constantly build more integrations and connectors on our platform too so that in case you know we are missing something which is very custom or something that you build yourself for your own company, like an application for your own company, you can just integrate it yourself as well. But out of box enterprise application integrations are core to our business and to our differentiation.

21:24Awesome. That's amazing. And Adit, would you be able to, I don't know if this is getting into deep or if it's easy to explain, but how is the process of teaching Lina or like integrate one as an individual into a certain platform that you need? Or how did you get it to become your agent, is this something like you tweak characteristics or you got to talk to and tell it, hey, once you get this result from this platform, I need you to go to this one and you send them the links. Like, could you give us a quick overview? No, absolutely. So the initial deployment, the end user, which is the employees, they don't do any configuration.

21:59They just use it, right? Initial configuration or the initial use cases are actually set up by IT teams, right? Nice. Yeah. So the IT teams would basically get a license of Lina and then set it up. I'll give you an example of a large top three financial services institute here in New York, like 100 ,000 people globally. So they had been trying to build their own agent for IT and HR over the last two years. So they wanted an agent that could integrate with Workday, ServiceNow, and SharePoint for knowledge. right they spent over one and a half years two years trying to build it with a team of around 50 odd people globally interestingly andres they came to lena and guess how much time they took to go live a day exactly we have out-of-box integrations with all these three and many more of course but we turned on the workday service now and sharepoint and they got all their knowledge into lena they got all the integration set up literally instantly and they literally were live within 24 hours and that's the impact of our platform think of us like an accelerator to the agent to get journey of large enterprises but then that's phase one right like once you're done with phase one you want to add and keep adding more and more applications you want to keep adding more and more use cases and then at some point you would also want to give lena the ability to create your own agents to every single employee as citizen developers to all of your company so a bunch of our customers have started giving access after setting it up correctly 200 use cases.

23:30After that, they allow employees to become citizen developers and continue to build their own agents and own use cases as well. So that's the power of the platform. That is so cool. Like I'm geeking out right now around everything that you're telling me that Lina does because it just sounds amazing. Do you guys cater to every single type of work industry? Are there some places where you know you shouldn't be entering? Like, I don't know, where you tell me about how easy it is to integrate and how it works. I just thought, Could Lina AI be applied to a hospital or is it HIPAA compliant or things like that?

24:03Do you guys handle that as well? Yeah, absolutely. We have SOC 2, HIPAA, IS-27K. We have GDPR. We have every possible certification from an information security and data privacy perspective. And we already have a bunch of hospitals working with us as well. So Christus Health is a customer of Lina. Houston Methodist is a customer. They have large healthcare organizations here in the US and many others as well. Nice. Wow, that is very cool because that is something that I've heard that is a bit of a fear that people have with these large language models that are more public, like ChatGPT. I'm not sure if Gemini, but it's like they're not sure of how private they actually are.

24:37And I know some businesses steer away from using them because they don't feel that their data is actually protected. So that's really cool. That's really cool. And OK, Adit, I have one more question before we reach the end of the podcast, because I know it's almost the end of the time. But as a CEO that you're building the future of the workplace with AI, what's one thing that you think every business leader should know before they implement it or they implement LENA or any other large language model in their organization? Yeah, I think it's a journey. So just like we have to start changing how you think, right?

25:11Like since the dawn of time or since a long, long time, we've had humans reporting into humans and working with humans, right? Like it's my colleagues, my managers, and then their managers, everyone's human. I think we have to accept the reality that AI colleagues will enter the workforce or already entered the workforce. And in the next 12 months, like we at Lena already have around six AI colleagues between our people. Like in our architecture, we already have six AI colleagues, right? By the end of the year, we scale that to over 50 AI colleagues. And by next year, we are looking to have more than 200, right?

25:45Now, large enterprises, of course, will be slower than that. So by the end of next year, they will have more than 20 AI colleagues. And I think people are just not able to understand that and be accepted. I think that's the change that needed at the top of organizations that, hey, the nature of org structures has already changed. And you will have AI colleagues reporting into humans and AI colleagues collaborating with human colleagues, right? That kind of fundamental shift in thinking needs to happen. and you know that's what we've been talking about to our customers and prospects and saying that hey you've got to start thinking of it this way while you're on this journey with lena because at the end of the day you know that's what we want to do in three years we want to give you 500 ai colleagues across the enterprise across different job roles in the back office so it's extremely important for you to understand this up front this also means just like humans ai colleagues or agentic ai will never be 100%, right?