WEBVTT

NOTE Timings are estimated because the source transcript contained no timestamps.

00:00:00.000 --> 00:00:04.314
<v Jörn "Joe" Menninger>If custom software used to be too expensive for small companies,</v>

00:00:04.394 --> 00:00:09.100
<v Jörn "Joe" Menninger>what happens when AI makes it cheap enough to compete with SaaS?</v>

00:00:09.180 --> 00:00:12.709
<v Jörn "Joe" Menninger>Can AI really make small and medium-sized companies faster,</v>

00:00:12.789 --> 00:00:16.711
<v Jörn "Joe" Menninger>or does it simply automate broken processes at higher speed?</v>

00:00:16.791 --> 00:00:22.281
<v Jörn "Joe" Menninger>How should founders, SME owners, and investors think about a company like Amadeni AI,</v>

00:00:22.361 --> 00:00:26.675
<v Jörn "Joe" Menninger>led by Delia Hallberg, which positions itself around AI-supported custom software</v>

00:00:26.755 --> 00:00:31.853
<v Jörn "Joe" Menninger>for small and medium businesses, says typical projects range from €15,000 to €50,000,</v>

00:00:31.933 --> 00:00:37.933
<v Jörn "Joe" Menninger>and argues that software should adapt to company workflows rather than forcing companies into rigid systems.</v>

00:00:38.013 --> 00:00:39.581
<v Jörn "Joe" Menninger>Hello and welcome everybody.</v>

00:00:39.661 --> 00:00:45.544
<v Jörn "Joe" Menninger>How do we separate real AI-enabled operational leverage from another wave of digital transformation theater?</v>

00:00:45.624 --> 00:00:47.192
<v Jörn "Joe" Menninger>Delia, welcome to Startuprad.io.</v>

00:00:47.272 --> 00:00:53.272
<v Jörn "Joe" Menninger>What is the operating problem in the Mittelstand that you think most people still describe incorrectly?</v>

00:00:53.352 --> 00:00:56.097
<v Delia Hallberg>Hi Jörn, thank you for having me.</v>

00:00:56.177 --> 00:00:59.315
<v Delia Hallberg>I would say, the Mittelstand is often misunderstood.</v>

00:00:59.395 --> 00:01:02.140
<v Delia Hallberg>Let's talk about it in the interview.</v>

00:01:02.220 --> 00:01:03.620
<v Jörn "Joe" Menninger>Great.</v>

00:01:03.700 --> 00:01:05.100
<v Jörn "Joe" Menninger>Let's get started.</v>

00:01:05.180 --> 00:01:10.278
<v Jörn "Joe" Menninger>When you look at a small or medium-sized business for the first time,</v>

00:01:10.358 --> 00:01:15.848
<v Jörn "Joe" Menninger>what tells you within the first hour whether the real problem is software, process,</v>

00:01:15.928 --> 00:01:17.496
<v Jörn "Joe" Menninger>or even the leadership?</v>

00:01:17.576 --> 00:01:20.714
<v Delia Hallberg>I would say it's almost never just one.</v>

00:01:20.794 --> 00:01:25.500
<v Delia Hallberg>And what makes this moment in time critical, when you introduce AI,</v>

00:01:25.580 --> 00:01:27.540
<v Delia Hallberg>you stress test every process.</v>

00:01:27.620 --> 00:01:33.503
<v Delia Hallberg>A bad process with AI on top of it breaks faster, louder, and more expensively.</v>

00:01:33.583 --> 00:01:37.896
<v Delia Hallberg>And I would say, because you asked about software and process,</v>

00:01:37.976 --> 00:01:40.722
<v Delia Hallberg>I would say the two are married.</v>

00:01:40.802 --> 00:01:45.115
<v Delia Hallberg>You won't find a perfect process with a software problem or</v>

00:01:45.195 --> 00:01:47.548
<v Delia Hallberg>terrible software with a great process.</v>

00:01:47.628 --> 00:01:51.942
<v Delia Hallberg>And that's why we approach development in an extremely process-driven way.</v>

00:01:52.022 --> 00:01:55.551
<v Delia Hallberg>There is usually a signal for each of them.</v>

00:01:55.631 --> 00:02:00.337
<v Delia Hallberg>So on the software process side, the signal is, I always ask,</v>

00:02:00.417 --> 00:02:05.123
<v Delia Hallberg>walk me through what happens when you receive, for example, an invoice.</v>

00:02:05.203 --> 00:02:10.301
<v Delia Hallberg>And then I count systems, how many times people copy and paste information,</v>

00:02:10.381 --> 00:02:13.126
<v Delia Hallberg>and how much inefficiency the process contains.</v>

00:02:13.206 --> 00:02:17.520
<v Delia Hallberg>For example, we had a bakery with around 40 media discontinuities,</v>

00:02:17.600 --> 00:02:20.345
<v Delia Hallberg>which cost 35 extra hours of work.</v>

00:02:20.425 --> 00:02:26.425
<v Delia Hallberg>So if you would have thrown AI on top of that you would have automated chaos.</v>

00:02:26.505 --> 00:02:28.858
<v Delia Hallberg>Then there is this leadership signal.</v>

00:02:28.938 --> 00:02:33.252
<v Delia Hallberg>For example, when the managing director describes a workflow completely different</v>

00:02:33.332 --> 00:02:38.822
<v Delia Hallberg>from what the team described 5 minutes earlier, that gap is a real risk.</v>

00:02:38.902 --> 00:02:42.431
<v Delia Hallberg>So we have to be very cautious of it.</v>

00:02:42.511 --> 00:02:48.002
<v Delia Hallberg>And that's why we run extensive workshops before writing a single line of code.</v>

00:02:48.082 --> 00:02:49.650
<v Delia Hallberg>The diagnosis determines everything.</v>

00:02:49.730 --> 00:02:51.130
<v Jörn "Joe" Menninger>Diagnosis determines everything.</v>

00:02:51.210 --> 00:02:52.610
<v Jörn "Joe" Menninger>That is interesting.</v>

00:02:52.690 --> 00:02:57.004
<v Jörn "Joe" Menninger>You have an interesting background because you actually studied sociology before</v>

00:02:57.084 --> 00:02:59.437
<v Jörn "Joe" Menninger>moving into technology and product work.</v>

00:02:59.517 --> 00:03:05.007
<v Jörn "Joe" Menninger>What did that teach you about organizations that are purely technical founder might miss?</v>

00:03:05.087 --> 00:03:09.009
<v Delia Hallberg>I would say I don't take requirements at face value.</v>

00:03:09.089 --> 00:03:13.402
<v Delia Hallberg>When a client says we need that tool that does X,</v>

00:03:13.482 --> 00:03:17.404
<v Delia Hallberg>or even a lot of conversations with clients start with,</v>

00:03:17.484 --> 00:03:21.405
<v Delia Hallberg>I need AI to do this and that for me.</v>

00:03:21.485 --> 00:03:25.015
<v Delia Hallberg>Maybe many technical founders would hear a feature request.</v>

00:03:25.095 --> 00:03:30.977
<v Delia Hallberg>So I come from systems thinking, and I think that's what sociology has taught me.</v>

00:03:31.057 --> 00:03:36.155
<v Delia Hallberg>And I hear a symptom and I ask, why do you need that?</v>

00:03:36.235 --> 00:03:38.196
<v Delia Hallberg>What's the process behind it?</v>

00:03:38.276 --> 00:03:42.198
<v Delia Hallberg>And understanding the entire system is sociology at its core.</v>

00:03:42.278 --> 00:03:47.376
<v Delia Hallberg>And for example, with one client, a youth welfare organization, the brief was,</v>

00:03:47.456 --> 00:03:49.809
<v Delia Hallberg>we need a better documentation tool.</v>

00:03:49.889 --> 00:03:53.418
<v Delia Hallberg>But when we start looking at the whole picture,</v>

00:03:53.498 --> 00:03:58.988
<v Delia Hallberg>like how do you actually work and where inefficiency occurs in the entire process,</v>

00:03:59.068 --> 00:04:03.774
<v Delia Hallberg>we very quickly understood that they don't need a better documentation tool.</v>

00:04:03.854 --> 00:04:08.168
<v Delia Hallberg>They need an entirely different process of how they interact with</v>

00:04:08.248 --> 00:04:10.601
<v Delia Hallberg>the software on a daily basis.</v>

00:04:10.681 --> 00:04:16.563
<v Delia Hallberg>So in the end, we wrote an entire ERP system for them with voice-first interactions.</v>

00:04:16.643 --> 00:04:21.741
<v Delia Hallberg>and semi-automated processes where documentation is only one smart part of the chain.</v>

00:04:21.821 --> 00:04:25.351
<v Jörn "Joe" Menninger>You basically taught them how to use the tools.</v>

00:04:25.431 --> 00:04:26.831
<v Jörn "Joe" Menninger>Okay, I see.</v>

00:04:26.911 --> 00:04:28.479
<v Jörn "Joe" Menninger>ERP with voice-first interaction.</v>

00:04:28.559 --> 00:04:30.520
<v Jörn "Joe" Menninger>That's also an interesting piece.</v>

00:04:30.600 --> 00:04:33.345
<v Jörn "Joe" Menninger>Would love to see that one day.</v>

00:04:33.425 --> 00:04:37.739
<v Jörn "Joe" Menninger>In the past, you have argued that AI made your own</v>

00:04:37.819 --> 00:04:42.525
<v Jörn "Joe" Menninger>company possible because it lets you operate with a much smaller team.</v>

00:04:42.605 --> 00:04:48.487
<v Jörn "Joe" Menninger>What exactly would have required 4 people before and what is now handled by agents?</v>

00:04:48.567 --> 00:04:49.967
<v Delia Hallberg>Exactly.</v>

00:04:50.047 --> 00:04:55.145
<v Delia Hallberg>So our situation is that we started with 2 people, with 2 founders,</v>

00:04:55.225 --> 00:05:00.715
<v Delia Hallberg>and were actually able to go a very long way with only 2 people.</v>

00:05:00.795 --> 00:05:05.109
<v Delia Hallberg>But I think we as you already said, we really did</v>

00:05:05.189 --> 00:05:08.326
<v Delia Hallberg>the work of 4 or even more people.</v>

00:05:08.406 --> 00:05:13.896
<v Delia Hallberg>The first area in which we work closely with agents is the most obvious:</v>

00:05:13.976 --> 00:05:15.376
<v Delia Hallberg>marketing and content.</v>

00:05:15.456 --> 00:05:18.594
<v Delia Hallberg>I would say 80% are outsourced to AI.</v>

00:05:18.674 --> 00:05:22.203
<v Delia Hallberg>Agents hold our strategies, our content plans, our pipeline.</v>

00:05:22.283 --> 00:05:25.028
<v Delia Hallberg>And I do the Q&A and finalization.</v>

00:05:25.108 --> 00:05:29.422
<v Delia Hallberg>So then there is this research because we work across industries</v>

00:05:29.502 --> 00:05:33.816
<v Delia Hallberg>and sometimes we start a project in a new industry that</v>

00:05:33.896 --> 00:05:36.249
<v Delia Hallberg>we are not yet familiar with.</v>

00:05:36.329 --> 00:05:40.642
<v Delia Hallberg>So we do deep dives into sectors and client processes before</v>

00:05:40.722 --> 00:05:45.820
<v Delia Hallberg>we start a first conversation, which AI is great at helping us with.</v>

00:05:45.900 --> 00:05:50.214
<v Delia Hallberg>And then there is the sales part, obviously, that were able</v>

00:05:50.294 --> 00:05:54.608
<v Delia Hallberg>to automate quite a big workflow that basically makes sure that</v>

00:05:54.688 --> 00:05:59.002
<v Delia Hallberg>during the sales process we have quality assurance, every piece of</v>

00:05:59.082 --> 00:06:03.395
<v Delia Hallberg>data we can gather enters the process so that when this</v>

00:06:03.475 --> 00:06:08.573
<v Delia Hallberg>lead converts into a project, we do not start the project from zero,</v>

00:06:08.653 --> 00:06:14.536
<v Delia Hallberg>but we already have, for example, a project outline created by the previous sales agent.</v>

00:06:14.616 --> 00:06:18.537
<v Delia Hallberg>And then the fourth and most significant area is coding.</v>

00:06:18.617 --> 00:06:22.539
<v Delia Hallberg>This is where we really have the biggest efficiency gains.</v>

00:06:22.619 --> 00:06:24.580
<v Delia Hallberg>We have multiple coding agents.</v>

00:06:24.660 --> 00:06:28.973
<v Delia Hallberg>We have, I would say by now we are replacing an</v>

00:06:29.053 --> 00:06:34.936
<v Delia Hallberg>entire IT department with a very coherent system of agents and it's not vibe coding.</v>

00:06:35.016 --> 00:06:36.584
<v Delia Hallberg>Please don't misunderstand me.</v>

00:06:36.664 --> 00:06:41.370
<v Delia Hallberg>It is an elaborate process designed to ensure consistent, coherent code quality.</v>

00:06:41.450 --> 00:06:46.940
<v Delia Hallberg>Previously, that would have required maybe one additional person that would somehow do sales,</v>

00:06:47.020 --> 00:06:50.550
<v Delia Hallberg>research, marketing combined, and at least three additional developers.</v>

00:06:50.630 --> 00:06:54.159
<v Jörn "Joe" Menninger>When you talked about preparing for a new engagement,</v>

00:06:54.239 --> 00:07:00.122
<v Jörn "Joe" Menninger>you were talking about all the data you add before the client actually says yes.</v>

00:07:00.202 --> 00:07:03.731
<v Jörn "Joe" Menninger>I was wondering how, because it's from the outside,</v>

00:07:03.811 --> 00:07:09.301
<v Jörn "Joe" Menninger>how do you ensure sufficient quality that you don't have the typical system problem,</v>

00:07:09.381 --> 00:07:10.950
<v Jörn "Joe" Menninger>garbage in, garbage out?</v>

00:07:11.030 --> 00:07:12.991
<v Delia Hallberg>That's a very good point.</v>

00:07:13.071 --> 00:07:17.384
<v Delia Hallberg>So first of all, there is always a human in the</v>

00:07:17.464 --> 00:07:22.170
<v Delia Hallberg>loop and I always have to be able to understand the results.</v>

00:07:22.250 --> 00:07:24.211
<v Delia Hallberg>Otherwise, yeah, that's a risk.</v>

00:07:24.291 --> 00:07:28.605
<v Delia Hallberg>And then again, it's about setting up a system where you</v>

00:07:28.685 --> 00:07:33.391
<v Delia Hallberg>lower the risks of the AI producing low-value results or wrong results,</v>

00:07:33.471 --> 00:07:37.784
<v Delia Hallberg>which you can do by certain steps that you do while</v>

00:07:37.864 --> 00:07:42.178
<v Delia Hallberg>creating your agent system and not by simply throwing a prompt</v>

00:07:42.258 --> 00:07:45.003
<v Delia Hallberg>in ChatGPT and hoping for the best.</v>

00:07:45.083 --> 00:07:46.652
<v Jörn "Joe" Menninger>Hoping for the best.</v>

00:07:46.732 --> 00:07:51.045
<v Jörn "Joe" Menninger>Yeah, I think for some people out there it's still strategy.</v>

00:07:51.125 --> 00:07:55.439
<v Jörn "Joe" Menninger>Where is the boundary between a lean AI-enabled company and a</v>

00:07:55.519 --> 00:07:59.049
<v Jörn "Joe" Menninger>founder simply hiding the operational fragility behind some automation?</v>

00:07:59.129 --> 00:08:03.050
<v Delia Hallberg>That's a very relevant question, and, especially at this time.</v>

00:08:03.130 --> 00:08:07.444
<v Delia Hallberg>So I would say the boundary is where you understand what</v>

00:08:07.524 --> 00:08:13.014
<v Delia Hallberg>your agents are doing well enough to do it yourself if they stop tomorrow.</v>

00:08:13.094 --> 00:08:15.839
<v Delia Hallberg>Agents don't replace competence, they replace capacity.</v>

00:08:15.919 --> 00:08:20.233
<v Delia Hallberg>The judgment, the quality standard, the strategic thinking, this always has</v>

00:08:20.313 --> 00:08:24.235
<v Delia Hallberg>to stay with the humans, in that case with me.</v>

00:08:24.315 --> 00:08:28.628
<v Delia Hallberg>And I always think about it it's very tricky because AI</v>

00:08:28.708 --> 00:08:33.022
<v Delia Hallberg>is so good at producing plausible output that you can fool</v>

00:08:33.102 --> 00:08:38.200
<v Delia Hallberg>yourself for much longer before you realize that you've actually lost the plot.</v>

00:08:38.280 --> 00:08:42.986
<v Delia Hallberg>That's also something that we really take care of with our clients,</v>

00:08:43.066 --> 00:08:49.066
<v Delia Hallberg>with the tools that we create for them, that they always stay in the driver's seat.</v>

00:08:49.146 --> 00:08:52.283
<v Delia Hallberg>We do a few things to manage that.</v>

00:08:52.363 --> 00:08:57.069
<v Delia Hallberg>First, as I already said, there's always a human in the loop.</v>

00:08:57.149 --> 00:09:00.678
<v Delia Hallberg>Second, we document agent architectures just like software architectures.</v>

00:09:00.758 --> 00:09:05.464
<v Delia Hallberg>It needs to be scalable and not just living inside my head.</v>

00:09:05.544 --> 00:09:09.466
<v Delia Hallberg>And third, I'm really honest about what AI can't do.</v>

00:09:09.546 --> 00:09:13.860
<v Delia Hallberg>For example, client relationships, reading the room, judgment calls in whether</v>

00:09:13.940 --> 00:09:16.685
<v Delia Hallberg>and how a project needs to shift.</v>

00:09:16.765 --> 00:09:21.078
<v Delia Hallberg>And I would say an indicator of operational fragility is when</v>

00:09:21.158 --> 00:09:25.472
<v Delia Hallberg>founders or people talk about AI replacing this and that department,</v>

00:09:25.552 --> 00:09:31.552
<v Delia Hallberg>AI replacing this and that person because they are thinking about AI in terms of replacement.</v>

00:09:31.632 --> 00:09:37.632
<v Delia Hallberg>I think the discussion that we should be having is AI in the sense of amplification.</v>

00:09:37.712 --> 00:09:42.026
<v Delia Hallberg>The competence was there first and then AI makes it scale.</v>

00:09:42.106 --> 00:09:46.812
<v Jörn "Joe" Menninger>Not only that I have learned that AI agents drift over time.</v>

00:09:46.892 --> 00:09:51.205
<v Jörn "Joe" Menninger>So maybe at the start you get a marketing department and</v>

00:09:51.285 --> 00:09:57.285
<v Jörn "Joe" Menninger>when you are towards the end of the year, the tool only does social media anymore.</v>

00:09:57.365 --> 00:10:03.248
<v Jörn "Joe" Menninger>So I don't think there's right now a case for complete AI replacement for humans.</v>

00:10:03.328 --> 00:10:08.818
<v Jörn "Joe" Menninger>So as you said, you always have to have a human in the loop.</v>

00:10:08.898 --> 00:10:13.996
<v Jörn "Joe" Menninger>Your public positioning says Amadeni, your company, starts with a workshop, process analysis,</v>

00:10:14.076 --> 00:10:17.605
<v Jörn "Joe" Menninger>and then decides whether custom software is actually needed.</v>

00:10:17.685 --> 00:10:23.176
<v Jörn "Joe" Menninger>What is your decision rule for saying nope, you should not build software here?</v>

00:10:23.256 --> 00:10:28.354
<v Delia Hallberg>So it's important to say we always look for a standard solution first,</v>

00:10:28.434 --> 00:10:32.747
<v Delia Hallberg>because if you have a standard business case, and then there</v>

00:10:32.827 --> 00:10:37.141
<v Delia Hallberg>is a SaaS product that genuinely fits, then I would say</v>

00:10:37.221 --> 00:10:39.966
<v Delia Hallberg>that this is the way to go.</v>

00:10:40.046 --> 00:10:43.184
<v Delia Hallberg>That's part of being an honest IT consultancy.</v>

00:10:43.264 --> 00:10:47.577
<v Delia Hallberg>But what I've observed in the Mittelstand, especially in Eastern Germany,</v>

00:10:47.657 --> 00:10:51.187
<v Delia Hallberg>is that businesses that have been established for 20,</v>

00:10:51.267 --> 00:10:55.973
<v Delia Hallberg>maybe 30 years are the ones that are sitting in a niche.</v>

00:10:56.053 --> 00:10:59.190
<v Delia Hallberg>And they're at the intersection of 2 industries.</v>

00:10:59.270 --> 00:11:05.152
<v Delia Hallberg>So what happens quite frequently is that there is no proper software solution for them.</v>

00:11:05.232 --> 00:11:09.546
<v Delia Hallberg>What they have done until now is they have tried to</v>

00:11:09.626 --> 00:11:13.940
<v Delia Hallberg>use software maybe from an industry that they're slightly part of.</v>

00:11:14.020 --> 00:11:18.333
<v Delia Hallberg>But this is creating so much problems than in daily processes.</v>

00:11:18.413 --> 00:11:24.413
<v Delia Hallberg>That for example, a straightforward case where we go, okay, custom development really is the answer.</v>

00:11:24.493 --> 00:11:28.807
<v Delia Hallberg>And one signal that is growing stronger and stronger, for example,</v>

00:11:28.887 --> 00:11:33.593
<v Delia Hallberg>with an engineering firm in Berlin that we are currently working with,</v>

00:11:33.673 --> 00:11:37.987
<v Delia Hallberg>they were very happy with their SaaS solution, with their sort</v>

00:11:38.067 --> 00:11:42.380
<v Delia Hallberg>of project management solution, but they had been asking for a</v>

00:11:42.460 --> 00:11:45.205
<v Delia Hallberg>long time for a very specific feature.</v>

00:11:45.285 --> 00:11:46.854
<v Delia Hallberg>They didn't get it.</v>

00:11:46.934 --> 00:11:48.895
<v Delia Hallberg>They didn't get an API.</v>

00:11:48.975 --> 00:11:54.465
<v Delia Hallberg>So this, I would say, minor issue is really what caused them to say,</v>

00:11:54.545 --> 00:11:58.075
<v Delia Hallberg>okay, we have to say goodbye to our solution.</v>

00:11:58.155 --> 00:12:02.076
<v Delia Hallberg>And they proactively decided to go for a custom development.</v>

00:12:02.156 --> 00:12:08.156
<v Delia Hallberg>Because if a SaaS product doesn't have an API, it means that nobody can add automation.</v>

00:12:08.236 --> 00:12:09.805
<v Delia Hallberg>There is no flexibility.</v>

00:12:09.885 --> 00:12:13.022
<v Delia Hallberg>And this is something that surprised even me.</v>

00:12:13.102 --> 00:12:15.847
<v Delia Hallberg>This is really coming from our clients.</v>

00:12:15.927 --> 00:12:19.456
<v Delia Hallberg>So the decision tree, I would say, is simple.</v>

00:12:19.536 --> 00:12:22.674
<v Delia Hallberg>Is there a good standard solution that exists?</v>

00:12:22.754 --> 00:12:24.715
<v Delia Hallberg>If yes, go for it.</v>

00:12:24.795 --> 00:12:29.500
<v Delia Hallberg>If no, then custom software is really becoming a very viable alternative.</v>

00:12:29.580 --> 00:12:32.325
<v Delia Hallberg>One point is especially important for clients.</v>

00:12:32.405 --> 00:12:37.896
<v Delia Hallberg>If they don't have an API or some way to interact with their data,</v>

00:12:37.976 --> 00:12:40.721
<v Delia Hallberg>then they're also going for the alternatives.</v>

00:12:40.801 --> 00:12:45.115
<v Jörn "Joe" Menninger>Yes, we should always keep in mind it is not only</v>

00:12:45.195 --> 00:12:49.508
<v Jörn "Joe" Menninger>a question of standard software or not, but also is the</v>

00:12:49.588 --> 00:12:52.333
<v Jörn "Joe" Menninger>standard software breaking the bank or not?</v>

00:12:52.413 --> 00:12:55.943
<v Jörn "Joe" Menninger>I think you know what I'm talking about here.</v>

00:12:56.023 --> 00:12:59.944
<v Jörn "Joe" Menninger>I've been going through a little bit of your data,</v>

00:13:00.024 --> 00:13:04.338
<v Jörn "Joe" Menninger>and in your Impuls KBB case, the reported gain was from</v>

00:13:04.418 --> 00:13:09.908
<v Jörn "Joe" Menninger>8 hours to 2 hours for development reports and a 4-week return on investment.</v>

00:13:09.988 --> 00:13:14.302
<v Jörn "Joe" Menninger>What had to be true operationally for that result to be</v>

00:13:14.382 --> 00:13:17.127
<v Jörn "Joe" Menninger>real rather than just a software demo?</v>

00:13:17.207 --> 00:13:21.129
<v Delia Hallberg>I really like it that you're looking at the numbers.</v>

00:13:21.209 --> 00:13:23.954
<v Delia Hallberg>So let's look at the whole picture.</v>

00:13:24.034 --> 00:13:28.347
<v Delia Hallberg>The four-week ROI and calculated savings of €21,000 per month did</v>

00:13:28.427 --> 00:13:33.525
<v Delia Hallberg>not come from one process, this efficiency increase from 8 to 2 hours.</v>

00:13:33.605 --> 00:13:37.919
<v Delia Hallberg>As I mentioned a few questions earlier, went through the entire</v>

00:13:37.999 --> 00:13:42.313
<v Delia Hallberg>process chain starting with case clarification from the youth welfare office,</v>

00:13:42.393 --> 00:13:45.138
<v Delia Hallberg>really to the very end of invoicing.</v>

00:13:45.218 --> 00:13:49.924
<v Delia Hallberg>And we really looked at every step where we can increase efficiency.</v>

00:13:50.004 --> 00:13:54.318
<v Delia Hallberg>Across the entire process, there was so much room for improvement</v>

00:13:54.398 --> 00:14:00.280
<v Delia Hallberg>that in that case were really able to achieve this, I would say, impressive ROI.</v>

00:14:00.360 --> 00:14:04.282
<v Delia Hallberg>So first the client made a very, I would say,</v>

00:14:04.362 --> 00:14:08.675
<v Delia Hallberg>bold move at the time because we had already started the</v>

00:14:08.755 --> 00:14:14.245
<v Delia Hallberg>project and soon after the project began they decided to replace their old software.</v>

00:14:14.325 --> 00:14:17.855
<v Delia Hallberg>They saw that it did not have an API.</v>

00:14:17.935 --> 00:14:22.249
<v Delia Hallberg>Certain features in their previous software were causing a lot of</v>

00:14:22.329 --> 00:14:26.642
<v Delia Hallberg>problems and they decided not to supplement it but replace it.</v>

00:14:26.722 --> 00:14:29.467
<v Delia Hallberg>And that unlocked the enormous automation potential.</v>

00:14:29.547 --> 00:14:33.861
<v Delia Hallberg>For example, the voice-first interactions, because these are social welfare workers</v>

00:14:33.941 --> 00:14:37.471
<v Delia Hallberg>that go to the homes of children or youth.</v>

00:14:37.551 --> 00:14:41.472
<v Delia Hallberg>And then afterwards they have to sit in their car,</v>

00:14:41.552 --> 00:14:46.258
<v Delia Hallberg>open the laptop and type up documentation about what they discussed Now,</v>

00:14:46.338 --> 00:14:50.652
<v Delia Hallberg>on their way back to the car, they can simply make</v>

00:14:50.732 --> 00:14:55.045
<v Delia Hallberg>a voice message which then gets transcribed and then translated into</v>

00:14:55.125 --> 00:14:59.047
<v Delia Hallberg>official bureaucratic language and then placed into an automated workflow.</v>

00:14:59.127 --> 00:15:03.833
<v Delia Hallberg>Second, we built the system for how the social workers actually work.</v>

00:15:03.913 --> 00:15:07.835
<v Delia Hallberg>As I said in the example of the voice notes,</v>

00:15:07.915 --> 00:15:12.228
<v Delia Hallberg>They also had a lot of unofficial communication about what was</v>

00:15:12.308 --> 00:15:17.798
<v Delia Hallberg>the status quo of this document and have you thought about this and that?</v>

00:15:17.878 --> 00:15:22.192
<v Delia Hallberg>So we took a lot from that informal communication that created</v>

00:15:22.272 --> 00:15:27.762
<v Delia Hallberg>really a lot of inefficiency and put it into a process within the software,</v>

00:15:27.842 --> 00:15:30.587
<v Delia Hallberg>which eliminates a lot of wasted time.</v>

00:15:30.667 --> 00:15:34.981
<v Delia Hallberg>And then third, we integrated it with the rest of the</v>

00:15:35.061 --> 00:15:38.198
<v Delia Hallberg>system because they had a separate HR system.</v>

00:15:38.278 --> 00:15:42.200
<v Delia Hallberg>So they had to do 2 types of time tracking.</v>

00:15:42.280 --> 00:15:47.378
<v Delia Hallberg>And we said, no, the system has to work for the social workers.</v>

00:15:47.458 --> 00:15:49.419
<v Delia Hallberg>They need only one interface.</v>

00:15:49.499 --> 00:15:53.813
<v Delia Hallberg>And in the background, we will create a flow of data</v>

00:15:53.893 --> 00:15:56.638
<v Delia Hallberg>so that they have only one touchpoint.</v>

00:15:56.718 --> 00:16:02.208
<v Delia Hallberg>And I would say this entire process optimization built or visualized in the tool,</v>

00:16:02.288 --> 00:16:05.033
<v Delia Hallberg>that's what really created this four-week ROI.</v>

00:16:05.113 --> 00:16:09.427
<v Jörn "Joe" Menninger>I had to smile when you said that you make a</v>

00:16:09.507 --> 00:16:13.428
<v Jörn "Joe" Menninger>voice memo and then it gets translated into bureaucratic language.</v>

00:16:13.508 --> 00:16:18.998
<v Jörn "Joe" Menninger>It's— that's really a good point, a translator between bureaucrats and normal human beings.</v>

00:16:19.078 --> 00:16:21.824
<v Jörn "Joe" Menninger>But let's go a little bit further.</v>

00:16:21.904 --> 00:16:25.433
<v Jörn "Joe" Menninger>When a client has 20 systems, 40 media discontinuities,</v>

00:16:25.513 --> 00:16:29.042
<v Jörn "Joe" Menninger>and 35 hours of avoidable duplicate work per week.</v>

00:16:29.122 --> 00:16:32.652
<v Jörn "Joe" Menninger>How do you decide what you will fix first?</v>

00:16:32.732 --> 00:16:37.045
<v Delia Hallberg>Yeah, so the numbers that you quote come from a case,</v>

00:16:37.125 --> 00:16:40.655
<v Delia Hallberg>from a bakery that produces its own sourdough goods.</v>

00:16:40.735 --> 00:16:45.441
<v Delia Hallberg>That's actually important for the process because sourdough makes things very complicated.</v>

00:16:45.521 --> 00:16:50.227
<v Delia Hallberg>But the CEO called us because they had a bad gut feeling.</v>

00:16:50.307 --> 00:16:54.620
<v Delia Hallberg>So we did an analysis in a 2-day workshop and this</v>

00:16:54.700 --> 00:17:00.191
<v Delia Hallberg>bad gut feeling translated into these numbers that are quite shocking or quite surprising,</v>

00:17:00.271 --> 00:17:01.671
<v Delia Hallberg>I would say.</v>

00:17:01.751 --> 00:17:04.104
<v Delia Hallberg>And I'm very open about it.</v>

00:17:04.184 --> 00:17:09.282
<v Delia Hallberg>We were a little bit overwhelmed at first, but what is really important,</v>

00:17:09.362 --> 00:17:13.675
<v Delia Hallberg>and I think that's great that the CEO actually took that</v>

00:17:13.755 --> 00:17:16.893
<v Delia Hallberg>step because you don't have to automate everything.</v>

00:17:16.973 --> 00:17:21.286
<v Delia Hallberg>You don't have to optimize for everything, but you need to</v>

00:17:21.366 --> 00:17:27.249
<v Delia Hallberg>prioritize and have a plan for how you want to deal with the other things.</v>

00:17:27.329 --> 00:17:30.074
<v Delia Hallberg>First, we looked for ticking time bombs.</v>

00:17:30.154 --> 00:17:32.115
<v Delia Hallberg>That's how I call them.</v>

00:17:32.195 --> 00:17:37.293
<v Delia Hallberg>So something that could potentially go off any day and threaten the business.</v>

00:17:37.373 --> 00:17:41.686
<v Delia Hallberg>So if you find a ticking time bomb, then that's obviously</v>

00:17:41.766 --> 00:17:44.511
<v Delia Hallberg>at the very top of your roadmap.</v>

00:17:44.591 --> 00:17:48.513
<v Delia Hallberg>And then you deal with it regardless of everything else.</v>

00:17:48.593 --> 00:17:52.907
<v Delia Hallberg>In that case, we did not find any ticking time bombs.</v>

00:17:52.987 --> 00:17:56.516
<v Delia Hallberg>But then you look at the fundamental structural things.</v>

00:17:56.596 --> 00:18:01.694
<v Delia Hallberg>In that case, it was actually that they didn't have a leading system.</v>

00:18:01.774 --> 00:18:05.304
<v Delia Hallberg>which led to a lot of distributed data flows.</v>

00:18:05.384 --> 00:18:09.697
<v Delia Hallberg>And so it was very clear that we first had to</v>

00:18:09.777 --> 00:18:14.091
<v Delia Hallberg>build a very sound foundation, like a building, and you don't</v>

00:18:14.171 --> 00:18:16.524
<v Delia Hallberg>want to build it on sand.</v>

00:18:16.604 --> 00:18:18.957
<v Delia Hallberg>And then we looked at impact.</v>

00:18:19.037 --> 00:18:23.351
<v Delia Hallberg>So where did we have the lowest investment or where could</v>

00:18:23.431 --> 00:18:27.352
<v Delia Hallberg>we create the greatest business impact with the lowest effort?</v>

00:18:27.432 --> 00:18:30.177
<v Delia Hallberg>Then you have a prioritization, a roadmap.</v>

00:18:30.257 --> 00:18:35.747
<v Delia Hallberg>And as I said initially, it's totally okay if you don't tackle every process,</v>

00:18:35.827 --> 00:18:39.749
<v Delia Hallberg>the important difference is that you are aware of them,</v>

00:18:39.829 --> 00:18:42.966
<v Delia Hallberg>have quantified them and you have a plan.</v>

00:18:43.046 --> 00:18:47.360
<v Delia Hallberg>And I think that makes a difference between a vague gut</v>

00:18:47.440 --> 00:18:49.793
<v Delia Hallberg>feeling and having worked through it.</v>

00:18:49.873 --> 00:18:55.873
<v Jörn "Joe" Menninger>Yeah, you just have to have a framework built on numbers that you can add to.</v>

00:18:55.953 --> 00:18:59.482
<v Jörn "Joe" Menninger>I was wondering, talking about numbers and metrics here,</v>

00:18:59.562 --> 00:19:03.876
<v Jörn "Joe" Menninger>what is the metric you trust most when evaluating whether an</v>

00:19:03.956 --> 00:19:06.309
<v Jörn "Joe" Menninger>AI automation project is worth building?</v>

00:19:06.389 --> 00:19:11.487
<v Jörn "Joe" Menninger>Do you think about hours saved, reduction in errors, adoption rate, revenue impact,</v>

00:19:11.567 --> 00:19:13.136
<v Jörn "Joe" Menninger>or maybe management relief?</v>

00:19:13.216 --> 00:19:17.529
<v Delia Hallberg>I would say I try to collapse all these things that</v>

00:19:17.609 --> 00:19:21.923
<v Delia Hallberg>you just said into a single number because at the end</v>

00:19:22.003 --> 00:19:26.317
<v Delia Hallberg>you can put a price tag on nearly all those things.</v>

00:19:26.397 --> 00:19:31.887
<v Delia Hallberg>What is it costing you right now in terms of time, errors, management firefighting,</v>

00:19:31.967 --> 00:19:34.712
<v Delia Hallberg>but also what would the solution cost?</v>

00:19:34.792 --> 00:19:39.498
<v Delia Hallberg>So if the solution has an ROI of less than six months,</v>

00:19:39.578 --> 00:19:43.500
<v Delia Hallberg>then it's a very straightforward decision to go for it.</v>

00:19:43.580 --> 00:19:48.678
<v Delia Hallberg>If it is 12 or 18 months, we have a more nuanced conversation,</v>

00:19:48.758 --> 00:19:54.248
<v Delia Hallberg>and it's also more about strategy, about potential risks also of not acting, right?</v>

00:19:54.328 --> 00:19:59.818
<v Delia Hallberg>But I would say there is a very big misconception that SMBs frequently have.</v>

00:19:59.898 --> 00:20:04.212
<v Delia Hallberg>It's a conversation that I have with a lot of especially</v>

00:20:04.292 --> 00:20:09.390
<v Delia Hallberg>potential clients at the beginning is they treat IT as its own budget.</v>

00:20:09.470 --> 00:20:14.960
<v Delia Hallberg>So for example, a €25,000 project looks like it would blow the IT budget,</v>

00:20:15.040 --> 00:20:18.177
<v Delia Hallberg>but and that's the difference from my perspective.</v>

00:20:18.257 --> 00:20:22.571
<v Delia Hallberg>You have to calculate the IT budget in terms of personnel</v>

00:20:22.651 --> 00:20:26.965
<v Delia Hallberg>costs of the inefficient process, and then when you compare the</v>

00:20:27.045 --> 00:20:31.358
<v Delia Hallberg>personnel costs versus the investment into optimizing these processes and the</v>

00:20:31.438 --> 00:20:37.321
<v Delia Hallberg>potential personnel costs that you save, not because you're letting go of people, but because.</v>

00:20:37.401 --> 00:20:42.891
<v Delia Hallberg>Suddenly you can come from a very reactive business to a more strategic one,</v>

00:20:42.971 --> 00:20:46.500
<v Delia Hallberg>or your people can actually do more meaningful things.</v>

00:20:46.580 --> 00:20:51.286
<v Delia Hallberg>That reframing is one of the most important conversations that we have,</v>

00:20:51.366 --> 00:20:54.504
<v Delia Hallberg>and then suddenly the decision-making becomes much easier.</v>

00:20:54.584 --> 00:20:55.984
<v Jörn "Joe" Menninger>The decision-making.</v>

00:20:56.064 --> 00:20:57.464
<v Jörn "Joe" Menninger>making becomes easier.</v>

00:20:57.544 --> 00:20:59.112
<v Jörn "Joe" Menninger>Talk about processes here.</v>

00:20:59.192 --> 00:21:05.075
<v Jörn "Joe" Menninger>You argue that a a poor process plus AI becomes a poor process on steroids.</v>

00:21:05.155 --> 00:21:10.645
<v Jörn "Joe" Menninger>What is the clearest sign that process must be redesigned before it's even automated?</v>

00:21:10.725 --> 00:21:13.078
<v Delia Hallberg>There is a very clear sign.</v>

00:21:13.158 --> 00:21:18.256
<v Delia Hallberg>And it is when nobody can explain the process without naming a person.</v>

00:21:18.336 --> 00:21:22.649
<v Delia Hallberg>So if people start saying, so I get the invoice and</v>

00:21:22.729 --> 00:21:27.827
<v Delia Hallberg>then Sabine does this and that and then Thomas does that for example,</v>

00:21:27.907 --> 00:21:31.045
<v Delia Hallberg>that's not a process, those are personal habits.</v>

00:21:31.125 --> 00:21:35.438
<v Delia Hallberg>And we don't want to automate people's personal habits because if</v>

00:21:35.518 --> 00:21:41.518
<v Delia Hallberg>you automate that and then the moment Sabine goes on holiday, the system doesn't work anymore.</v>

00:21:41.598 --> 00:21:45.912
<v Delia Hallberg>So especially in the Mittelstand, that is a process that we</v>

00:21:45.992 --> 00:21:50.698
<v Delia Hallberg>go through to get away from this person name-driven thinking to, okay,</v>

00:21:50.778 --> 00:21:52.739
<v Delia Hallberg>what are the actual steps?</v>

00:21:52.819 --> 00:21:57.525
<v Delia Hallberg>A second straightforward sign is if information gets typed more than twice.</v>

00:21:57.605 --> 00:21:59.565
<v Delia Hallberg>A third is, shadow systems.</v>

00:21:59.645 --> 00:22:04.351
<v Delia Hallberg>Shadow systems—shadow IT—are a very strong indicator that the process isn't working.</v>

00:22:04.431 --> 00:22:08.745
<v Delia Hallberg>And if you automate that process without integrating the shadow systems</v>

00:22:08.825 --> 00:22:13.139
<v Delia Hallberg>or without looking at why people had to create shadow systems,</v>

00:22:13.219 --> 00:22:17.140
<v Delia Hallberg>then you'll create a tool that nobody's going to use.</v>

00:22:17.220 --> 00:22:21.534
<v Delia Hallberg>I'm talking about private Excel files where the real numbers are,</v>

00:22:21.614 --> 00:22:25.536
<v Delia Hallberg>or the WhatsApp group that's the actual project management tool.</v>

00:22:25.616 --> 00:22:29.929
<v Delia Hallberg>So these are indicators where it's very clear that you have</v>

00:22:30.009 --> 00:22:32.362
<v Delia Hallberg>to redesign first and automate second.</v>

00:22:32.442 --> 00:22:36.756
<v Jörn "Joe" Menninger>Yes, I was just smiling talking about shadow IT and WhatsApp</v>

00:22:36.836 --> 00:22:39.973
<v Jörn "Joe" Menninger>groups that actually are used for project management.</v>

00:22:40.053 --> 00:22:45.936
<v Jörn "Joe" Menninger>I have seen that and in my experience, startups tend to do this more frequently.</v>

00:22:46.016 --> 00:22:50.722
<v Jörn "Joe" Menninger>There was a heyday for Slack, but I think it's partly gone.</v>

00:22:50.802 --> 00:22:53.547
<v Delia Hallberg>WhatsApp has remained popular in the Mittelstand.</v>

00:22:53.627 --> 00:22:56.764
<v Jörn "Joe" Menninger>Yeah, maybe it just, it just moved there.</v>

00:22:56.844 --> 00:22:58.244
<v Jörn "Joe" Menninger>That's also fine.</v>

00:22:58.324 --> 00:23:03.814
<v Jörn "Joe" Menninger>Talk a little bit about more cases that I've seen from you for example,</v>

00:23:03.894 --> 00:23:05.463
<v Jörn "Joe" Menninger>the Ostdeutscher Bankenverband case.</v>

00:23:05.543 --> 00:23:09.856
<v Jörn "Joe" Menninger>The promise was source traceability and zero hallucination for an AI.</v>

00:23:09.936 --> 00:23:15.819
<v Jörn "Joe" Menninger>What architectural choice prevents hallucination by design rather than by, yeah, you know, just hope?</v>

00:23:15.899 --> 00:23:19.820
<v Delia Hallberg>Yeah, hope and prayers are not something that you want,</v>

00:23:19.900 --> 00:23:23.038
<v Delia Hallberg>in your system design or in your IT.</v>

00:23:23.118 --> 00:23:27.431
<v Delia Hallberg>so what we did is, I think the key is obviously</v>

00:23:27.511 --> 00:23:31.825
<v Delia Hallberg>the key is not literally zero hallucinations, but lowering the risk</v>

00:23:31.905 --> 00:23:35.042
<v Delia Hallberg>and having a system that can catch hallucinations.</v>

00:23:35.122 --> 00:23:39.436
<v Delia Hallberg>Because in that case with the Ostdeutscher Bankenverband, they provide information</v>

00:23:39.516 --> 00:23:44.222
<v Delia Hallberg>to their members that is crucial that their members base decisions on.</v>

00:23:44.302 --> 00:23:49.400
<v Delia Hallberg>And if you have a hallucinated piece of information, that's reputation damaging, right?</v>

00:23:49.480 --> 00:23:54.578
<v Delia Hallberg>So what we did is obviously human in the loop at every step.</v>

00:23:54.658 --> 00:24:00.148
<v Delia Hallberg>Then second, we gave different, unconnected LLMs small tasks, because the smaller the task,</v>

00:24:00.228 --> 00:24:05.326
<v Delia Hallberg>the more concrete the information, the less the AI is forced to hallucinate.</v>

00:24:05.406 --> 00:24:06.975
<v Delia Hallberg>When does AI hallucinate?</v>

00:24:07.055 --> 00:24:12.545
<v Delia Hallberg>When models lack information but still have to come up with an answer, right?</v>

00:24:12.625 --> 00:24:16.939
<v Delia Hallberg>And so the answer was to cut the task very small</v>

00:24:17.019 --> 00:24:20.940
<v Delia Hallberg>and to include in the system a traceability of information.</v>

00:24:21.020 --> 00:24:25.334
<v Delia Hallberg>So every piece of information that I had, I could trace</v>

00:24:25.414 --> 00:24:29.728
<v Delia Hallberg>back to the original website, to the original source of information.</v>

00:24:29.808 --> 00:24:32.945
<v Delia Hallberg>And this was especially important for the humans.</v>

00:24:33.025 --> 00:24:38.515
<v Delia Hallberg>who were in the loop to be able to prove every piece of information.</v>

00:24:38.595 --> 00:24:42.909
<v Delia Hallberg>I would say that these two or three design decisions lowered</v>

00:24:42.989 --> 00:24:45.342
<v Delia Hallberg>the risk of hallucinations very much.</v>

00:24:45.422 --> 00:24:49.736
<v Jörn "Joe" Menninger>So you give the AI a very small area of responsibility</v>

00:24:49.816 --> 00:24:54.129
<v Jörn "Joe" Menninger>at each step and enough information to fulfill it in order</v>

00:24:54.209 --> 00:24:56.562
<v Jörn "Joe" Menninger>that they Don't need to hallucinate.</v>

00:24:56.642 --> 00:24:58.042
<v Jörn "Joe" Menninger>Great.</v>

00:24:58.122 --> 00:25:01.260
<v Jörn "Joe" Menninger>Delia, it was a pleasure talking to you.</v>

00:25:01.340 --> 00:25:06.045
<v Jörn "Joe" Menninger>We agreed that we're going to cut this interview into 2 pieces.</v>

00:25:06.125 --> 00:25:10.831
<v Jörn "Joe" Menninger>We'll be back with number 2 for you in a few days,</v>

00:25:10.911 --> 00:25:13.264
<v Jörn "Joe" Menninger>for us in a few minutes.</v>

00:25:13.344 --> 00:25:14.913
<v Jörn "Joe" Menninger>Talk to you soon.</v>

00:25:14.993 --> 00:25:16.393
<v Jörn "Joe" Menninger>Thank you.</v>

00:25:16.473 --> 00:25:17.873
<v Jörn "Joe" Menninger>Bye.</v>

00:25:17.953 --> 00:25:19.353
<v Delia Hallberg>Bye.</v>
