3193: How AI-Powered Document Automation is Transforming Business Productivity

[00:00:04] What does the future of AI-Powered Document Generation look like in 2025? And as businesses race towards automation and efficiency, are they fully considering the risks that come with it as they dive forward headfirst? Yes, productivity is the buzzword of the moment, along with AI, but not when it comes at the expense of compliance and security.

[00:00:29] And apologies as the ex-IT guy that has to bring that up, but today I'm going to be rejoined by Christian Lund, co-founder of a company called Templify. And they are a leading AI-powered document generation platform that are trusted by over 800 industry leading companies. And when I say leading companies, I mean everyone from KPMG to BDO and Adobe.

[00:00:56] And with more than 4 million users and 30 days saved per employee annually through automation, I want to learn more about how Templify is reshaping how businesses create accurate, compliant and on-brand documents. So today we'll talk about how AI can revolutionize document workflows. And my guest will also explain why the best automation happens when humans aren't heavily involved in the process.

[00:01:24] And in doing so allows employees to focus on high value tasks rather than manual document management. So how can businesses harness AI to streamline their document generation without sacrificing compliance? And what does the next wave of AI agents mean for your organization's workflow? And how can reducing human input actually lower risks and improve outcomes? These are some of the big topics we're going to explore together and hopefully give you

[00:01:54] a few valuable takeaways along the way. But enough from me. Let's get Christian onto the podcast now. So a massive warm welcome back to the show, Christian. We last spoke back in 2018. We've probably got a, well I would hope we've got a few more listeners since then, but can you remind everyone listening with a little about who you are and what you do? Absolutely. Thanks for having me, Neil. So I'm Christian Lund. I'm a co-founder at Templify, automation of business documents with AI.

[00:02:24] And I'm also very active in especially the Nordic tech community, supporting businesses that are trying to become real businesses as an investor and as a, as a, as a divisor as well. And I was thinking before you joined me back on the podcast, the world was a very different place where we last spoke. There was a small matter of a global pandemic, the shift to hybrid working. And of course the last three years, everyone's gone crazy on AI and businesses are now increasingly

[00:02:51] turning to all things AI for automation and efficiency. But as I've seen that you've pointed out by looking at some of your content online, AI is only as good as the data and context that it receives. I suspect the one thing we had, do have in common when we last spoke was garbage in, garbage out. It's always been a problem. So how do you at Templify ensure that AI is given that right context to generate accurate and compliant documents and avoid that garbage in, garbage out problem?

[00:03:22] Yes. Um, and first of all, you're absolutely right. It's nothing has changed really because, uh, before AI, if you wanted to get something done and you had a task and you had two options, you could go to a random person and give them a little bit of input, or you could go to a subject matter expert and give them a lot of input and even best practices. It's pretty clear where you would get the best results. And we think that has sort of been inherited into, into the world of AI as well, that you really

[00:03:50] have to be on top of how you inform AI and be very deliberate on both the questions you ask, which AI you ask, which additional content you bring to that and how it's built into the workflows that you're trying to support as well. So, um, I would say the way that we do it and I've been investing very heavily on for the past few years is to allow technology had to have a bigger role in that, uh, be a little bit less

[00:04:15] dependent on the user being the one both asking the question and selecting the best AI and also building it into their workflows. But as much as possible to make sure that, uh, companies has a way to orchestra this a little bit better through technology. So, uh, to, to some extent being able to, uh, be a place where a lot of the pre-prompting can take place. So if you have an idea about, uh, some direction that the user taking that you could be more

[00:04:42] supportive to getting that in there, in the right direction, in terms of the question you want to ask AI, you can be more deliberate on which specialized AI or the most secure AI one you allowed to use can be used. And also, uh, that you can add additional context to it. So things that have worked well in the past around what you're trying to do, make sure that that influences the results. And finally, that it is automatically built into the workflows of Ross. That means the process of building out a business document.

[00:05:11] And if something has been built in the past that has performed very well, why not reuse that? But technology can really play a big role in, in, in some of that orchestration. And another trend that I'm seeing at the moment is this move away from traditional search engines, but we're kind of using the same habits when we turn to AI and many users, are prompting AI as they would a search engine using a few keywords.

[00:05:38] So I'm curious from what you're doing here, why is that approach problem problematic in document generation? And how do, uh, you at templify guide users towards better outcomes? Because it seems that search is evolving the way we use things are evolving, but we're going back or still using those bad habits maybe. Yeah. Yeah. And, uh, I think it's very understandable why you would, because as a user, you, you, what comes to mind is something very simple.

[00:06:04] I need a proposal proposal for this company would be an example that we would see in a, in a prompt. And, uh, and that's understandable because typically what's ranks the highest for a user is how quickly they can get it done. But especially sort of in a business context, and you want to use this output, not only for documents, but an AI in general, there is another sort of access that speaks more to the quality and the accuracy of the output.

[00:06:30] Um, and, uh, to get to, to that, the, that is really where the problem sits of, of having these, um, I would say under educated, uh, prompters with most users simply because, uh, the context that you're giving it, I like to call them briefs instead of, um, instead of prompts actually, uh, makes a big difference in terms of which direction you're sending the AI. And so, um, uh, the way that we're supporting that is to allow businesses and essentially

[00:06:59] users to use it. These pre-populated prompts that are a lot more rich that you can essentially trigger through a button. So to take something like extremely simple, uh, that I think most people have seen as well as like, if you have some, uh, some text, for example, and you want to, you want to simplify that or, or rewrite it in a different way, you could do that with a prompt. But if you have a button that does it, that Mike might have a lot of extra information

[00:07:26] built into that prompt, but just with a button where you could sort of improve a piece, uh, a piece of text with a button that says improve. Then underneath, you might have a very rich information that says how that should be improved in this context, which tone of voice to be used to say something simple, but it could also go way beyond that. So it's to use, um, uh, try to, try to capture what the user is trying to do and then be supportive

[00:07:51] of helping them use things that others have helped, uh, sort of, uh, think out to the end to make it more beneficial for the user as well. Uh, so that is something that we're trying to do a lot. And, uh, we think that there is a lot of opportunity of not necessarily having the users, uh, you know, needing to become prompt engineers, but they should just continue to do their work, but then something else needs to support that. So the end result is actually something that looks like it was done by a prompt engineer.

[00:08:22] And I was reading online that you said that the best automation happens when humans are less involved. So how does re reducing human interaction in document workflows benefit businesses as a whole, especially in terms of productivity and risk reduction, all those cool features that businesses have been trying to unlock for as long as I can remember. Yeah. Well, first of all, users are, and people, of course are very important, uh, important part of any business process. So I'm not neglecting that.

[00:08:51] I think they're extremely important, but there were some areas where, uh, we actually prefer users to do the work that they were hired to do, uh, which for example, if you're a commercial person, there's more to be with your customers, do selling, uh, do all the parts that's kind of sit, sit in that interaction, which technology has a, as a harder time do. But if you look at business documents, again, there are a lot of things there really rely on accuracy. If we take it like from a clear compliance perspective, there are certain things that just needs to

[00:09:20] sit in a business document simply for legal reasons sometimes. And you just don't want to leave that necessarily to the user just because they have other priorities. What they're typically trying to achieve is to get to their, what they think is their, their end goal, the purpose of the document, for example, to, to, to, uh, make sure that that turns into some sort of business at some point. But as part of it, you need to get the structure, right. And get all the disclaimers in and get all the metadata injected into the document.

[00:09:46] And we don't think that is something you would necessarily want to put that weight on the user to make sure that that happens, both because they don't, they don't want to go through the work to do it. Uh, and also because they have better things to do with such. Um, so that's one example. Um, then when we go further into the business document, we think that technology and especially now in the age of AI as, uh, so much ability to capture all the best practices, all the, all

[00:10:14] the things that have really worked well in the past, all the knowledge that is just on, uh, on catchable for an individual user and built that into the workflows and essentially support the user like a butler to get to just better outcomes, uh, than they would be able to do on their own. One example I could use is, is, is a copilot for example, with Microsoft, we were wondering what that came out, uh, and they gave it the name copilot. Well, who is the pilot then?

[00:10:41] And I think in the world of Microsoft, the pilot would probably be the user that then has a copilot. We think in a business context, it's a little bit different. Actually, the user is more the, the, the passenger because they know the destination they're going on, but we don't want them flying the plane. That's actually the business that does that through technology. And then they have these copilots around them that can be supportive in terms of what they're trying to achieve. So that's a sort of a, uh, a slight nuance to, uh, to, to some of that thinking maybe.

[00:11:11] And as an overcautious X IT guy, I'm also compelled to raise a question around how compliance and security risks, they also remain major concerns in document creation too. So how do you, how does Templify's AI powered platform help enterprises minimize some of those risks while also remaining, uh, or maintaining efficiency? Well, uh, we've been, we've been speaking to, to some of it, but like the V on the very simple path,

[00:11:40] uh, in terms of AI, what we heard from a lot of companies is that they're building out their own models, very specialized models that does something very specific. We see literally thousands of those projects going on with our customers right now. And they do that not only for, because they can be specialized in terms of what the output should be because, but also because they trust it. And one worry that they definitely have is that users go rogue and just pick up whatever

[00:12:08] they can find that they, uh, that that is available to them. So there's a big part of making sure that, uh, the information that, that flows and the, the prompts that are sent goes to a very deliberate AI seen from the perspective of the company to support the user, but also take care of control of the, of the security that sits within that. And that kind of expands all the way to what actually goes onto the business documents that, um, you need to be able to have something that is reliable and that performs well, but also

[00:12:38] something that doesn't put you in a breach situation. So you have to be able to under two, to make sure that what goes onto a document, even created with, uh, with AI is actually something that is sufficiently accurate and you can, you can trust and will not put you in like in legal problems because you're telling things that is not true or that you're not allowed to use for example. So we do that, uh, in templify with a very strong orchestration of this. So again, which question is asked, where does that question go to, which model? Wow.

[00:13:08] Do we ground that information through other means like racks and so on? And how is that put very, very deliberate places inside of the business document to get to this very high accuracy of the output. And that is definitely one of the areas where we are very unique in terms of that ability to be very orchestrational towards, uh, detailed information that is created by, uh, generative AI.

[00:13:33] And since we last spoke, you've secured partnerships spanning from the big four accounting firms to global brands, such as Adobe. I'm curious, how do you at templify tailor your solutions to meet such a diverse needs of, of such a broad client base? Cause they seem to be at so many different ends of the spectrum. Uh, you're, you're absolutely right. And we do have like, uh, I would say a preference for what we call knowledge workers.

[00:13:58] So the ones that essentially rely on business document to deliver their value to their customers. Uh, so there, there is kind of that sort of narrowing in, but that is still a very broad field because you essentially have like hundreds of millions of those types of users that are doing that. Uh, but dog gen is a very generic thing because it falls into so many different business processes where you need a document to,

[00:14:21] to essentially, uh, document the work that you're doing to be able to deliver that to a customer or even internally. Um, so the way that we do it is by having built, I would say a generic, a general, uh, very strong document compilation engine that sits on top of AI, uh, that is able to do these things that is easy to, to sort of translate into very specific use cases.

[00:14:45] And that gives us the ability to speak with our customers in terms of what is required, uh, from the, for their specific need to connect all the different, uh, uh, data sources and AI models that they're using and build that into something that supports the outcome that they're trying to achieve. So in that way, we can be a little bit, a little bit less sort of specific on the document type and be a little bit more generic.

[00:15:08] But then having said that we definitely, as I said, support the ones that have the highest need to deliver an output and especially the ones that sit in a commercial role. Uh, there is a strong focus on the ROI of every tech project now, along with that measurable value, tangible results that AI will deliver. If it will deliver on those promises and when doing a little research, I was reading that Templify has reported that it can save employees 30 days annually by automating document creation.

[00:15:38] And just to bring that figure to life that I was reading about there, are there any real world examples you can share of, of how these time savings translate into tangible business value? Anything you can share around that? Uh, yes. I mean, lots as, as maybe, as you can imagine, we're actually business document is one of the areas where most time is being consumed. And I would, I would say lost in businesses today and people really need to go back to their real work. And there is so much automation you can do.

[00:16:08] The one very concrete example would be, uh, Adobe maybe that in, in, in sort of one of their, uh, bigger sales departments is actually 70, 72% of the time that they're taken off. Of the time they, they, they, uh, they had to spend before creating out their, their big, their pitch decks and, uh, and, and sales collateral with that.

[00:16:29] But there are just tons of examples of that with these very, uh, another, another example could be like bid management going through the process of answering to RFPs and going through the, the, the steps of building out the bits and proposals that sit on the back of that. But it's just such a time consuming task to do that.

[00:16:48] Uh, there is just enormous amounts of, uh, of time to be found to give back to users, to do their real jobs, which is essentially do all the discussions with the customers, uh, rather than sitting down and building business documents. Absolutely love that. Yeah. And, uh, of course, as AI continues to evolve already three years into the generative AI high, but are there any potential risks that businesses should be watching out for when integrating AI into document generation?

[00:17:16] And how are you at Templify helping to mitigate some of these challenges? Because it felt like for the first couple of years, people are going headfirst, diving in. I'm only just starting to notice some of those challenges, but what are you seeing here? Yeah. Well, again, the, um, the big promise of AI and all the first videos we saw that came out that was, uh, on, on doctrine were super impressive in terms of how quickly you could get from, from like an intent to an output.

[00:17:44] So the, the, the first sort of promise that really came through was really speed and what was underestimated for a while. And I think a lot of businesses have picked up now is that speed is not enough on its own. That's definitely an expectation that AI would give you speed to, to get from A to B. But the other axis again, is like quality has to be there. Still cannot be worse off than you were before AI came in, because that's just going to put you in a very problematic situation.

[00:18:13] And there was a risk that your, your cons is not going to perform as good as it could, but there's also a risk that you do things that you're not allowed to do. That would put you in legal risk as discussed just earlier. So, um, we're, we're definitely seeing, uh, business, a lot of businesses, pretty much everyone we work with investing quite heavily in getting that other access in, in a lot of the use case.

[00:18:36] Also the ones that are not, has nothing to do with dog gen, but, uh, how is it that we add the quality layer to that and give, uh, um, uh, technology, a more orchestrational layer on top of AI to make sure that we can actually achieve that.

[00:18:50] And make sure that it's the company that still has kind of a pretty high level of control, uh, to make sure that we, we, we end up with quality that is sufficient to do, uh, to do, to get to our business deliverables in a better way than we were able to do before and not worse. And as someone right in the heart of this space, you've seen it evolve over the last decade and possibly beyond. What trends do you see shaping the future of document generation?

[00:19:19] There's so many different AI solutions, solutions and use cases appearing now, but what are you seeing and how are you preparing for some of these changes as well? It feels like quite an exciting time and indeed a good time for opportunities for you too. Oh yes. It's, it's a, it's a very sort of pivotal moment for sure.

[00:19:37] And, uh, again, I would, I would say when, when, when the big wave came out, uh, Microsoft launched, uh, co-pilot, for example, um, there are still things to fix in terms of what, uh, how doctrine is done with AI also with co-pilot. But what that definitely gave that push was that the expectation in the market is that documents is some will be something that is going to be generated with AI for sure. That expectation now sits in the market.

[00:20:05] So there's no doubt that the ones that don't have that focus right now, I think are going to be in trouble if they don't find very good ways to figure out how to, how to support their doctrine engines with AI. And that's of course the reason why we've taken that, uh, that approach as well. Uh, what is also common in is kind of next wave AI with the agents as well, which is super exciting.

[00:20:28] Uh, that goes on top and there's a, there's a big push going on in many companies, including ours around how agents can play a bigger role with that. So it's AI as we know it, it's the agents coming in and then definitely, uh, the big piece as we've discussed today, the whole orchestration around it, that doesn't put the businesses in the, in, in the dark, but actually keep them up front and in control. Uh, those are the big trends we see. Exciting times indeed.

[00:20:55] And thank you so much for taking the time to sit down with me again and share your insights with not only myself, but everyone listening today. And I want to see if there's something we can do for you now, because some of the biggest names in business, VC funding and tech have either been guests or even listened to this podcast, or maybe there's somebody who is a friend of a friend out there. So is that a person that you'd love to have a breakfast or lunch with? And he or she may just get to listen to this. Let's see what we can manifest together today, but who would it be? And why?

[00:21:26] You know what? I'm going to give you a cheesy one for that one, uh, Neil, actually, because I, I've been following this show and there is a, there, you know, I don't think there's anyone I wouldn't love to have a lunch with a breakfast with or a lunch, but I would pick you actually, just because you've, you've done this for so long and you met, you know, thousands of people. So I think you might have the best insights of, of using that and maybe a good leverage to, to maybe not make my own connection, but maybe multiple as a part of that, as a part of that work. So I would love that.

[00:21:55] Ah, well, we can make that happen. Nice and easy. So are you still based in the UK? Christian, are you in London still? Whereabouts are you? I'm actually in Copenhagen in Denmark. Ah, Denmark. Okay. Right. So I'm going to see what we can do there because I, I do travel to tech conferences a lot. I know Copenhagen is just a short flight away and usually an inexpensive flight too. So I want to see what we can make happen there. Maybe record something in person as well.

[00:22:22] And for everyone listening, wanting to find out a little bit more information about you, about Templify and everything. Where would you like to point everyone listening? Well, um, very welcome for Templify to go to our website, templify.com. Or myself, Christian Lund. You can definitely look me up on LinkedIn and other social media and, uh, um, happy to engage with anyone. So just please reach out. Well, we covered so much there, especially some great points around the less time people spend doing manual document or data work,

[00:22:51] the more time they can spend on driving things that deliver real business value. And I love that line you've used before the less people involved in automation, the less compliance security risk businesses face so many big talking points there. And not to mention that we've also got a lunch date to sort out. So thank you so much for sitting there with me today. Of course. Thanks so much for having me, Neil. Having spoken with Christian, I think the future of AI-powered document generation,

[00:23:19] it isn't just about speed and automation. Most importantly, it's about precision, context, and compliance. And as he pointed out there, AI works best when humans don't have to intervene in the manual aspects of document creation. And anyone looking for ROI or measurable difference saves a fair bit of time, up to 30 days per employee each year.

[00:23:43] But also, possibly more important, it significantly reduces compliance risks. So today, we've learned that the quality of AI-generated documents depends heavily on rich context, structured briefs, rather than simple prompts. But Tempify's approach to orchestrating AI models, data sources, and workflows seems to ensure that businesses can maintain control, security, and accuracy throughout that entire process.

[00:24:13] But the question remains, how ready is your business to embrace this level of automation? Is your organization adopting AI-powered solutions that deliver on those promises of value without compromising security and compliance? Or will the risks of automation keep you tied to that old-school, manual, time-consuming processes? I'd love to hear your thoughts.

[00:24:36] Please, join the conversation, share your perspective by emailing me, techblogwriteroutlook.com, LinkedIn, X, Instagram, just at Neil C. Hughes. And my website, techblogwriter.co.uk, where you can send me a message or just browse through 3,200 interviews that we have now. Let me know. I'll be back again tomorrow with a completely different topic of how technology is transforming our work, world, life, and everything in between.

[00:25:05] And hopefully I will be able to speak with you all once again. But that's it for today. Bye for now.