Masters of Search Conference 2027
← All episodes
Masters of Search episode 50: We built our own agency OS with Claude Code | Artur Kosch, Co-CEO @ KKP cover art
EP 50·Nov 20, 2026·3711

We built our own agency OS with Claude Code | Artur Kosch, Co-CEO @ KKP

Show notes

How KKP runs a finance-only search agency on Claude Code

Artur Kosch is co-founder and Co-CEO of KKP, a search agency from Saarbrücken that works only for finance and insurance brands like Vanguard, Raisin and DA Direct. What I find most interesting about KKP is how they work. Artur and his team built their own operating system, the KKP OS, on Claude Code. The whole team works in it, with AI agents plugged into their tools, and Artur says it makes them twice as fast as a year ago.

So we opened it up. Why they built it instead of buying something, what a consultant's day looks like inside it, and which tools they cancelled along the way. We also talked about what bank and insurance legal teams ask when an agency works with AI, what they had to rip out again, and why an in-house team does not need an agency to get faster. And we talked about the question every agency is asking right now: if AI does the junior work, how does anyone become a senior?

My 6 biggest takeaways from our conversation:

1. With no-code automation you have to know every step. With an agent, you only have to know the goal.

KKP started where most teams did: ChatGPT, then custom GPTs, then automations in Make and n8n. The turn came when a colleague from their paid team showed Artur what you can build with Claude Code, skills and your own knowledge base. In n8n you define every step and connection. In the KKP OS you say what you want. A consultant's day starts with a morning briefing: the OS scans their emails, Slack and Teams messages, calendar, CRM and project management, flags a client email that sounds worried, points to a new analysis in the Peec AI workspace and suggests what can wait until next week. From there a content workflow pulls the client's prompts from Peec AI, search data from DataForSEO and the brand voice, and pushes the result to a dashboard each client can log into.

2. Build only what no tool does for you, and cancel every tool you no longer open.

KKP still uses Teams and HubSpot, because those do the job. They built their own client dashboard and moved their website from WordPress to code written with Claude Code. Security they left to a partner agency, because KKP has no software engineers. Artur says his own work is now about 90% MCP and API and 10% user interface. KKP cancelled a lot of tools because of DataForSEO, whose MCP he puts at around 50 dollars a month. They keep Sistrix for classic numbers like the visibility index and for its prompt research module, and use Peec AI for about 95% of their AI search work. They did build their own AI visibility reporting with Claude Code, but paying Peec AI is cheaper than paying the API costs of every model. His rule for keeping a tool: watch your own behaviour, and cancel what you no longer open.

3. Bank and insurance legal teams don't ask whether you use AI. They ask who is accountable.

Artur says these talks are less dramatic than you would expect. KKP's contracts state openly that they use AI tools under editorial control. Legal teams want to know which data goes where and who signs off. The answer is always the same: a named person is responsible for every piece, and every number in a report is checked by hand against its source. No client data leaves without the client's approval.

4. The best sign an internal tool works is that the team builds most of it.

Artur checked the week we recorded: 59% of all deployments and changes in the KKP OS came from his team, not from him. He had expected to be the only one pushing it. Not everything worked. Pulse, the part of the OS meant to bring the agency's finances and data together, became so complex that Artur stopped using it, and Excel was better. After three months of building, they ripped it out.

5. An in-house team doesn't need an agency to get faster. It needs a boss who makes room for it.

If AI makes a good SEO twice as fast, why would anyone still hire an agency? Artur encourages companies to build their own SEO team, and says KKP gets the best results with clients who have one. If that team works and the KPIs go up, great. If the KPIs have been stuck for three years, that is when to call someone. For teams that want to move on their own, his advice is that the change has to come from the boss, with time, room and resources. Artur himself is not sure he still has Excel or PowerPoint installed, because Claude Code builds his tables and slides.

6. If AI does 95% of junior tasks, juniors have to be trained like apprentices.

Artur estimates that AI does about 95% of the tasks a junior used to do, faster, cheaper and more precisely. So he looks at junior hires like the German Ausbildung, a three-year apprenticeship where the company puts in more than it gets back. You hire three juniors and maybe one stays long enough to become a senior, and that risk belongs in your planning. For juniors, his advice is to build your own website, the way he and his co-founder learned SEO with affiliate sites on weekends next to their corporate jobs.

Also in this episode: what Artur still does not do in Claude, and why decisions stay with people. How 90% of his LinkedIn posts come out of Claude Code, starting from a team-fed database of AI search news. Whether passion is required to be in the top one or two percent. And the pressure the constant change puts on teams, and what leaders owe the people who are not as nerdy about it as we are. In Artur's words, the company can roll out a red carpet, but you still have to walk it yourself.

Let's connect

Niklas on LinkedIn: https://www.linkedin.com/in/niklas-buschner/
Radyant on LinkedIn: https://www.linkedin.com/company/radyant/
Artur on LinkedIn: https://www.linkedin.com/in/artur-kosch/
Check out KKP: https://kkp.agency
KKP's weekly AI Search Update newsletter (in German): https://kkp.agency/insights

Full conversation

via podigee
  • 00:00:00For me, it's not just like a tool.
  • 00:00:02It is transformation and working.
  • 00:00:04I work completely different than when i worked one year ago on my LinkedIn posts And all the things that I do in LinkedIn are done by Claude Code.
  • 00:00:13If Claude makes good SEO twice as fast The client's in-house team
  • 00:00:19could
  • 00:00:19potentially also get same boost from buying Claude.
  • 00:00:23Why they still need an agency?
  • 00:00:25I always encourage companies to do that, build your own SEO team.
  • 00:00:29If you can do it by yourself if we can do in-house and at work its great but with the KPIs stuck for three years call us.
  • 00:00:36maybe
  • 00:00:38they will help You!
  • 00:00:42When you talk to junior who have all of theoretical knowledge And then when compared to talking with experienced like with US you see immediately a difference.
  • 00:00:52Get the knowledge try to learn.
  • 00:00:55Before we dive in, you're listening to the Masters of Search podcast with me your host Niklas Buschner.
  • 00:01:01Each week I sit down with some of these smartest people around the world an SEO and AI search to bring their strategies mental models top pieces actionable advice.
  • 00:01:12if You enjoy this podcast don't forget like subscribe follow it on your favorite podcasting app or YouTube.
  • 00:01:19helps us get top-notch guests and create the best possible content for you.
  • 00:01:23Let's dive into todays episode!
  • 00:01:38My guest today is Artur Kosch, co-founder and Co-CEO of KKP a search agency from Saarbrücken in the wonderful Saarland Germany.
  • 00:01:48If you don't know it ,you should definitely visit at some point.
  • 00:01:52that works only for insurance & finance brands like Vanguard, Raisin, DA Direct.
  • 00:02:00The part of KKP I find and found most interesting also when i followed Artur is how they work.
  • 00:02:06Artur and his team built their own internal operating system, the KKP OS. The whole team works in it with AI agents plugged into their tools And Artur says that makes them two times faster than a year ago.
  • 00:02:19So today we'll open up why They build instead of buying something where broke.
  • 00:02:24what clients think Of It?
  • 00:02:29Welcome to the podcast, Artur.
  • 00:02:30Hey Niklas thanks for having me!
  • 00:02:32Do you feel like the introduction was a alright representation of the work that you do at KKP?
  • 00:02:39It's on point.
  • 00:02:41not even my AI agent could introduce me better.
  • 00:02:45You already stole my follow up question because I wanted to ask what did the AI agents say about it?
  • 00:02:51He said it is perfect and nothing too corrects.
  • 00:02:57on-point
  • 00:02:58Awesome.
  • 00:02:58Thank you so much, cool!
  • 00:03:01Yeah let's start with going back a little bit in time.
  • 00:03:06please take us back to the moment you decided that KKP needed its own operating system like what was going on in the agency at that point.
  • 00:03:16and Was there an aha moment where it said damnit we have to build something like this?
  • 00:03:23Yes Think like back then, almost everyone who's working in the search industry was trying to figure out how we can implement AI and our workflows.
  • 00:03:37And daily doings... We all started with ChatGPT.
  • 00:03:40The next part is custom GPTs.
  • 00:03:43I still remember that it wasn't even able to upload something if you copy-paste it three years ago.
  • 00:03:50Next step were automations with Make or n8n.
  • 00:03:54And then one of my colleagues from the paid department in our agency, she came up with something like... She read about a course.
  • 00:04:10Like to build your own brain in Claude Code with IDEs and skills and all that stuff.
  • 00:04:16I thought it was very interesting!
  • 00:04:18A week later he showed me what is possible All that stuff and it was the moment for me when clicked.
  • 00:04:26And I say, okay let's screw all that.
  • 00:04:28Make and n8n automations way better to automate workflows really get them most out of AI.
  • 00:04:39And that was the moment when we started building our KKP OS.
  • 00:04:43So yeah, it was Claude Code and all of our knowledge, workflows... We do daily in our agency to have our clients get more visibility on AI search helps us with Claude Code and all this stuff.
  • 00:05:02But what precisely made you feel n8n, Zapier, Make all these like no code flow building things didn't live up to your needs anymore.
  • 00:05:19Because still today a lot of people
  • 00:05:21built
  • 00:05:22AI agents would also be interested.
  • 00:05:25in your opinion if people call a Zapier automation an AI agent is that actually correct representation?
  • 00:05:32but what precisely made you feel This is something that we have outgrown and we have to go beyond that.
  • 00:05:39Yeah, I think how you used automatic tools like that was not for my... I would say naturally using it.
  • 00:05:52so For me the KKP OS we had a window.
  • 00:05:59Um, here type in your wish and I will like do it.
  • 00:06:05Uh And we've made n8n.
  • 00:06:08It's-it's like you have to Know exactly what do you want?
  • 00:06:12What's?
  • 00:06:12the steps are step one, step two and it disconnects only this API.
  • 00:06:16And with our KKP OS we just type in what you want or to wish for what.
  • 00:06:21your repeat workflow and With all the knowledge that knows when it grows with more even more we're gonna use is It just builds naturally without You know what's the task, but you know the goal.
  • 00:06:36And that was for me... The reason we said okay and all in automation tools are nice maybe for some purposes may be for some tasks But it is not a way to go.
  • 00:06:46We want to go.
  • 00:06:48I think when i looked back It wasn't right decision.
  • 00:06:53So because You also obviously have some expertise In finance if i would come To your KKP OS Telling hey How about you make me half a million euros?
  • 00:07:07Will it be able to fulfill my
  • 00:07:09wish?".
  • 00:07:10I don't think so, no.
  • 00:07:11Unfortunately not!
  • 00:07:13Okay okay
  • 00:07:14this is something that cannot do yet but let's talk about what it can already do.
  • 00:07:18maybe You can walk us through one workflow or Something that you find as very cool use case and how your team actually leveraging because i know A lot of people out there are currently building their operating systems, knowledge bases and tools.
  • 00:07:43I think it's good to have a real life.
  • 00:07:46look behind the scenes like what can you actually do?
  • 00:07:52I think that it's very, um... Very funny when you walk through our agency here and look at the screens.
  • 00:07:58You would say like even if see a call or you see Claude Code on every desk?
  • 00:08:05And i think nobody is working in Claude Code and maybe ten percent calls!
  • 00:08:15So let me give you small examples.
  • 00:08:17so for example a consultant opens his laptop, get his coffee or his tea and starts today.
  • 00:08:26May you know the good morning start in Claude Code?
  • 00:08:33And what it does is it scans all emails, scans all messages like Slack or Teams.
  • 00:08:40It scans calendar meetings... go to the CRM, project management.
  • 00:08:49To figure out exactly what is doing for consultants today?
  • 00:08:54For example it reads an email and says like oh some clients text you in e-mail... And sounds a little bit concerning.
  • 00:09:06may look deeper into this.
  • 00:09:10you have to get answer soon but it sounds like that.
  • 00:09:14and because the Our OS knows the client, he knows like how his text and you know what its purpose is.
  • 00:09:20And now your acts that our client gets to do list What?
  • 00:09:23He's uh...what she's looking for today in particular way or another part Is Like hey there was a new analyzes from the Peec AI workspace from the MCP and You may look into this This week and Other tasks.
  • 00:09:39maybe it's time For next Week.
  • 00:09:41so um Our consultants and everybody who's working with our OS gets a very good scan from all the amount we have to do, all day.
  • 00:09:52And focusing on really good tasks because the OS has access of data or information and knows what is important.
  • 00:10:02now maybe it isn't that important.
  • 00:10:05And for example, if the consultants say okay that's a very good point.
  • 00:10:10we have to produce these five content pieces because it is not very crucial.
  • 00:10:15Because data says and clients need them.
  • 00:10:18For example you can start your content production workflows one of biggest work flows built back then in the first part.
  • 00:10:28so what workflow does?
  • 00:10:33the prompts from Peec AI.
  • 00:10:35We do our own prompt research before, of course he gets data from DataForSEO.
  • 00:10:42we do all other analysis and text is produced by brand voice and all data for the client.
  • 00:10:55So every client has a dashboard.
  • 00:10:57You can log into it, you have all your data, your reportings, your Peec AI data, Search Console data... All you want is ranking keywords and you've got all the documents we work for.
  • 00:11:08so when clients are ready with their document or some delivery or analysis as they push to the client he gets a message like hey!
  • 00:11:17We have new data here.
  • 00:11:19that's the analysis.
  • 00:11:21if you wanna talk on our next call review That will be reviewed And after that, we're going to make some decisions.
  • 00:11:28So I think maybe a day in the consultancy of our agency is over.
  • 00:11:38Yeah so cool overview!
  • 00:11:39There was already so much food for thought.
  • 00:11:42my brain has already buzzing.
  • 00:11:46Let's talk about one thing first because there are lots to unpack.
  • 00:11:51Obviously, you have the OS.
  • 00:11:54It's something that you build for the agency and let the whole agency uses.
  • 00:11:58but already now mentioned a couple of external tools You still use.
  • 00:12:03so you mention Peec AI?
  • 00:12:04You mentioned DataForSEO.
  • 00:12:06How do you see this balance between your own tool that you built And then external tools?
  • 00:12:13So when Do you still use external tools and why?
  • 00:12:17how has Like software buying, decision process also changed as an agency owner.
  • 00:12:24When do you decide to buy something?
  • 00:12:26when Do You Decide To Build Something?
  • 00:12:29That's a very good question and for us it is like this tool doesn't fulfill our needs hundred percent.
  • 00:12:39I think that we consider maybe we can do better not in terms of building the best tools but It's more for our specific needs.
  • 00:12:52For example, we use Teams because I think teams are okay and don't want to build a new messaging tool or HubSpot.
  • 00:13:02is our CRM not necessary to build something else?
  • 00:13:06For example, the search intelligence we built is our dashboard for clients because they want to build specific dashboards and data.
  • 00:13:21that we think and there was our need.
  • 00:13:23Of course, you could use WordPress or something like that but... That's the point!
  • 00:13:30Another example is back then we used Wordpress for a website.
  • 00:13:33now it's fully HTML in JavaScript built with Claude Code because its way easier.
  • 00:13:39I can't remember back then when you just want to, new clients arrived and they have no marketing.
  • 00:13:45They had to search for the logo download the logo put it to Photoshop Then we have make the color out And then put on a website.
  • 00:13:54Now just type in.
  • 00:13:56please include this client logon website.
  • 00:13:58It's done in a minute Very faster.
  • 00:14:05So I would say if it's better and fulfill our needs more or If its a faster and cheaper for us, so think that the biggest reasons why we decide to build by own
  • 00:14:20Now around building the OS?
  • 00:14:22I suppose please correct me from wrong That you also built the OS with Claude Code but what i know you don't have.
  • 00:14:31or maybe now it changed, but you didn't have a dedicated software engineers and software developer in-house.
  • 00:14:40This is something that obviously not so new – people building their tools with Claude Code!
  • 00:14:46It's something that in the specific marketing bubble feels like common practice now….
  • 00:14:52But what I care about?
  • 00:14:56How did you implement things like security mechanisms governance ensuring that not everybody on the team.
  • 00:15:05Like can go beyond certain let's say best practices but sort of thing where use a?
  • 00:15:12hey this is our gold standard how we work in the agency because he also have A high quality standard.
  • 00:15:18from what I know, like how did you implement these?
  • 00:15:21Unsexy things and how.
  • 00:15:23so take us through the journey of You opening the first Claude Code session and starting the OS project until now.
  • 00:15:30Like What were also this Maybe let's call it boring unsexy but still very important parts.
  • 00:15:39Yeah as a good question.
  • 00:15:41So all tools we use its are just for our own usage.
  • 00:15:47so we know don't sell any software or try to make money with it.
  • 00:15:52So just for our own use, I would say already secure environment because you're not selling it... It's a way different thing.
  • 00:16:01and are we not planning too?
  • 00:16:04Not good as team!
  • 00:16:05Because you were very correct that there aren't engineers here.
  • 00:16:10but all the security stuff like i said is boring.
  • 00:16:15we said no, that's not our expertise.
  • 00:16:18We need the expert.
  • 00:16:20so we work with a partner agency who is specialized in this area like security logins databases and all of stuff whose critical or could be crucial And helps us with this.
  • 00:16:36So yeah, we have a partner tool to help you know for that stuff.
  • 00:16:41So good.
  • 00:16:41Did you your already?
  • 00:16:44I think partly answered that you have no intention or at least right now do not sell Something from the OS in itself, it's always something that enables us an agency.
  • 00:16:55But did you along the way also right now ever thought about hey this is something that maybe also provides value in itself?
  • 00:17:05Two clients that may be also cannot afford hiring you as an agency with the extra human power.
  • 00:17:13So how do you think about that turning something, which also solves obvious problems so it has value into something that could sell on its own?
  • 00:17:26I was never a friend of doing two things at the same time owning and leading an agency, then a side project like selling software.
  • 00:17:45I think they're too big to split that apart.
  • 00:17:53so maybe in the future when i would think okay it's very nice idea Yeah, the only thing I can imagine is like to step away from agency and then focusing on a software because I love what i do every day.
  • 00:18:16Every morning And focus like hundred percent in this in this agency?
  • 00:18:21And couldn't you mention two have like another thing Like that would be too much.
  • 00:18:25so all your reason will be like okay The idea's nice.
  • 00:18:28maybe it We can.
  • 00:18:30we can start up from here But not two things at the same time.
  • 00:18:37Would there be any amount of money or something?
  • 00:18:42Or external feedback where people tell you, hey this is so cool.
  • 00:18:49I heard about what your building like.
  • 00:18:50we also need that and yeah obviously i can try building it myself but It will take me probably a year or something.
  • 00:18:57So if anything were could see yourself revisiting the decision coming to different conclusion.
  • 00:19:04Yeah, I think so.
  • 00:19:08It's hard to tell because all we built is so specific for the client.
  • 00:19:12of course there are a lot more modules and things that were used... ...for every client but The things we build for their clients was also so specific that I think like maybe older All the analyzes data and maybe deliver words.
  • 00:19:30creation could be may be tool.
  • 00:19:34I could revisit my decision in maybe years or months, but at the moment there's no plans.
  • 00:19:46Okay!
  • 00:19:47And i don't want to talk you into it and i'm just genuinely curious because... ...i feel like there are some agency founders that operate on service businesses that always had this dream of building products and now obviously we are enabled to build product.
  • 00:20:08And it makes absolute sense to build products, let's solve our own problems first but then the obvious question is why not go out into market with its own?
  • 00:20:19But you gave a very convincing answer
  • 00:20:21for
  • 00:20:21your conclusion.
  • 00:20:24A lot of SEO tools just started, like you said.
  • 00:20:27It was a tool inside and now they are like Screaming Frog or write back then to start it that way.
  • 00:20:35Yeah That's good keyword pun intended here SEO Tools.
  • 00:20:43so Peec AI, DataForSEO.
  • 00:20:47these two.
  • 00:20:49So, Peec AI is a rather new tool.
  • 00:20:50DataForSEO.
  • 00:20:51I think it's not an really new tool but they have come to new fame due their like API nativeness and being able to implement in workflows with Claude etc.
  • 00:21:05How do you see the let us call them legacy SEO tools?
  • 00:21:09Like Semrush, Sistrix, Ahrefs?
  • 00:21:12Do you still use them?
  • 00:21:14if yes what do we still use then?
  • 00:21:16four And If no When did you cancel them and why?
  • 00:21:21Yeah, it's a good question.
  • 00:21:22So we cancelled a lot of tools because of DataForSEO And you mentioned that they are now pretty famous.
  • 00:21:34I don't think.
  • 00:21:34back then there was just some nerds technical SEO.
  • 00:21:38so It is more like an API database than the tool.
  • 00:21:45Now with all the APIs MCPs get, I think pretty famous now.
  • 00:21:52So how to answer this question?
  • 00:21:56It's like we use still Sistrix for the old school SEO KPIs rankings visibility index right and rankings and all that stuff cpc something like that.
  • 00:22:09but you can't get it from DataForSEO as well And i liked.
  • 00:22:13the key is prompt research module from Sistrix.
  • 00:22:18So yeah, but for all the AI search is like ninety-five percent Peec AI.
  • 00:22:26And I think it's a nice view of point because and i think all the tool SEO and Search tool industry reflects very good what is going on in our industry companies, agencies and software providers slept on the trend.
  • 00:22:53And didn't saw it coming this hard I think because like you said Peec AI was two years that didn't even exist.
  • 00:23:04now they're like eighty people.
  • 00:23:06i know the founders pretty good.
  • 00:23:09Yeah, they come from already for nothing and I think a lot of old companies or software providers and agencies struggling with this fast growing AI search market.
  • 00:23:25And maybe are too stuck in the old world?
  • 00:23:35Like I said, we use mostly the old school SEO tools for the old-school data and new tools.
  • 00:23:43But i think they are trying.
  • 00:23:46it's always worth looking into what they're doing in terms of AI search data.
  • 00:23:55Let us talk about DataForSEO A little longer for a second here because I feel like you and me, so we also use DataForSEO.
  • 00:24:04We see the value already.
  • 00:24:06So yeah For us it already clicked.
  • 00:24:09What would you say to people that maybe are a little bit skeptic still where they say?
  • 00:24:13Yeah DataForSEO.
  • 00:24:16i mean if i look at their website It doesn't look as fancy Maybe as Ahrefs And Like.
  • 00:24:21They don't have a big conference like Ahrefs Evolve etc...so it feels like maybe smaller-ish provider, why is it still worth looking into?
  • 00:24:31And even thinking about maybe I can cancel my SEMrush or my Ahrefs subscription and replace that with DataForSEO.
  • 00:24:41What makes Data For SEO so interesting?
  • 00:24:45if people think about agentifying their SEO and GEO workflows?
  • 00:24:57I think it's more in the human than data, this decision.
  • 00:25:01Because if you work agentic like with Claude or other AI agents It is a way different approach to work.
  • 00:25:15then clicking through your eyes.
  • 00:25:18So what we say?
  • 00:25:20people who are still concerned and not sure about that would tell, try the DataForSEO MCP connected with your Claude Code.
  • 00:25:30Or you're AI of choice and just tried to ask him questions.
  • 00:25:36Try to automate your daily workflows like asking them for a keyword research or as come for content topic hub topics are Yeah, something you always do manually.
  • 00:25:50And so we see the magic because all of data I think it's a magical data and connecting all this data put into one task is the magic.
  • 00:26:05So try the data MCP.
  • 00:26:08It's very cheap like fifty dollars per month Something like that.
  • 00:26:14Try it then It's amazing and you will see by your own behavior that you're not using the UIs anymore, also often any more.
  • 00:26:25And how much of a role do think plays exactly what you mentioned?
  • 00:26:32The pricing where for example Semrush so... I can transparently say we also canceled our Semrush subscription.
  • 00:26:41Locks you into a subscription with the very high base fee, and you get a lot of products that like features but it feels different product apart are basically stuck together and bundled together And pay for a lot stuff That don't really use.
  • 00:27:00Then there's other providers AI native or these more like agentic providers, where you just pay for what you need.
  • 00:27:15How much of a role do think that also plays in how you decide which software to keep and which software cancer?
  • 00:27:23Pricing is always the thing experienced SEO or consultant, I think you already know what you want and what you need.
  • 00:27:40So for example when i log into Semrush have the same feeling like oh my god it's so much!
  • 00:27:47Like open photoshop but just...I wanna change the color.
  • 00:27:51everything that can do The rest is not necessary to me.
  • 00:28:03the operative work like SEO AI search, it's like ninety percent MCP API and maybe like ten percent UI.
  • 00:28:11And yeah I would just track my own behavior.
  • 00:28:15so when i'm not using it anymore is that necessary for me?
  • 00:28:18So cancel it!
  • 00:28:19That what you use in what you need.
  • 00:28:23stick with it.
  • 00:28:25cool let's talk about something where you actually build But then decided to discontinue it.
  • 00:28:33actually buy something because I saw.
  • 00:28:36I think was opposed on LinkedIn where you said that you build your own AI visibility tracker with Claude Code in ten minutes for ten dollars or ten euros don't know.
  • 00:28:46but Then in the end, he still concluded that it makes sense paying for Peec AI.
  • 00:28:51Because yeah what is the reason?
  • 00:28:56Yeah It wasn't.
  • 00:28:57Actually, it was not a tracker.
  • 00:28:59It wasn't reporting but the hook was very good so I used that.
  • 00:29:04But yeah The biggest reason for us to stick with Peec AI is of course data because paying for Peec AI cheaper than paying for API costs from all models So price and features All the things they are building next to data is a very, very good product.
  • 00:29:33So for example like the action plans that were introduced or all of their features bring out there's so interesting data and UI makes a lot sense!
  • 00:29:47I think it's pricing and the features i use especially in such new environment like AI search easier to provide good features because everything is new and in the twenty years SEO world we already saw every feature, so it's not that new anymore or nothing new for us SEOs.
  • 00:30:12So I think its the pricing of the features.
  • 00:30:14you see what they're building with passion.
  • 00:30:24get away from this, because I think that's a very important part in our world.
  • 00:30:30to stay on what is happening.
  • 00:30:38Exactly!
  • 00:30:39And Peec AI represented it very good and their feature.
  • 00:30:42so yeah... That are the
  • 00:30:45reasons.
  • 00:30:46Yeah, the Peec AI team feels really cutting edge.
  • 00:30:50Cool!
  • 00:30:51Another thing I was curious about...I already said it in the introduction you completely specialized and working with banks and insurers like the finance industry?
  • 00:31:01And i wondered what do they say when they hear that an increasing part of work for them is done by agents And I especially care about the compliance or legal teams here.
  • 00:31:20What do they say, what did they say?
  • 00:31:23Take us through these conversations a little bit.
  • 00:31:27Yes so honestly it's less drama than you would expect!
  • 00:31:31So our contract said openly we use AI tools under editorial control Yeah, and what legal teams want to know is who's accountable.
  • 00:31:44So which data goes where?
  • 00:31:46And whose signs off... Our answer always the same.
  • 00:31:52A name-it person is responsible for every piece.
  • 00:31:55Every number is reported check it by hand against a source something like that.
  • 00:32:01so there are some little points we have to look deeper into because in this industry yeah but its not dead big difference like compared to our e-commerce industry or fashion, something like that.
  • 00:32:16And have you also looked into the whole model?
  • 00:32:23provide data retention?
  • 00:32:25Where does the data go?
  • 00:32:26so using Claude on the Anthropic API directly US based using Claude through something like Google I think now it's called AI Studio.
  • 00:32:38It used to be called Vertex, which then allows EU routing versus even using EU data centers with for example open-weight models like GLM
  • 00:32:50etc.,
  • 00:32:52how do you think about that?
  • 00:32:53For a performance having the most capable model but also price and compliance so to say or is something that you feel like it's important at all.
  • 00:33:07Yeah, I think we work with our partners in this part as well because he is doing a lot of stuff for the industry and so he helps us alot.
  • 00:33:24but i can't tell you much about details.
  • 00:33:34But what I know is like that everything we do, everything we work for and every data it goes in is transparent.
  • 00:33:45We communicate with the clients so they don't want to do anything or no data goes out too.
  • 00:33:55It's just everything with approval from the client.
  • 00:34:00I want to wrap up the part about KKP OS a little bit with looking at the development process again, not like with the nitty gritty tech details.
  • 00:34:11But what would you say?
  • 00:34:15What worked better than expected?
  • 00:34:17and were there also things that you built then had to rip out again saying oh man we really overdid it here this is getting out of hand.
  • 00:34:28let's redo it and delete the last like five thousand lines of code.
  • 00:34:35Yeah, very
  • 00:34:37good!
  • 00:34:37The do's and don'ts and glory moments maybe also the fuck ups.
  • 00:34:43So things better than I thought is... I checked it.
  • 00:34:52like this week, that's a fifty-nine percent of all deployments and changes in our OS is done by our team.
  • 00:35:02So not my me!
  • 00:35:03And so thats things goes better than i thought because back then i felt like ah its just going to be me.
  • 00:35:11now we have to tell please use it but A great way to work, that the people and team is loving this OS.
  • 00:35:24And continue developing it by its own.
  • 00:35:28so like I said a ninety-five percent are built not by me.
  • 00:35:32So thats... yeah.. a good sign.
  • 00:35:38you build something great.
  • 00:35:42The thing we have to screw with
  • 00:35:44Was our rapid out again.
  • 00:35:47Yeah, was our I would say like the agency we call it Pulse.
  • 00:35:55that was our agency operating system to get all the finances and data together And the financial part was so complex.
  • 00:36:06So many data, so many points.
  • 00:36:09I wasn't even able to use it anymore and didn't know what... So Excel is better than this tool.
  • 00:36:17That's why we last week said after three months of developing Okay let's rip that apart.
  • 00:36:23It doesn't fulfill the needs any more.
  • 00:36:25its way more complex.
  • 00:36:27Take us more time then helps.
  • 00:36:31Okay, I want to play Devil's Advocate a little bit now because... ...I read you saying that the KKP OS, Claude.
  • 00:36:43The whole like agentification basically makes your team twice as fast.
  • 00:36:48Now if Claude Let's simplify here a little it If Claude makes good SEO and i know You only have good SEOs on Your Team Makes A Good SEO Twice As Fast.
  • 00:36:59The client's in-house team could potentially also get the same boost from buying Claude or ChatGPT, whatever.
  • 00:37:10Why do they still need an agency?
  • 00:37:13Or why is it still a better decision... ...or when is not even the best decision to hire an agent as KKP?
  • 00:37:25Very important questions and I think you can ask this question five years ago.
  • 00:37:32Because there always been companies, they do everything or a lot of things in-house and what I think is great because... ...I always encourage companies to do that.
  • 00:37:45build your own SEO team And i think we can bring the best results when We work with clients who have their old SEO Team because you have responsibility Bring good results.
  • 00:37:58So Yeah, if you can do it by your own If we can do in-house and that works And bring this the KPIs up It's great but with the KPI stuck for three years Call us and maybe help You.
  • 00:38:16That is a very on point response.
  • 00:38:23Let me still challenge A little bit.
  • 00:38:25For example, it's someone so let's look for options where maybe people can still get value from this as an in-house team.
  • 00:38:36So let's say okay that for whatever reason there may be skeptical of hiring agency what would you say if?
  • 00:38:44For example is someone ahead of marketing marketing lead or something is listening to this and he says Okay I Get It would love to hire someone like KKP, but right now I don't have the budget approval or something.
  • 00:38:59Like that what are maybe your top three things?
  • 00:39:03That you will tell them how they can enable their team and move their team a little bit more towards where for example your team already is having this AI transformation.
  • 00:39:16owning this AI Transformation What Are Your Top Three Top Three Things Maybe
  • 00:39:23Yeah, so when we for example get in touch with a new client We never say, hey make it like this and don't make that because they're the experts.
  • 00:39:37They know what is going on in his company to know KPIs.
  • 00:39:41there are their bosses who knows everything.
  • 00:39:46but I think we can always bring a different perspective with our expertise.
  • 00:39:56I think we saw everything in the SEO world, in this search word.
  • 00:40:04In this industry... The most important part is that encouragement has to come from the boss.
  • 00:40:15so for me it's not just a tool but transformation and working one year ago.
  • 00:40:29And this transformation needs room and time, resources to do it.
  • 00:40:35so its not like a project you have to throw in okay let's work with the eye for couple of weeks.
  • 00:40:42So It Needs The Time, It Neads The Right Place To Do This.
  • 00:40:47And Its A Transition That Needs Time.
  • 00:40:53And it was like, for us the same.
  • 00:40:55Like I said one year ago we worked differently completely differently and a lot of people were faster another person was slower.
  • 00:41:02they needed help.
  • 00:41:04there are concerns.
  • 00:41:05now.
  • 00:41:05all that stuff is going on.
  • 00:41:07so to answer your question as i would say The boss have started transformation work more with AI Yeah, and have to encourage a team.
  • 00:41:25show them that it's the way to work because if you're younger team.
  • 00:41:30You have to say hey guys your twenty or thirty years of work in front of you And now change how we work.
  • 00:41:41It is not Excel anymore nor Powerpoint anymore.
  • 00:41:45I don't know if I even still have Excel on my computer or PowerPoint, because it's Claude Code building my tables and Claude Code building my slides.
  • 00:41:57So no necessarily needed... And that is the biggest point.
  • 00:42:03so i would say the transformation needs room time and the bosses need to encourage their team.
  • 00:42:14invest in this transformation.
  • 00:42:17Where do you still have things that you DO NOT do, and Claude?
  • 00:42:37I
  • 00:42:41think everything in terms of decision-making, that's nothing i do include.
  • 00:42:50so it helps but its never on point.
  • 00:42:54It is still manually using by our consultants.
  • 00:43:03Yeah...I think thats the... Its hard to answer this question because have to think about.
  • 00:43:15There's almost nothing I think nowadays, so maybe for example all you see on my LinkedIn posts and the thing that i do in LinkedIn is ninety percent by Claude Code it gives me... So we have a database of AI search news that gets provided by our teams or every one providing this feed with news As a start of the week, I get a content plan for my LinkedIn posts.
  • 00:43:48They say okay analyze this and then create all the slides or text.
  • 00:43:55Then you have to check it out and the comments as well.
  • 00:44:04So I'll barely use the link in UI anymore, just get my reports from my agents on the phone.
  • 00:44:13so yeah it's hard to tell a question maybe calls and making decisions but everything else is Claude Code.
  • 00:44:20Crazy!
  • 00:44:21So cool for probably um...the final part I want to focus on something that i feel like is especially important for younger people.
  • 00:44:35Because there's this discussion about agents basically being able to replace juniors or replacing the work that juniors did so with juniors.
  • 00:44:47obviously we mean interns, working students and junior SEO managers etc.
  • 00:44:54My question is if agents do their job?
  • 00:44:58How can anyone become a senior if they are not yet?
  • 00:45:03Yeah, that's very good question and I think That's the question get asked in every digital industry Because i think when we It it can gets bigger problem.
  • 00:45:17In few years When the old heads retire And do you need to new people young people And I think we have to look at it in a different perspective.
  • 00:45:29Like you said, and that's the same in our company there is
  • 00:45:31no...
  • 00:45:32like i would say ninety-five percent of anything necessary do by a junior because the AI faster quicker cheaper more precisely than a junior.
  • 00:45:47but We have to Look at it like educational building Way like in Germany with the Ausbildung.
  • 00:45:56so we go three years and you get a little bit for the
  • 00:46:00English listeners apprenticeship.
  • 00:46:02And they get just a little amount of money very low, that's why it is because its more putting work into this employee than getting out.
  • 00:46:14So I think that's the way we have to do it.
  • 00:46:17so and of course there is a risk because like imagine you hire in junior or for internship.
  • 00:46:25And then after three years, okay now he's ready to work with our clients and then he is leaving your company.
  • 00:46:31but yeah i Yeah, that's a part you have to include in your risk decision-making and in your finances.
  • 00:46:41Like okay let's hire three juniors maybe one stays after two years then we'll have senior.
  • 00:46:49I think it is the perspective of looking at this Do this thing manually or do with AI but learn and get experience needed before we can throw on our client.
  • 00:47:04And what would you say?
  • 00:47:06What do people actually have to learn, especially the younger or more junior people if things keep changing so fast and will change.
  • 00:47:16So let's zoom back in time maybe ten years.
  • 00:47:24of course also things changed but I would say specially in the SEO industry it was slower.
  • 00:47:31Maybe you had a new Google Core update, maybe you have a new tactic or strategy.
  • 00:47:36that came about and then somebody said hey look they did this cool thing.
  • 00:47:40interesting!
  • 00:47:41This is how we can get a lot of backlinks.
  • 00:47:43let's also try it... And then maybe you tried in the couple weeks.. A couple months and then you assessed whether work etc.
  • 00:47:51Now I feel like this happening real time Like i see people David Konitzny from your team previously, now Peec AI, publishing research about what changes in the new ChatGPT models etc.
  • 00:48:07basically every week like there's something happening probably everyday or at least every second day.
  • 00:48:13so What is it actually that junior people have to learn?
  • 00:48:16because I suppose?
  • 00:48:17It's not Like a best practice or a tactic.
  • 00:48:22Or what ChatGPT retrieves as the top sources under GPT-Five point six soul or what's your take on that?
  • 00:48:33Yeah, it was a good question.
  • 00:48:35I think even experienced people struggle on desk.
  • 00:48:38so um What i will say young people is like try to stay Um Try to keep up with all the changes especially in the AI search and uh In the AI workflow agentic world two big parts that's placed together and I think the biggest part in work as a search marketing, search marketer.
  • 00:49:06And when you talk to junior who have all of theoretical knowledge then compared with experience like us You see immediately difference in the first two sentences.
  • 00:49:22So I think because they know what you're doing, to do a lot is this... They saw everything and just like experienced.
  • 00:49:31so i think it's get the knowledge try to learn and get experience And then you are confident with what your'e doing and the confidence which tasks or recommendations will give to clients or how to raise your visibility.
  • 00:49:50So things like try to keep up, maybe it helps.
  • 00:49:54We have a newsletter once a week for AI Search.
  • 00:49:58may that help.
  • 00:49:59so you or follow me of Follow You.
  • 00:50:02That Helps To Keep Up With All The AI Search News and Keep Up with the AI agentic world, how it's working and literally worth changing.
  • 00:50:13And yeah try to get experience!
  • 00:50:15If there is no company that can help you just maybe build your own website.
  • 00:50:19That´s what I learned back then with SEO.
  • 00:50:21we had this affiliate websites, twenty or thirty of ads... There was the best way to learn and building our own website is very cheap.
  • 00:50:33now You can build content, you can build backlinks To.
  • 00:50:37prompt research is, I mean almost everything for free online and i think that helps a lot.
  • 00:50:43So it's knowledge and experience like was back then was the same.
  • 00:50:50We will put the link to the newsletter in the description and I also have agents that help with post production.
  • 00:50:59You can find it under kkp.agency/insights, scroll a little bit.
  • 00:51:05then there's this AI search update weekly.
  • 00:51:08It's in German, but even if you're an English speaker maybe you can ask Claude to translate it for you or maybe Artur will at some point offer a English version too.
  • 00:51:16For some people with there is enough demand.
  • 00:51:20But so If I got your correctly You basically made the speech for side projects again right?
  • 00:51:26Junior people having to do their own projects learning it hands-on.
  • 00:51:31i had Julian Redlich on the podcast.
  • 00:51:34he was Involved in building up.
  • 00:51:36The seo team at Booking.com.
  • 00:51:39back in days Went, up until head of seo and Booking.com, now working at Permira.
  • 00:51:45this I think private equity ish company?
  • 00:51:48I don't know enough about finance but they basically own Kleinanzeigen, mobile.de, FlixBus, etc.
  • 00:51:56And he basically said the same.
  • 00:51:58He feels like ten or fifteen years ago everybody was doing side projects.
  • 00:52:04Now it has become less But maybe its'the best time to start again for junior people Setting up your own domains trying get excited Trying to get into AI responses.
  • 00:52:17Would you agree?
  • 00:52:18Is that something That is The
  • 00:52:20only
  • 00:52:20track even To get this experience You talked about?
  • 00:52:24Yeah, I don't think it's the only way but give you this slight edge compared to other people.
  • 00:52:31They just do their regular stuff and go into company because they have no limits.
  • 00:52:38I was working back then in a company as an SEO instead of going out on weekends with.
  • 00:52:47now my co-founder did SEO and affiliate websites.
  • 00:52:55Saturday, Sunday when we closed our laptop on Monday went back to the corporate job And that's how we learned it.
  • 00:53:05I think gave us this light edge in knowledge & experience.
  • 00:53:12There was a big part for me.
  • 00:53:17It gives you permission another permission or the Shows you that maybe the boss are your Your co-workers there.
  • 00:53:29You're passionate about this because it gives
  • 00:53:32you also credibility.
  • 00:53:33exactly Yeah, exactly, credibility That's yeah, you were passionate about and even Put your free time into it so I don't think it's necessary but i'm pretty sure It gives you to the edge Next to your colleagues or people who don't do it.
  • 00:53:54Maybe that leads us to a philosophical question, but I still want to ask you... Do you feel like one can become really good in anything?
  • 00:54:06Be at AI Search, be at SEO if your not truly passionate about it.
  • 00:54:11so is being passionate the necessary premise Deliver top-notch work.
  • 00:54:22I think if you want to be top notch and the last I would say one or two percent, I will say yes Yeah hundred percent.
  • 00:54:32we like you have to.
  • 00:54:33when i look at our people You have to be on the couch On the evening and scroll through linkedin because your curious.
  • 00:54:41what's new?
  • 00:54:42new things are.
  • 00:54:45Listen to podcasts, no?
  • 00:54:46Like.
  • 00:54:46I listen a lot too.
  • 00:54:47To your podcast because you have very nice guests and the topics are very interesting.
  • 00:54:52so listening on when i drive my car or when i enter gym Or...I don't know When do my dish is something like that.
  • 00:55:00So it's not- I don't think It's.
  • 00:55:02maybe You can be good..you Can.
  • 00:55:04we have a good life.
  • 00:55:05but if you want really Be top notch And another one percent you Have to be passionate about it and everything Else comes by its own, because you want to do it and interesting.
  • 00:55:19You just cover this
  • 00:55:22very good speech!
  • 00:55:25Artur, we are already approaching the hour mark and since... ...we now know that a lot of people listen to us in The Gym.
  • 00:55:33We want give them proper one-hour workout companion.
  • 00:55:38I always wrap up.
  • 00:55:41if you're a frequent listener You know it!
  • 00:55:44There's always final question at end.
  • 00:55:45That comes as great surprise.
  • 00:55:49But i wanna know what didn't we talk about?
  • 00:55:52I think we should talk about
  • 00:55:57the... I think like more, or maybe with your next guests.
  • 00:56:05About pressure in our industry?
  • 00:56:09I think that's a pressure of all changes especially an old SEO and new AI search And the change from working manually to now with AI agents.
  • 00:56:24For a lot of people there's a lot pressure.
  • 00:56:26Yeah,
  • 00:56:27to keep up with your thoughts we still have some time.
  • 00:56:30I think it is very interesting point
  • 00:56:33Because for us its like oh nice that new.
  • 00:56:36i want try this agent and build this workflow.
  • 00:56:41A poster today linkedin post wrote about Start some Claude sessions and bring it to the gym or my phone on a remote control and prompting while I'm working out because i love It but there are other people they're not.
  • 00:56:58Not built this way, so deep into it which is totally okay.
  • 00:57:05I mean, it's totally okay.
  • 00:57:06don't have to be this nerdy like we are and i think there's a lot of pressure in.
  • 00:57:12oh my god that's too fast!
  • 00:57:14I cannot keep up with it?
  • 00:57:16how am able get into next to my eight hour work so busy?
  • 00:57:22So maybe we talked about another previous question.
  • 00:57:30It always has come from the boss.
  • 00:57:32He has to give you this room and the space, resources... ...to try to keep up with it.
  • 00:57:41I think that builds a lot of pressure because people may be in their thoughts.
  • 00:57:51So think like, oh my god maybe I'm not good enough anymore.
  • 00:57:54And i think that needs a space especially in the HR when bosses talk one-to-one with their employees and it should be something leaders have in his mind That everybody is running this fast as we do And that's totally okay, but you have to be... You had this in mind.
  • 00:58:25That
  • 00:58:26builds a lot of pressure.
  • 00:58:30Maybe one additional thought don't know what do think about it?
  • 00:58:33I'm curious to hear!
  • 00:58:45they rightfully have.
  • 00:58:46They want us to deliver results and so, they say to us you have to be on the cutting edge!
  • 00:58:52You have to tell me what's going on?
  • 00:58:53And what the implication is for
  • 00:58:55us?!
  • 00:58:55I feel like that.
  • 00:58:57it's actually our duty as business owners... ...to build systems that allow our teams to stay on top of it without having to be this nerdy guys.
  • 00:59:09go through LinkedIn, attend events, attend conferences.
  • 00:59:14Listen to podcasts on a Sunday morning when we pick up some bread or etc.
  • 00:59:24for the breakfast.
  • 00:59:24so I think this was at least my feeling that We also have a duty and also is the capability now to encode all that knowledge The great research, for example also that comes out or we do on our own to encode into systems.
  • 00:59:45so Monday morning the team as you described it so nicely opens up their laptop and is not feeling overwhelmed by oh my god!
  • 00:59:55So much happening.
  • 00:59:56What the hell?
  • 00:59:57But actually feels calm because they can trust this system that helps them stay on top.
  • 01:00:03what do think about that Also from a business owner perspective here?
  • 01:00:08So it's a great answer to this question.
  • 01:00:10That is why always I say our employees, I don't want to be always say like, Be happy you work in this company.
  • 01:00:21Like with us because we provide everything that you said exactly.
  • 01:00:25We are on the cutting edge where providing everything building me pushing and it's almost rolling out red carpet agent carpet but you have to walk by your own.
  • 01:00:39That's a perfect answer you gave.
  • 01:00:43But there are other companies, they're not even a carpet.
  • 01:00:46it is not the floor.
  • 01:00:47You have to fly or jump and theres way bigger gaps between what we want to achieve And whats really capable because theyre not like this pushing AI nerds Like We Are.
  • 01:01:03Yeah so cool Artur, it has been a pleasure talking to you.
  • 01:01:09We will put the link to your LinkedIn profile in description.
  • 01:01:12A very big recommendation to follow You.
  • 01:01:15I think most content is German speaking But i think linkedin we'll do a good job and auto translating.
  • 01:01:21if someone here Is native English then?
  • 01:01:31Thank you
  • 01:01:35so much.
  • 01:01:36See you, bye-bye!

This episode is also available as Markdown and PDF.