Tines Co-Founder Thomas Kinsella on Governing the Rise of Wild Code

0:00So you make it so that the easier path, the more productive path to using AI and connecting to your MCP servers, that's the path that's governed, that's the path that has visibility and that's the path that has control. And then people can also take a lot of credit and take a lot of joy from shipping those because you're able to ship them across the enterprise very, very quickly.

0:22Peter High: Welcome to Technovation. I'm Peter High, president of Metastrategy. My guest today is Thomas Kinsella. Thomas is the co-founder and chief operating officer of Tynes, an intelligent workflow platform that helps organizations automate complex and repetitive work securely. Founded in Dublin, Ireland in 2018, Tynes began by helping security teams automate manual processes and has since expanded more broadly across security and information technology. The company serves customers ranging from startups to global enterprises, including Mars, SAP, Booking.com, McKesson, and Logitech, and is approaching$100 million in annual revenue.

0:58Peter High: Prior to co-founding Tynes, Thomas spent roughly a decade in information security, including roles at eBay and at DocuSign. His frustration with how difficult it was to automate repetitive security work helped inspire Tynes and its original low-code, no-code approach to building workflows. Today, the company is expanding that vision for the AI era, helping organizations combine deterministic automation, AI agents, and humans within secure governed workflows. The company is also tackling the emerging challenge of wild code as AI makes it possible for people across an enterprise to build software and automations at unprecedented scale.

1:34Peter High: In today's conversation, we'll discuss the evolution of Tynes, how enterprises should determine when to use traditional automation versus AI, why humans remain an important part of intelligent workflows, and how CIOs can govern a rapidly expanding population of AI builders, and what Thomas has learned building a global technology company from Ireland. Thomas, welcome to Technovation. It's great to speak with you today. Thanks so much for having me, Peter. I'm excited about it. I am as well. Yeah, I'm really looking forward to this conversation. So many interesting topics for us to tackle as that introduction would certainly highlight and looking forward to delving into a variety of different directions with you.

2:29Peter High: the tone, the emotion, the rhythm of real speech, and wraps them in the full stack you need to actually deploy. Knowledge integrations, workflow tooling, multilingual support, and enterprise-grade reliability. Build human-level conversational experiences into your products, services, and operations. Head to 11labs.io backslash technovation to get started for free. And now on to the interview. You spent most of the past decade working on Tynes as a co-founder of the business. And I wonder, what problem did you repeatedly encounter prior to founding the company that ultimately led to you to conclude that there was an opportunity to do something and to start a business such as the one you did?

3:16yeah so uh i'm a security engineer by by trade uh but myself and my co-founder own uh we spent as you said a decade or so working in in information security in various different roles in some large organizations and we faced a challenge that is now faced by a whole load of different companies but when you're working in cutting edge security it was a little bit more niche that's like there's way too much work for security professionals there's not enough staff have to hire, your budget may grow, but you don't necessarily have everything that you want on your plate. And there's just way too many tools to connect.

3:52And that like too much work, not enough staff was combined with the sense of inevitability that some incident or other was going to happen. And this is a challenge or was a challenge that was faced by lots and lots of security teams. And we asked a lot of people like, hey, what are you doing? All of our peers in various different organizations. And they all came back with some answer of like, we're trying to throw more money at it or we're trying to automate and we tried to automate so Owen and I we like tested about 10 different automation tools out there and we frankly just didn't like them we thought they were way too complex and we said we could do better so we rolled the dice and we started Tynes to be the platform that we wished we had when we were working in security a simple powerful automation platform that allowed those people that were familiar with the work to automate their own task but also to be the company that we wish we could have worked with and worked for as well And so we're still at that mission today.

4:45We're still trying to allow people to build, run, and monitor their most important workflows.

4:49Peter High: And it's been a lot of fun. Yeah, very interesting. You allude to this, as I did in the introduction. You spent the first seven or eight years primarily helping security teams build low-code, no-code workflows. Now you're positioning the company much more broadly as an intelligent workflow platform. What changed and how different is the opportunity you see today from the company you originally founded? Yeah, completely. So the original aim, we always knew the platform should be agnostic. We always knew that we wanted to empower people to build whatever they wanted. We knew that it didn't really matter what they were automating.

5:23The platform was going to be able to help you automate that. But what we realized was that the people that were technical enough were usually the people that were automating. So people that were familiar with APIs, people that were familiar with JSON, people that knew how to secure something. they were the people that naturally gravitated to what we call time stories our original product and we always had some it teams that were using the platform when you always had some other like teams across the enterprise but it was always that little bit more challenging to build so we pioneered a uh like a low-code no-code platform that allowed everybody to drag and drop uh like tools onto a canvas that was very powerful what changed in the last like 18 months is that AI just made it simple for anybody to build.

6:08So when we think about the challenge that we set out to solve, to build a platform so that people could trust us to build, run, and monitor their most important workflows, the build became commoditized. People were using cloud code. People were using codecs. And now we've got another challenge. And that challenge is what we call wild code. It's that invisible, ungoverned, like unknown agents, automation scripts, workflows being created across the team or across teams and across the enterprise. But there's no central owner. There's no central governance. There's no central visibility. So we said, actually, like we have a decade of experience in building, running and monitoring these workflows.

6:49We think we can make a platform that allows like CIOs, allows IT teams, allows security teams, like centrally managed, but allow those people that just want to build to build wherever they can. and allow us to take over that management layer for them. So we built Tynes 3B as a platform to help solve that problem, but it's a different problem. Same kind of solution, same flavor, same experience, but the problem of wild code is every enterprise is experiencing it. So we've really branched far beyond that original kind of ideal customer.

7:21Peter High: I want to delve further into this notion of wild code. So interesting, Thomas. You know, CIOs and their equivalents want thousands of employees experimenting and building with AI naturally. But they also need to know what's being built and the data it accesses, what it costs, whether it's secure. So I wonder, how do you think further about, if we can delve a bit deeper than you just did, how do you think about democratizing software creation without democratizing the technology risk, if you will? yeah so the first the first thing i'll say is that like uh like democratizing is a is a super noble aim and you should absolutely be trying to get it into everybody's hands but you don't necessarily want every single person just going ahead and building because they're going to build the exact same thing multiple different times so even the idea of saying hey we want everybody to build but we don't want everybody building the same thing is you know is part of the uh is part of the challenge but the goal should be again build people are going to build wherever they want So you have to allow people to build using Cloud Code or using whatever tool they want or using Times 3B.

8:24The secure part in the running is, for us, isolated containers, making sure that there's, as you said earlier, shared secrets that are not accessible, able to be used but not accessible to anybody who's doing that building. Isolated steps, centrally stored source codes that you have visibility over it. And then the monitoring part is what we think is so important for those CIOs and for those IT and security teams. So that's like understanding, first of all, everything that's running across the environment. More than that, it's understanding what it's spending. It's understanding how much it's run.

9:03It's understanding your return on investment. And then the next step is turning what traditionally would be like a tech debt problem. Like, you know, 10 people have built 10 different workflows or 100 people have built 100 different workflows. and turning that into a solution for the rest of the enterprise. So the challenge is if somebody's built one thing, how do I share that to 50 other people securely? And how do I make those services that somebody has shipped or somebody has built, how do I make that a service for other people on the team? And then I can go into this much more. When something breaks, you should be able to auto-suggest a fix.

9:36When something improves, we should be saying, hey, maybe you want to codify that into your workflow. But it's really trying to build up that secure environment so that anybody can ship. But you don't necessarily want to ship absolutely everything. You want to make sure that what's visible and what's used is actually working for the enterprise and is returning on your investment.

10:01Peter High: For time immemorial, it certainly seems, as somebody who's worked with CIOs for three decades now, I worry was the notion of shadow IT, that people were buying IT solutions without the knowledge of the IT team and the CIO. Now, with Cloud Code, Codex, other AI development tools, the notion of shadow AI is something that's real. And you're alluding to that perhaps without yet using the terminology. And so there's this risk of people building tools in ways that the IT organization isn't governing appropriately, that the company is not governing appropriately. And so I wonder, talk a little bit about overcoming the idea of shadow AI and making sure that the risks associated with that are not something to proliferate.

10:44Yeah, so the reason Shadow AI exists, the same way Shadow IT exists, is that people want to build and people suddenly feel empowered to build. They suddenly feel empowered to use AI. What you don't want to do is there's certain organizations that have certain risk profiles that may need to say no. I don't think you can go down the restrict access completely approach. But I also don't think you can go down the give access to absolutely everybody for absolutely everything approach. which there has to be that governed route. So the way to give access is to allow people to use the tools that they have, but make it better.

11:22This is the paved path approach that a lot of engineering teams take. So that instead of having to go out and get a credential for Salesforce, or instead of having to go out and find access to Snowflake or whatever information you want, you make it so that that access already exists in a central place, and then they can ship securely to the rest of their teammates in a central place. And if they need access to find out how to access somebody's Google calendar or how to access NetSuite, that that information is also provided. So you make it so that the easier path, the more productive path to using AI and connecting to your MCP servers, that's the path that's governed, that's the path that has visibility and that's the path that has control.

12:05And then people can also take a lot of credit and take a lot of joy from shipping those because you're able to ship them across the enterprise very, very quickly. But it's, yeah, making, I think that paved path is what we really try to push that instead of making it, yeah, a shadow AI problem, we're saying actually there's a better way and a more secure way and a faster way. And if you make something better, faster, more secure, people are going to use it.

12:31Peter High: Yeah, fascinating. I mentioned in the introduction, Thomas, the remarkable progress that you all have made. A hundred million dollars in annual revenue, a goal of getting to 125 very shortly, with customers ranging from startups now to much more established scaled companies. I mentioned some of them in the intro as well, companies as large as Mars and McKesson and SAP, among many others. As you move from cloud-native technology forward customers into much more traditional enterprises, what have you learned about how their automation needs differ? Yeah, so they definitely have some more governance controls that are that are necessary their risk tolerance is a little bit uh is a little bit lower uh but actually what we found i was very nervous when we started supporting some of these like much larger organizations because like they are just that little bit uh yeah a little bit a little bit slower a little bit more deliberate but i love that and so what we learned is that like a lot of companies that are like cloud native they're much happier to move faster but they also shoot from the hip a little bit.

13:35And that's great, especially when you're selling software, you're like, okay, you're going to make a decision quickly. The traditional companies are those, the phrase we use is like companies that sell things that rust, they move a little bit more deliberately, but they start with the problem that they want to solve, which is kind of how we've actually started trying to approach bringing value to some other companies. So instead of trying to say, hey, I have an AI transformation initiative, they're like, okay, what is the specific business problem that I'm trying to solve. And then you can decide whether or not AI is the, or using an agent is the most appropriate way.

14:12Using a human being is the most appropriate way. Using a automation is the most appropriate way or workflow. And what you'll see is that typically it's a combination of all three of those. It's not just like one that you should be using an agent for one portion and a human being for another portion. but working with those large organizations they'll come with a project management team they'll come with like set business goals and those set business goals make it so much easier for me to say okay uh yeah mars you want to reduce your alerts by 80 that is not going to be just you know throwing an agent at this problem that's going to be deeply understanding what that process is what the individual workflows are and then we can figure out which part of the you know which part of the stack we should be focusing on which part of your workflow you should be automating with the deterministic part and then which part the human beings still need oversight so yeah i think it's almost uh it's a little bit weird but as we've matured we're we're almost more comfortable like we see much we don't see much higher growth rates but we see like higher n or or we see like yeah less like less churn more satisfaction not necessarily faster uh yeah faster expansion across the enterprise, but more deliberate expansion across the enterprise.

15:25And that's exactly what you want. You want those mature partners. So it's been a lot of fun working with a lot of those large organizations and people in the defense space as well. It's been a real learning curve.

15:35Peter High: I can only imagine. Yeah, fascinating. And again, congratulations on the progress you've made. And now for a word from our partner, Retool, and the company's co-founder and chief executive officer, David Hsu. Every CIO I talk to is feeling the same thing right now. Their teams are building AI apps faster than anyone can track. It looks like progress, but if you're the one who owns security and compliance, it's chaos. Software is running all across your business that nobody reviewed and nobody can see. And oftentimes, it's on the public internet, accessible to anybody. With AI writing more code than ever before, you need a way to govern all of it, to manage it centrally.

16:12That's where Retool comes in. your team can generate all the code they want and deploys to one central platform. Every app that is generated is automatically behind single sign-on, can be hosted on-premise, and is managed all centrally by you. You can control who can connect to what data sources, such as Salesforce, Workday, and other databases, as well as who can access which app. This way, anyone in your company can go and build applications when you can assure that everything is safe, governed, and that your data never ends up on the public internet. Deploy applications safely and securely with Retool.

16:53Learn more at retool.com slash technovation.

16:59Peter High: And now back to the interview. One of the things I've found very refreshing about getting to know you, Thomas, is, you know, Many of your peers who run AI native organizations want AI to be the solution kind of most of the time to the task that the customers might wish to perform. But you've made that point, as you've alluded to, that if someone performs, for example, the same task 100 times a day and the correct action is known, there may be little reason to ask an LLM to decide what to do each time. you know so I wonder I'd love your perspective are companies currently overusing AI for problems that conventional automation can solve more cheaply reliably predictably yeah definitely like and it's because it's so easy to build with AI it's so easy to build an agent and give it access to like certain tools ship it and now all of a sudden you're solving that problem but frequently when you're solving that problem you're yeah you're using an agent so it's being like Tokens are being spent every single time and you're kind of lighting money on fire and just gently whispering it in the direction of OpenAI or Anthropic or whatever other model provider you want to send your money to.

18:09When we were trying to think about this, we took a step back and we tried to think about the history of workflows. And the very first type of workflow was a human being following a checklist or an SOP, like a human saying, okay, this is what you should do in this situation. When tools like, you know, code or like Python came around, people started automating those workflows. But very quickly, they realized that the best workflows, they didn't just use like deterministic logic and code. They actually combined deterministic logic with like human beings in the loop. And that like human in the loop was so much more powerful because you had the power of extremely fast, extremely cheap automation and like human judgment.

18:56When you think about then when to use AI, AI is incredible. It's so powerful. It's so good at what it does. And you're able to provide it with tons of business context to help it make decisions. But ultimately, if you know a particular way of doing things, you should not be using an agent to do that because there's a much cheaper, much more faster, reliable deterministic method rather than a probabilistic expensive method of using an agent. Similarly, like if you know there's a problem that's actually like a little bit fuzzy, but you know the rough outcome that you want, an agent is actually substantially better than a deterministic logic at that.

19:38And then if you've got a judgment call to isolate the CEO's laptop or to, you know, put something in a board deck, you absolutely should not be using an agent or deterministic automation. You should be using a human. And our view is that like the future is a combination of all those three, but it's probably like the, on the spectrum, people have very much verged like over towards using agents for a lot of things. And that's a lot of the time why we see those, you know, people say, oh, we haven't seen the return on investment in agents because they're exceptionally good, but you're not necessarily using them for, for the right thing every single time.

20:10And so, yeah, we encourage you to use all three in your workflows rather than just using one. If you're just using one, you're setting yourself up for maybe solving the problem, but much more expensive than it needs to be.

20:22Peter High: So if deterministic logic, AI agents, and humans each have different strengths, what does the ideal intelligent workflow look like? It's a combination of all three. And you need to be able to ask yourself, like, hey, how predictable is this task? What's my tolerance for risk or error? How fast does it need to be? how much money am I prepared to spend on it? And if you ask yourself those questions, you'll figure out at that point in time, whether or not you should be using an agent, whether or not you should be using deterministic logic or whether or not you should be using a human. So if you've got something that like has to be correct all the time, and you know, there's a correct way of doing it.

20:58You should be using deterministic logic. If you've got something that like, you know, is, has to be done exceptionally quickly, you should probably be using deterministic logic. If you've got something that needs like company specific judgment, you should probably be using a human being to make that decision. But if you've got something that like there's very broadly a right way of doing it, or it's an unknown process that you may not have seen before, agents are incredibly good at doing that. And you should be using an agent. And it's at each point in the workflow, like you should be saying, okay, should I be using an agent?

21:32Should I be using deterministic logic? And our view is actually that like workflows evolve so workflows it's not your traditional like checklist that used to exist 10 years ago or 10 years ago like 50 years ago 100 years ago and it's certainly not like you know deterministic logic that like was static for several months or several years as your workflows continue so should your like as as you as you run them several times so your your logic should like should evolve so let's say you know you've got a process to uh like onboard an employee and you're saying hey i'm going to give them access to these 10 tools that's perfect if every single time you know that employee is in i don't know the finance department they come back and they request access for another tool and a human being has to be in that loop and they press go or an agent says yes it seems likely that they should have access to that tool and the agent says I approve access, that's now something that you know should happen every single time.

22:37And you should no longer be including the human and you should no longer be including the agent in that part of the workflow. You should just say, hey, I've looked at the last 10 times this has run or the last 100 times this has run and update your deterministic logic. And the next step in the future is actually like looking at your tickets, looking at your SOPs, looking at what actually happens and what decisions humans have made and saying, can I codify this into those workflows? So, like, that's a very simple example, but, like, workflows are alive. They've always been alive. You've always, like, tweaked them.

23:11Now, what you'll see a lot of people doing is they'll say, oh, I'm going to, you know, prompt my agent to do this every single time. That's, for me, that's the wrong approach. If you're prompting an agent to do it every single time, you know that's the correct way of doing it. You should have deterministic logic do that. And if you've got human in the loop every single time and a human does the same thing, you should be removing the human from the loop. So the workflows are alive. And our view is that, again, most workflows will involve an element of all three. Not all will. But you should be prepared to see them evolve over time and grow.

23:44And that's what's really exciting about it is working with some of these organizations and seeing their workflows grow.

23:49Peter High: Does the human intervention ultimately become knowledge that can be encoded back into the workflow, Thomas? yeah sometimes but not always uh like like in in in many ways yeah if like a human is making a judgment call to like give access to something or spend money on something or uh yeah like take a security decision or even like you know uh grant approval like there's a there's a business logic that can like that you can try to extract out of the human but the human has so much context on like why they might or might not want to do something. So oftentimes the underlying exact question can easily be codified.

24:27But the rationale for somebody saying, yeah, I think that's okay. Like the fuzziness there is really hard to codify. But what we see is that if somebody does something, yeah, like 50 times, that's very, very clear. The harder part is when like 50 % of the time they do, 50 % of the time they don't, that's where you really need to dive that little bit deeper and scratch under the surface and be like, hey, why is this happening? What is going on? And if it's worth it, if they're spending enough time on it, you can codify that. And agents are exceptionally good at, if you have a chat with an agent, dragging that out of you and figuring out that logic.

25:05But it's not all the time that it's able to be codified. But yeah, the fuzzy logic is the hard part.

25:11Peter High: Yeah, makes sense. You mentioned earlier that we might return to the notion of when something breaks, develop auto, an ability to auto suggest a fix. Talk more about the method, the best method to do so. yeah so so like when you ship when you put a workflow into production the challenge with you know uh what we'd call like tech debt is that you know people don't really understand what's you know what's in uh what's in production and they will like frequently be relying on something that's like mission critical to their company or mission critical to their team and it will break and the challenge is that if that's running on some person's laptop or it's running in a cloud you can't actually see like both what the process is supposed to be or the fact that it's broken.

26:01You just get a ping being told, hey, I think this tool is down. And if you've worked in operations for any period of time, you'll definitely be gotten a ping saying this tool is down. When you ship something with like Tynes 3B or a bunch of other platforms out there, the monitoring part is just as important. So what you're able to see is you're able to see the like the workflow that's running, but you're also able to see the output and the expected output. So the output could be, yes, you know, approval granted to X, or it could be approval denied. But if you get like a 404, so the authorization has lapsed, or you get like a malformed response, it's like, oh, that's very weird.

26:42That shouldn't have happened. What typically happens then is, you know, again, the user on the other end will say, my workflow is broken, or I'm not able to do this. when you ship something in times 3b that monitoring part says okay i know the expected output was a grant at 200 in api terms or 201 instead we received a 404 it can now spin up a an investigation into what has happened and it can then auto write new code to fix or to investigate and then fix that problem and tell you hey peter it looks like this workflow broke do you want to ship a fix into production and i can say yes and that's like that i suppose that monitoring part to tell you hey this is like this is you know this is a like yeah this is broken typically this is also like it can happen when you ship new codes this is like the concept of unit tests and then with agents uh there can be like configuration drift as you ship a new a new um like model into production for example that like something can like the agent behaves slightly differently So as a result, the output can be a little bit different.

27:47So what you want is you want a bunch of tests in your workflow. So it's not just the API call, but the output from certain parts of that workflow. You want to be able to test and say, okay, this is still working or this isn't working. And if it isn't working, you need to alert and potentially auto fix. And then again, that next step is what I was talking about. Is that like auto improve and auto expand that workflow as it goes. So, yeah, that's the idea of one of the parts of monitoring. But it allows you to, I suppose, move from that wild code to a much more governed state.

28:27Peter High: You and your co-founder Owen Hinchy built Tynes in Dublin at a time when there were relatively few examples of Irish startups becoming global software companies, Thomas. That would seem daunting, frankly, from the outside here. What was your inspiration to become a trailblazer? I don't think we set out to be like a trailblazer necessarily. We set out to solve the problem that we wish we had. But we really wanted to, we knew that there was an opportunity. So we were both in DocuSign when DocuSign was growing fast. And we were both in eBay when, yeah, eBay was growing fast. And DocuSign went public while I was still there, which was a really exciting time.

29:06And we also got to work with a bunch of exceptional startups while we were there. So, and yeah, some of which got acquired, some of which, yeah, like went on to become like enormous, enormous companies. And it was hard and we saw the hard work. But when we were both like very privileged, when we like had the idea of times said, okay, there's an opportunity to solve a problem here. And we'll see if we can do it. We didn't set out to be a$10 billion company or$100 billion company. We did set out knowing that we could be, but the goal was very much to build a product that we wished we had. Building it from Dublin was a motivation that I wanted to know that it can be done from Dublin.

29:53There weren't too many people that were, as you say, shipping and building impressive software companies. That flywheel had kind of just about started when like the likes of, you know, Google or Meta had their operations in Dublin and then a bunch of smaller companies. I say smaller, but like the HQ of Salesforce or the HQ of LinkedIn or the HQ of like MongoDB or HubSpot, like they created their, you know, their European or their EMEA HQ in Dublin. That allowed people to start seeing the like the process to setting up an organization. So, you know, when to hire your first salespeople, when to hire your first like CS people, when to like, you know, in-source recruitment, when to bring on a finance team, all that sort of stuff.

30:39And some of that is centralized in the US, but that really allowed you to see, OK, this is the process. And it's no surprise that like OWN, for example, like OWN was the very first employee of DocuSign in Dublin. and so when we started like times we said okay we want to ship this from dublin we want to see can we like can we build this product i think if i'm completely honest we probably would have ended up in the u.s if it wasn't for code so i'm saying all this with like a slight asterisk uh but yeah when we were like so when we took on some investment our investors were very much telling us hey the tried and tested method is to go over spend some time in the bay we were quite reluctant to do that we wanted to try build it from dublin and then when covid hit there was no option we basically had to stay had to stay put but like yeah we grew from like five to about 80 people during like lockdown and that just allowed us to say okay hey we can do this we hired you know like 50 people in the u.s during lockdown and we said okay this is you know this can be done and it's been hard, but it's been a lot of fun.

31:44And there's a ton of great startups that we've also been on that journey with and taken inspiration from Ireland as well.

31:51Peter High: Last time you and I spoke, Thomas, you observed that American companies have historically been more willing than European counterparts to take a risk on buying critical software from young startups. And I wonder as Ireland and Europe's technology ecosystems mature, is that gap closing? Yeah, I think it is. So yeah, definitely. Like the US and the West Coast in particular, there's a much more familiarity and much more willingness to take a risk on buying software from a smaller company. Like that's ingrained in their nature. And that's not nearly as or was nearly as ingrained. But what we've seen now is like there's in Europe, there is that like pace of change, pace of adoption.

32:35There's, you know, yeah, some exceptional startups, neobanks, some AI startups that have been that have been created. And there's much more familiarity with the tech ecosystem and respect for that tech ecosystem. So while we still see a faster adoption in the US, that gap is definitely the opposite of widening. It's definitely narrowing. And it's translated into, I suppose, over the last few years, much greater success in Europe. And then, yeah, as you kind of mentioned earlier, the likes of SAP, like the largest software company in, like outside of the US, like they're now a customer. And again, that would be the sort of thing that a couple of years ago, we just couldn't have, couldn't have hoped for.

33:22But that's also the case with, you know, European government institutions and like large startups in France and or large startups, large companies in France and the UK and the rest of the rest of Europe as well. So, yeah, there's much more familiarity and much more comfort with it. And I think there's also a little bit of, if I'm completely honest, there's a little bit of like native pride as well that they're like, hey, actually, we want to support European startups because we need to grow that ecosystem ourselves as well.

33:49Peter High: Fascinating. You've drawn inspiration from the Farmer's Almanac as you've thought about ways in which to convey what will come in the future. Talk a bit about that. I found that fascinating. Oh, yeah. So this time, I'm not going to take complete credit for this. This is our CEO, Owen. He's fantastic. He's an incredible leader, and he's a little bit eccentric, and he won't mind me saying that. So, yeah, the Farmer's Almanac is, for those familiar, it's like an annual publication that tells people, hey, here's how you should tend your crops, and here's how you should, like, when you should plant your beet.

34:25here's how you should like when you should you know bring in your cattle and like we're expecting it to be a much more wet year so as a result you should maybe do x y and z if it starts getting wet at this point in time and the whole idea was that in an era or in an age when it was really hard to know what to do there were some rules that you could follow and and some things that like you could have comfort in and we knew this time last year when we were starting like planning for like FY27, so calendar year 26 for us, we knew that like 2026 was going to be a very, very challenging year for the world.

35:04The SaaSpocalypse was, you know, just about starting AI labs. Then those frontier AI labs were just shipping software so fast and growing so quickly. And frankly, like as I alluded to earlier, like our job, build, run and monitor like workflows we realized that actually that building part was getting completely commoditized and that was going to be scary for a lot of the team and suddenly we started seeing like Claude Code pop up in like you know notes from calls and things like that so we said okay we're going to like try to use this farmer's almanac analogy and we built or we shipped a you know it's blurred so you're not going to see it but like it's a it's a little book and the little book is like telling us hey telling the team and we've published it online you can uh you can go check it out uh hey here's what we can control and here's what we know is going to happen next year and here's what we're going to do and here's what our plan is around it because i can't control what anthropic are going to do anthropic are an incredible partner by the way but i can't control what they're going to do and i can't control what their markets are going to say uh but i can control like how we show up i can't control how we treat our customers i can't control what our philosophy is around like treating our staff correctly.

36:18I can control how we use our data. And here's what our strategy is going to be. And we wrote it all down. And we said, hey, here's the goals that we're going to ship. And we said, we're going to ship 3B next year. We're going to hit$125 million in revenue next year. We're going to triple the number of workflows built on the platform next year. And we're on track for all, well, we've shipped 3B, but we're on track for all of those. And it's just trying to, we still use it in plenty of Teams goals where there's so much information and the pace of just news coming at you is intimidating. It's like, I can't do any, I can't shield you from that, but I can tell you, we can help you control what you can control and what you can control is like these factors and this is our strategy around it.

36:59So it's a, yeah, it's been a lot of, it's a lot of fun. It was a lot more work than we thought it was going to be. It was a great idea. And then we put a huge amount of effort into it and we're like, ah, so it may not, It may be the one and only Tynes annual Almanac, but it's really been a good inspiration for the team. And just something that we keep on coming back to is actually this is today's market. It's so important to know what you can do and what we're not going to focus on.

37:27Peter High: That's really interesting. I mean, the fact that you include qualitative and very specific quantitative aspects to what you're describing there must be useful in sort of driving the organization in a common direction, hopefully with inspiration. Yeah, definitely. And like the method, the things that we had, like$125 million in AOR, like that's like absolutely like very specific. And, you know, our board key point to do it being like, hey, are you, and so far we're on track, which is amazing. But it's also like the other, some of the other goals we set, you know, so tripling the number of workflows, doubling the number of users using the platform.

38:05all of that is something that we're trying to like it's very aligned with both like our success our customer success like they're not metrics that can be fudged easily so you know the typical like hey we'll like say something and then we'll um you know what the instant you the instant you put in a metric is the instant it can be gained these really can't be gained they're they're objective metrics for you know who's using the platform what they're building what our revenue anyways um but it does yeah put big numbers uh picked up big put big numbers in front of the team and on the other side like people like like extreme challenges and when we decided to build 3b like we did so at the start of the year we had time stories our original platform time 3b we said hey this has to be like 100x easier 100 100x faster to build because we know we need to have more people using the platform.

39:00We know we need to have more workflows built on the platform. We know we need to hit this revenue goal. So like that allowed the team to be like, okay, it's not just a, you know, simple, like, Hey, this is like 10 % better, 20 % better. It has to be a complete change in how you do it. And it has to be so that, yeah, like Dave from finance is able to automate their workflow. Cause if it's not, we're not going to hit those goals.

39:27Peter High: Yeah, fair point. It's been well articulated. I wonder, Thomas, given these dynamic times, what do you read or listen to that is of influence to you? In order to stay current, beyond, of course, working with great colleagues and customers, which no doubt are part of the answer, what are resources that you leverage that you might recommend to others? Yeah, I think there's a few different things. Every day I try to read across the spectrum. so I try to read like the Wall Street Journal the New York Times the Financial Times that gives me like high level hey here's what's happening in business here's what's happening in politics and hopefully a little bit of here's what's happening in tech and yeah it tries to keep me out of my a little bit out of like the I suppose yeah my circle which can be definitely have like one or two opinions that are yeah that may not be may not be shared by everybody and the other thing that I read like all the time is is hacker news so news.y combinator that's for startups probably and tech probably the best like notice board that you can possibly go on or that you can possibly go on uh there's a ton of founders there's a ton of engineers with incredible experience and that like post there and if you want to find out about the latest models or you want to find out about you know quantum uh like post the post-quantum world you want to find out about like a new cancer drug that's being developed typically there'll be a thread that's discussed by professionals who know what they're talking about there and then there's a bunch of like niche podcasts so uh niche maybe not totally niche but uh you know podcasts in the security world where I still spend like a lot of my time podcasts in the startup world like 20 bc podcasts in the tech world like yourself and then like just you know history podcasts and a bunch of things that I try to read.

41:20But the thing that I'm seeing across my team now, they'll share a bunch of like sub stacks and things like that. But it's also that in order to like keep current, it's not really, it's no longer the case that you can just like listen to podcasts or like read a bunch of articles. You actually have to go in and get your hands dirty. Unfortunately, the barrier to entry has just come down so much you have to go and try the latest model you have to like you know get hands on keyboard with some like new software and that's how you're gonna figure out what's gonna change the world and that's what we what you're gonna that's how you're gonna like that's certainly for me but for a lot of people that's how they're learning and you know with and there's tons of amazing companies out there if you take one a week or one a month and just like take out a free trial or take out an explore edition or community edition and give it a whirl you'll learn so much so quickly and fortunately with the you know uh vc money coming in uh there's plenty of like plenty of ways of trying something for free and you know getting a bunch of tokens to play around with but yeah i just encourage anybody who's listening to genuinely just get your hands

42:30Peter High: dirty that's how you're going to learn yeah great great advice all uh thomas thank you so much for a great conversation. Fascinating to learn about the evolution of times, the Genesis story, but what it's become across the past, most of the past decade, the inspiration behind it, and your remarkable success. Congratulations. And thank you for a terrific conversation. Thank you so much. I really appreciate you having me on. Thanks for listening to Technovation. If you enjoyed this conversation, subscribe and leave a rating. It helps more leaders find our podcast. Thank you. And I look forward to seeing you next time.

43:09Thank you.