Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
Jörn "Joe" MenningerIf custom software used to be too expensive for small companies, what happens when AI makes it cheap enough to compete with SaaS? Can AI really make small and medium-sized companies faster, or does it simply automate broken processes at higher speed? How should founders, SME owners, and investors think about a company like Amadeni AI, led by Delia Hallberg, which positions itself around AI-supported custom software for small and medium businesses, says typical projects range from €15,000 to €50,000,
Jörn "Joe" Menningerand argues that software should adapt to company workflows rather than forcing companies into rigid systems. Hello and welcome everybody. How do we separate real AI-enabled operational leverage from another wave of digital transformation theater? Delia, welcome to Startuprad.io. What is the operating problem in the Mittelstand that you think most people still describe incorrectly?
Delia HallbergHi Jörn, thank you for having me. I would say, the Mittelstand is often misunderstood. Let's talk about it in the interview.
Jörn "Joe" MenningerGreat. Let's get started. When you look at a small or medium-sized business for the first time, what tells you within the first hour whether the real problem is software, process, or even the leadership?
Delia HallbergI would say it's almost never just one. And what makes this moment in time critical, when you introduce AI, you stress test every process. A bad process with AI on top of it breaks faster, louder, and more expensively. And I would say, because you asked about software and process, I would say the two are married. You won't find a perfect process with a software problem or terrible software with a great process.
Delia HallbergAnd that's why we approach development in an extremely process-driven way. There is usually a signal for each of them. So on the software process side, the signal is, I always ask, walk me through what happens when you receive, for example, an invoice. And then I count systems, how many times people copy and paste information, and how much inefficiency the process contains. For example, we had a bakery with around 40 media discontinuities, which cost 35 extra hours of work.
Delia HallbergSo if you would have thrown AI on top of that you would have automated chaos. Then there is this leadership signal. For example, when the managing director describes a workflow completely different from what the team described 5 minutes earlier, that gap is a real risk. So we have to be very cautious of it. And that's why we run extensive workshops before writing a single line of code. The diagnosis determines everything.
Jörn "Joe" MenningerDiagnosis determines everything. That is interesting. You have an interesting background because you actually studied sociology before moving into technology and product work. What did that teach you about organizations that are purely technical founder might miss?
Delia HallbergI would say I don't take requirements at face value. When a client says we need that tool that does X, or even a lot of conversations with clients start with, I need AI to do this and that for me. Maybe many technical founders would hear a feature request. So I come from systems thinking, and I think that's what sociology has taught me. And I hear a symptom and I ask, why do you need that?
Delia HallbergWhat's the process behind it? And understanding the entire system is sociology at its core. And for example, with one client, a youth welfare organization, the brief was, we need a better documentation tool. But when we start looking at the whole picture, like how do you actually work and where inefficiency occurs in the entire process, we very quickly understood that they don't need a better documentation tool. They need an entirely different process of how they interact with
Delia Hallbergthe software on a daily basis. So in the end, we wrote an entire ERP system for them with voice-first interactions. and semi-automated processes where documentation is only one smart part of the chain.
Jörn "Joe" MenningerYou basically taught them how to use the tools. Okay, I see. ERP with voice-first interaction. That's also an interesting piece. Would love to see that one day. In the past, you have argued that AI made your own company possible because it lets you operate with a much smaller team. What exactly would have required 4 people before and what is now handled by agents?
Delia HallbergExactly. So our situation is that we started with 2 people, with 2 founders, and were actually able to go a very long way with only 2 people. But I think we as you already said, we really did the work of 4 or even more people. The first area in which we work closely with agents is the most obvious: marketing and content. I would say 80% are outsourced to AI.
Delia HallbergAgents hold our strategies, our content plans, our pipeline. And I do the Q&A and finalization. So then there is this research because we work across industries and sometimes we start a project in a new industry that we are not yet familiar with. So we do deep dives into sectors and client processes before we start a first conversation, which AI is great at helping us with. And then there is the sales part, obviously, that were able
Delia Hallbergto automate quite a big workflow that basically makes sure that during the sales process we have quality assurance, every piece of data we can gather enters the process so that when this lead converts into a project, we do not start the project from zero, but we already have, for example, a project outline created by the previous sales agent. And then the fourth and most significant area is coding. This is where we really have the biggest efficiency gains.
Delia HallbergWe have multiple coding agents. We have, I would say by now we are replacing an entire IT department with a very coherent system of agents and it's not vibe coding. Please don't misunderstand me. It is an elaborate process designed to ensure consistent, coherent code quality. Previously, that would have required maybe one additional person that would somehow do sales, research, marketing combined, and at least three additional developers.
Jörn "Joe" MenningerWhen you talked about preparing for a new engagement, you were talking about all the data you add before the client actually says yes. I was wondering how, because it's from the outside, how do you ensure sufficient quality that you don't have the typical system problem, garbage in, garbage out?
Delia HallbergThat's a very good point. So first of all, there is always a human in the loop and I always have to be able to understand the results. Otherwise, yeah, that's a risk. And then again, it's about setting up a system where you lower the risks of the AI producing low-value results or wrong results, which you can do by certain steps that you do while creating your agent system and not by simply throwing a prompt
Delia Hallbergin ChatGPT and hoping for the best.
Jörn "Joe" MenningerHoping for the best. Yeah, I think for some people out there it's still strategy. Where is the boundary between a lean AI-enabled company and a founder simply hiding the operational fragility behind some automation?
Delia HallbergThat's a very relevant question, and, especially at this time. So I would say the boundary is where you understand what your agents are doing well enough to do it yourself if they stop tomorrow. Agents don't replace competence, they replace capacity. The judgment, the quality standard, the strategic thinking, this always has to stay with the humans, in that case with me. And I always think about it it's very tricky because AI is so good at producing plausible output that you can fool
Delia Hallbergyourself for much longer before you realize that you've actually lost the plot. That's also something that we really take care of with our clients, with the tools that we create for them, that they always stay in the driver's seat. We do a few things to manage that. First, as I already said, there's always a human in the loop. Second, we document agent architectures just like software architectures. It needs to be scalable and not just living inside my head.
Delia HallbergAnd third, I'm really honest about what AI can't do. For example, client relationships, reading the room, judgment calls in whether and how a project needs to shift. And I would say an indicator of operational fragility is when founders or people talk about AI replacing this and that department, AI replacing this and that person because they are thinking about AI in terms of replacement. I think the discussion that we should be having is AI in the sense of amplification.
Delia HallbergThe competence was there first and then AI makes it scale.
Jörn "Joe" MenningerNot only that I have learned that AI agents drift over time. So maybe at the start you get a marketing department and when you are towards the end of the year, the tool only does social media anymore. So I don't think there's right now a case for complete AI replacement for humans. So as you said, you always have to have a human in the loop. Your public positioning says Amadeni, your company, starts with a workshop, process analysis,
Jörn "Joe" Menningerand then decides whether custom software is actually needed. What is your decision rule for saying nope, you should not build software here?
Delia HallbergSo it's important to say we always look for a standard solution first, because if you have a standard business case, and then there is a SaaS product that genuinely fits, then I would say that this is the way to go. That's part of being an honest IT consultancy. But what I've observed in the Mittelstand, especially in Eastern Germany, is that businesses that have been established for 20, maybe 30 years are the ones that are sitting in a niche.
Delia HallbergAnd they're at the intersection of 2 industries. So what happens quite frequently is that there is no proper software solution for them. What they have done until now is they have tried to use software maybe from an industry that they're slightly part of. But this is creating so much problems than in daily processes. That for example, a straightforward case where we go, okay, custom development really is the answer. And one signal that is growing stronger and stronger, for example,
Delia Hallbergwith an engineering firm in Berlin that we are currently working with, they were very happy with their SaaS solution, with their sort of project management solution, but they had been asking for a long time for a very specific feature. They didn't get it. They didn't get an API. So this, I would say, minor issue is really what caused them to say, okay, we have to say goodbye to our solution. And they proactively decided to go for a custom development.
Delia HallbergBecause if a SaaS product doesn't have an API, it means that nobody can add automation. There is no flexibility. And this is something that surprised even me. This is really coming from our clients. So the decision tree, I would say, is simple. Is there a good standard solution that exists? If yes, go for it. If no, then custom software is really becoming a very viable alternative. One point is especially important for clients.
Delia HallbergIf they don't have an API or some way to interact with their data, then they're also going for the alternatives.
Jörn "Joe" MenningerYes, we should always keep in mind it is not only a question of standard software or not, but also is the standard software breaking the bank or not? I think you know what I'm talking about here. I've been going through a little bit of your data, and in your Impuls KBB case, the reported gain was from 8 hours to 2 hours for development reports and a 4-week return on investment. What had to be true operationally for that result to be
Jörn "Joe" Menningerreal rather than just a software demo?
Delia HallbergI really like it that you're looking at the numbers. So let's look at the whole picture. The four-week ROI and calculated savings of €21,000 per month did not come from one process, this efficiency increase from 8 to 2 hours. As I mentioned a few questions earlier, went through the entire process chain starting with case clarification from the youth welfare office, really to the very end of invoicing. And we really looked at every step where we can increase efficiency.
Delia HallbergAcross the entire process, there was so much room for improvement that in that case were really able to achieve this, I would say, impressive ROI. So first the client made a very, I would say, bold move at the time because we had already started the project and soon after the project began they decided to replace their old software. They saw that it did not have an API. Certain features in their previous software were causing a lot of
Delia Hallbergproblems and they decided not to supplement it but replace it. And that unlocked the enormous automation potential. For example, the voice-first interactions, because these are social welfare workers that go to the homes of children or youth. And then afterwards they have to sit in their car, open the laptop and type up documentation about what they discussed Now, on their way back to the car, they can simply make a voice message which then gets transcribed and then translated into
Delia Hallbergofficial bureaucratic language and then placed into an automated workflow. Second, we built the system for how the social workers actually work. As I said in the example of the voice notes, They also had a lot of unofficial communication about what was the status quo of this document and have you thought about this and that? So we took a lot from that informal communication that created really a lot of inefficiency and put it into a process within the software,
Delia Hallbergwhich eliminates a lot of wasted time. And then third, we integrated it with the rest of the system because they had a separate HR system. So they had to do 2 types of time tracking. And we said, no, the system has to work for the social workers. They need only one interface. And in the background, we will create a flow of data so that they have only one touchpoint. And I would say this entire process optimization built or visualized in the tool,
Delia Hallbergthat's what really created this four-week ROI.
Jörn "Joe" MenningerI had to smile when you said that you make a voice memo and then it gets translated into bureaucratic language. It's— that's really a good point, a translator between bureaucrats and normal human beings. But let's go a little bit further. When a client has 20 systems, 40 media discontinuities, and 35 hours of avoidable duplicate work per week. How do you decide what you will fix first?
Delia HallbergYeah, so the numbers that you quote come from a case, from a bakery that produces its own sourdough goods. That's actually important for the process because sourdough makes things very complicated. But the CEO called us because they had a bad gut feeling. So we did an analysis in a 2-day workshop and this bad gut feeling translated into these numbers that are quite shocking or quite surprising, I would say. And I'm very open about it.
Delia HallbergWe were a little bit overwhelmed at first, but what is really important, and I think that's great that the CEO actually took that step because you don't have to automate everything. You don't have to optimize for everything, but you need to prioritize and have a plan for how you want to deal with the other things. First, we looked for ticking time bombs. That's how I call them. So something that could potentially go off any day and threaten the business.
Delia HallbergSo if you find a ticking time bomb, then that's obviously at the very top of your roadmap. And then you deal with it regardless of everything else. In that case, we did not find any ticking time bombs. But then you look at the fundamental structural things. In that case, it was actually that they didn't have a leading system. which led to a lot of distributed data flows. And so it was very clear that we first had to
Delia Hallbergbuild a very sound foundation, like a building, and you don't want to build it on sand. And then we looked at impact. So where did we have the lowest investment or where could we create the greatest business impact with the lowest effort? Then you have a prioritization, a roadmap. And as I said initially, it's totally okay if you don't tackle every process, the important difference is that you are aware of them,
Delia Hallberghave quantified them and you have a plan. And I think that makes a difference between a vague gut feeling and having worked through it.
Jörn "Joe" MenningerYeah, you just have to have a framework built on numbers that you can add to. I was wondering, talking about numbers and metrics here, what is the metric you trust most when evaluating whether an AI automation project is worth building? Do you think about hours saved, reduction in errors, adoption rate, revenue impact, or maybe management relief?
Delia HallbergI would say I try to collapse all these things that you just said into a single number because at the end you can put a price tag on nearly all those things. What is it costing you right now in terms of time, errors, management firefighting, but also what would the solution cost? So if the solution has an ROI of less than six months, then it's a very straightforward decision to go for it.
Delia HallbergIf it is 12 or 18 months, we have a more nuanced conversation, and it's also more about strategy, about potential risks also of not acting, right? But I would say there is a very big misconception that SMBs frequently have. It's a conversation that I have with a lot of especially potential clients at the beginning is they treat IT as its own budget. So for example, a €25,000 project looks like it would blow the IT budget,
Delia Hallbergbut and that's the difference from my perspective. You have to calculate the IT budget in terms of personnel costs of the inefficient process, and then when you compare the personnel costs versus the investment into optimizing these processes and the potential personnel costs that you save, not because you're letting go of people, but because. Suddenly you can come from a very reactive business to a more strategic one, or your people can actually do more meaningful things.
Delia HallbergThat reframing is one of the most important conversations that we have, and then suddenly the decision-making becomes much easier.
Jörn "Joe" MenningerThe decision-making. making becomes easier. Talk about processes here. You argue that a a poor process plus AI becomes a poor process on steroids. What is the clearest sign that process must be redesigned before it's even automated?
Delia HallbergThere is a very clear sign. And it is when nobody can explain the process without naming a person. So if people start saying, so I get the invoice and then Sabine does this and that and then Thomas does that for example, that's not a process, those are personal habits. And we don't want to automate people's personal habits because if you automate that and then the moment Sabine goes on holiday, the system doesn't work anymore.
Delia HallbergSo especially in the Mittelstand, that is a process that we go through to get away from this person name-driven thinking to, okay, what are the actual steps? A second straightforward sign is if information gets typed more than twice. A third is, shadow systems. Shadow systems—shadow IT—are a very strong indicator that the process isn't working. And if you automate that process without integrating the shadow systems or without looking at why people had to create shadow systems,
Delia Hallbergthen you'll create a tool that nobody's going to use. I'm talking about private Excel files where the real numbers are, or the WhatsApp group that's the actual project management tool. So these are indicators where it's very clear that you have to redesign first and automate second.
Jörn "Joe" MenningerYes, I was just smiling talking about shadow IT and WhatsApp groups that actually are used for project management. I have seen that and in my experience, startups tend to do this more frequently. There was a heyday for Slack, but I think it's partly gone.
Delia HallbergWhatsApp has remained popular in the Mittelstand.
Jörn "Joe" MenningerYeah, maybe it just, it just moved there. That's also fine. Talk a little bit about more cases that I've seen from you for example, the Ostdeutscher Bankenverband case. The promise was source traceability and zero hallucination for an AI. What architectural choice prevents hallucination by design rather than by, yeah, you know, just hope?
Delia HallbergYeah, hope and prayers are not something that you want, in your system design or in your IT. so what we did is, I think the key is obviously the key is not literally zero hallucinations, but lowering the risk and having a system that can catch hallucinations. Because in that case with the Ostdeutscher Bankenverband, they provide information to their members that is crucial that their members base decisions on. And if you have a hallucinated piece of information, that's reputation damaging, right?
Delia HallbergSo what we did is obviously human in the loop at every step. Then second, we gave different, unconnected LLMs small tasks, because the smaller the task, the more concrete the information, the less the AI is forced to hallucinate. When does AI hallucinate? When models lack information but still have to come up with an answer, right? And so the answer was to cut the task very small and to include in the system a traceability of information.
Delia HallbergSo every piece of information that I had, I could trace back to the original website, to the original source of information. And this was especially important for the humans. who were in the loop to be able to prove every piece of information. I would say that these two or three design decisions lowered the risk of hallucinations very much.
Jörn "Joe" MenningerSo you give the AI a very small area of responsibility at each step and enough information to fulfill it in order that they Don't need to hallucinate. Great. Delia, it was a pleasure talking to you. We agreed that we're going to cut this interview into 2 pieces. We'll be back with number 2 for you in a few days, for us in a few minutes. Talk to you soon. Thank you.
Climate tech hardware in Germany rarely fails on technology. It fails between pilot and plant, in a €5 million to €50 million gap that is too big for angels and too risky for most funds. Business angel Birte...
Helsing Clarity: How AI Agents Fuse Battlefield Data Helsing's Clarity runs hundreds of specialised AI agents over satellite imagery, drone video, signals and written reports. They share one military data model and one...
ICEYE already delivers sovereign satellite systems for seven European governments, and the Scaleup Europe Fund co-led its €1 billion financing round. Mistral AI raised €3 billion. Jörn “Joe” Menninger asks whether...
Germany chose physical AI: NVIDIA plans stakes in Agile Robots, Avelios, Mirelo and Reflex Aerospace; RobCo passes $1B; Zurich's Enact raises a $30M pre-seed. A note on timing: because of my travel schedule, this...
E 783 - Laser Fusion Spin-Off: How TU Darmstadt Built It Laser fusion is the hardest deep tech bet in Germany: huge capital, more research, a plant targeted for the mid-2030s. Focused Energy took it with a $240M Series...