LinkedIn thought leader ad examples

Positief nieuws! Eerder deze maand mopperde ik nog dat onze pensioenfondsen vooral in Amerikaanse bedrijven beleggen. Welnu, ABP gaat €1 miljard investeren in Europees durfkapitaal. Techleap laat in hun rapport ‘State of Dutch Tech’ al jaren zien waarom dat nodig is. Europese pensioenfondsen beheren zo'n $9 biljoen, maar stoppen daarvan jaarlijks 0,01% (!!) in Europees durfkapitaal. Dus springen Amerikaanse investeerders in het gat: in Nederlandse groeirondes van €50–100 miljoen verdrievoudigde hun aandeel vorig jaar naar 40%. En ABP begint op de goede plek: de eerste €250 miljoen gaat naar het Scaleup Europe Fund, dat juist in grote rondes en strategische deep tech investeert. Met de rekenmachine erbij: €1 miljard is 0,18% van ABP's vermogen. Klein voor ABP. Groot voor Europese scale-ups.

Compliance is boring, slow, and sometimes even flat out painful. Or at least, that's what it used to be. We helped seQura turn a classic bottleneck into a revenue driver by 16x'ing their case handling efficiency. We also increased first-time-right by 38%, only requiring follow ups in 12% of their cases. This is the magic of our policy engine, combined with smart AI automations. Read the full story here: https://lnkd.in/ekhytMk6 Lastly, if you love digging into gnarly business processes and build delightful experiences, Duna might just be the perfect place for you: https://duna.com/careers

Finanzbildung braucht eine Datenbasis Wie gewinnen wir mehr Menschen für den Kapitalmarkt? Die jüngste BSW-Trendumfrage liefert eine aufschlussreiche Antwort. 27,6 Prozent der Teilnehmenden setzen auf bessere Finanzbildung. 41,7 Prozent meinen, dass Bildung, finanzielle Rahmenbedingungen und weniger Bürokratie zusammenkommen müssen. Ich würde noch eine Voraussetzung ergänzen. Informationen müssen im Moment der Entscheidung verständlich und verlässlich verfügbar sein. Wer ein Wertpapier vergleicht, braucht mehr als einen aktuellen Kurs. Entscheidend sind auch Produktmerkmale, Kosten, Chancen und Risiken und die Möglichkeit, Angaben über verschiedene Angebote hinweg einzuordnen. Hier liegt unsere Aufgabe als Infrastrukturunternehmen. ARIVA bereitet Finanzdaten auf und stellt sie Banken, Brokern und Finanzportalen bereit, damit diese daraus hilfreiche Anwendungen entwickeln können. Finanzielle Selbstbestimmung entsteht nicht durch eine Datenlieferung allein. Aber ohne zugängliche, belastbare Informationen wird sie unnötig schwer. Deshalb interessiert mich an unserer Arbeit nicht nur, wie viele Daten wir verarbeiten. Entscheidend ist, was Menschen mit ihnen besser verstehen können. Die BSW-Umfrage ist eine Online-Abstimmung unter 1.298 Teilnehmenden, keine repräsentative Befragung aller Menschen in Deutschland. Ihre Leitfrage ist dennoch richtig: Wie machen wir den Kapitalmarkt zugänglicher?Wie kann aus einem ersten Interesse an einem Investmentthema eine informierte Auseinandersetzung mit konkreten Wertpapieren werden? Ein Ansatz sind datenbasierte Investment-Playlists. Welche Informationen dafür nötig sind und wie sie in eine Finanzanwendung gelangen, besprechen Katharina Herzog von money:care und Florian M. Roebbeling von ARIVA in unserem Webinar „Von Inspiration zur Investition“. Alle Infos zu diesem und allen weiteren Webinaren sind auf ariva.ag zu finden. #ARIVA #Finanzdaten #Data #Banken #Börsen #Broker #Fintech #Finanzportale #API #Schnittstelle #Finanzbildung #Altersvorsorge #BSW #money:care #Kiel

Den finansielle konsolidering behøver ikke være fyldt med fejl. På vores kommende webinar viser Pernille og jeg, hvordan I kan få bedre overblik og færre fejl og samtidig automatisere det manuelle arbejde, der unægteligt er til stede ved manuel konsolidering i Excel. Tilmelding via linket i kommentarfeltet - vi glæder os! 🎉

When will AI agents execute trades in fixed income? ‘The desire is there, but there are significant barriers to that.’ That’s David Rickard, Global Head of Fixed Income at LSEG, on the latest Capital Markets FinTech Forum podcast. He thinks adoption will come faster than we expect, but the barriers are real: models still struggle with real-time data, handing execution to a model faces heavy scrutiny and trading systems on many desks are ‘deeply entrenched’. For now, his view is grounded. Most firms are experimenting with the tools, proving value and embedding them into real workflows. Agentic, multi-step automation is still the frontier. I also liked his explanation of why chat and voice persist in fixed income. How electronic a market becomes depends on two things: how standardised the instrument is and how liquid it is. Where both are low, relationships matter. The question becomes ‘who’s the most likely to show you a price in times of stress’. That makes the unstructured data in chat a huge opportunity, but only if desks can trust where the data feeding their models comes from. Three takeaways: 1. AI on the desk is moving from experiments into real workflows. Execution will come later. 2. Chat persists where liquidity and standardisation are low, and that won’t change quickly. 3. Data provenance is becoming the gating factor for AI, and compliance teams are being asked to move faster than ever. David is hosting a buy side only workshop on 'designing fixed income workflows for the AI era' at FILS on Wed 14th at 1605 CEST. If you’re going, see you there. Full episode: https://lnkd.in/efV8eVVb #FILSEU2026

Boligjagt i Aarhus har lært mig, at man godt kan få hjertesorg over mursten 🥲 Man når at forelske sig i en bolig, indrette den oppe i hovedet og regne på, hvordan hverdagen ville se ud. Og så bliver den revet væk, før man reelt når at få en chance. Det er én slags boligkaos. Den anden opstår ofte længe efter, man har købt. Når boliglånet bare har kørt videre i årevis, mens renter, vilkår og muligheder har ændret sig undervejs. Ikke fordi man er ligeglad, men fordi der er rigeligt andet at holde styr på, når man allerede ejer bolig. Det er den del, Finobos låneberegner er lavet til. Den giver boligejere en første, vejledende vurdering af, om deres nuværende lånepakke er værd at se nærmere på, og om der kan være en mulig besparelse ved at ændre noget. For nogle er det nok til at få lidt mere ro i maven. For andre er det starten på at få set deres lån efter i sømmene. Se link til beregneren i første kommentar, hvis du også vil prøve den. #boligjagt #beregner #privatøkonomi

𝗬𝗼𝘂 𝗽𝗮𝘆 $𝟭𝟬𝟬 𝗽𝗲𝗿 𝗰𝗮𝗹𝗹, 𝗮𝗻𝗱 𝗼𝗻𝗹𝘆 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗰𝗮𝗹𝗹𝘀 𝘆𝗼𝘂 𝗱𝗲𝗰𝗶𝗱𝗲 𝘄𝗲𝗿𝗲 𝘂𝘀𝗲𝗳𝘂𝗹. We don't ask for recordings. Your word is enough. Here is what it costs, what you get, and where the limits are. 𝗪𝗵𝗮𝘁 𝘆𝗼𝘂 𝗽𝗮𝘆 ➜ $1,000 for three months of infrastructure and outreach: domains, mailboxes and tools ➜ $100 for each call that took place and that you find useful ➜ You decide which calls count 𝗪𝗵𝗮𝘁 𝘆𝗼𝘂 𝗰𝗮𝗻 𝗲𝘅𝗽𝗲𝗰𝘁 ➜ 10–50 qualified investor meetings in 90 days. ➜ All of the outreach on our side. We pick the funds, write the emails in your name, and put meetings in your calendar. You take the calls. 𝗪𝗵𝗮𝘁 𝘄𝗲 𝗱𝗼𝗻'𝘁 𝗽𝗿𝗼𝗺𝗶𝘀𝗲 ➜ A closed round or a check. You close those. I'd rather you hear that from me now than find out later. 𝗪𝗵𝗲𝗿𝗲 𝗰𝗹𝗶𝗲𝗻𝘁𝘀 𝗲𝗻𝗱 𝘂𝗽 On average, we generate 22 calls per startup. 𝗕𝗼𝗼𝗸 𝗮 𝟑𝟎-𝗺𝗶𝗻𝘂𝘁𝗲 𝗰𝗮𝗹𝗹 𝗮𝗻𝗱 𝘄𝗲'𝗹𝗹 𝘄𝗼𝗿𝗸 𝗼𝘂𝘁 𝗵𝗼𝘄 𝗺𝗮𝗻𝘆 𝗺𝗲𝗲𝘁𝗶𝗻𝗴𝘀 𝘆𝗼𝘂𝗿 𝗿𝗼𝘂𝗻𝗱 𝗻𝗲𝗲𝗱𝘀 𝗮𝗻𝗱 𝘄𝗵𝗮𝘁 𝘁𝗵𝗲𝘆'𝗱 𝗰𝗼𝘀𝘁. https://lnkd.in/eQhtSwMc

970 replies came back from 7,500 cold emails in one 2026 campaign. Founders ask me why our call numbers are higher than what they get on their own. Here are both funnels, side by side. A founder, on a call with us 500 emails ➜ 1 fund Our funnel, one 2026 campaign 500 funds ➜ 2,500 people ➜ 7,500 emails ➜ 970 replies ➜ 40–60 calls ➜ 1–4 term sheets The gap is 40 to 60 times, and it comes from two multipliers: ➜ 15 times more emails: 5 people from the investment team at each fund, and each gets one email and 2 follow-ups ➜ 3 to 4 times more calls per email: 0.5–0.8 calls per 100 emails against 0.2 Most of the gap comes from reaching more people at each fund and from follow-ups. If your own outreach is stalling, start with the list, the recipients and the follow-ups. Why each of our emails works better ➜ The list. A database lists everyone who has ever invested. Around 10,000 funds are investing right now, and we pick 500–1,000 of them for your company by their deals. ➜ The email. One line about the company, traction at the top, round size, links to the deck and Calendly. ➜ The infrastructure. Two weeks to set up separate domains and mailboxes before the first send. The founder above kept hitting sending limits, and his emails landed in spam. On its own, cold outreach rarely closes a round. It gets you the calls, and the money usually comes through warm intros. We take all of the outreach. Send us your current numbers, and on a 30-minute call we'll show where your funnel leaks and how we'd get you 10–50 qualified meetings in 90 days. https://lnkd.in/eUFp3XZC

“𝐈𝐧𝐬𝐚𝐧𝐢𝐭𝐲 𝐢𝐬 𝐝𝐨𝐢𝐧𝐠 𝐭𝐡𝐞 𝐬𝐚𝐦𝐞 𝐭𝐡𝐢𝐧𝐠 𝐨𝐯𝐞𝐫 𝐚𝐧𝐝 𝐨𝐯𝐞𝐫 𝐚𝐧𝐝 𝐞𝐱𝐩𝐞𝐜𝐭𝐢𝐧𝐠 𝐚 𝐝𝐢𝐟𝐟𝐞𝐫𝐞𝐧𝐭 𝐫𝐞𝐬𝐮𝐥𝐭!” These words from Albert Einstein ring just as true today as they did roughly 100 years ago. We are roughly half a year into our business year. Amazing how quickly time is ticking away. But nevertheless, it’s a good moment to look at the performance at EOS Group: The first few months of this year have demonstrated our capabilities. Weve once again invested significantly in NPL purchases, which clearly shows that our strategy is paying off and that our teams are working hard. The market environment is anything but easy. Pricing remains competitive, and conditions are volatile. Nevertheless, we are doing well in all regions. Additionally, a few hidden champions, such as Slovenia and North Macedonia, are gaining momentum. We are confident that we will exceed our targets. And what has Einstein to do with all this? Well, EOS has been and is a very successful organization. But we will not stay this successful if we repeat what we did in the past. AI will be one of our drivers of our future success. We are intensively working on processes to be transformed and we are looking at every opportunity in order to be seized. This, for me, is the one message to take with me regardless of good numbers.

Technology opens possibilities. But it is our people who turn them into real transformation. Our latest Santander Innovation Challenge is a great example of that. In this new episode of Voices of Our Transformation, we meet the teams behind the winning projects that use AI to reduce workloads, support better decisions, connect trusted data and accelerate product development. These four projects show what transformation looks like in practice. They are helping us rethink how we work, unlock greater efficiency and create value across the bank. This is exactly the kind of impact we want to generate through our One Transformation. At our Investor Day, we committed to generating more than €1 billion of business value annually from data and AI initiatives by 2028. That is what being an AI-first bank means: combining the best technology with the best talent. We have both and, most importantly, we have the ambition to keep raising the bar. Congratulations to the winners and to everyone who took part, contributing 3,400 ideas from across Santander.

TradFi is going 24/7. So what happens when crypto loses its favorite competitive advantage? Just got off stage at the Digital Asset Yield Summit in Singapore for a panel appropriately titled “Why Hyperliquid Will Eat the World.” I think it can. The ability to trade crypto, global equities, oil, gold, and eventually almost any asset class on one always-on platform is incredibly powerful. That convergence is exactly what we’re building for with CRXTrade.com and CoinRoutes: giving traders one infrastructure stack to access crypto, futures, equities and emerging perpetual markets. But 24/7 itself isn’t the moat. Traditional markets are moving there too. The harder problem is certainty. Institutions don’t care about your L1, L2 or wallet architecture. They care about contract standards, clearing, margin rules, settlement, and whether a thin market on Sunday can liquidate them against an equity that hasn’t traded since Friday. Crypto has moved incredibly fast on technology. Now the market structure needs to catch up. Hyperliquid may eat the world. But the next trillion dollars comes when on-chain markets become boring enough for institutional risk committees to trust them. Jennifer Shaughnessy Michelle Herman, MBA

new summation pod with my friend Roy Schwartz, co-founder of Axios and author of the new book Simplify. we get into meeting bankruptcy, why AI made everything longer, and why average is worthless now. (and pickleball) https://lnkd.in/ektjsFUy

Thoroughly enjoyed being on stage at SuperReturn CFO/COO, showcasing what Steward is all about, a well as talking through the impact of AI in investor operations and AML/KYC. There is an exciting shift we are leading in the investor ops world, with opportunities to increase operational strength and risk management, without hiring a massive team to do this. Further, having spoken to many GPs during the event, it was clear that AML is still a big pain point for managers - so glad that we are making every onboarding a little more magical. For those of you that missed the event or just wanted to learn a little more about Steward, feel free to reach out here: https://lnkd.in/eYuXhjya

Steward ships Adverse Media Screening - for Social Media! 🚀 This has been a wish list item for our clients for a while. Our AI Agents can trawl through social media posts and uncover risks before they appear in your cap table, LP register or vendor list. Screening is not one-dimensional - we already source search engine results and existing screening providers to give comprehensive AML coverage. We now have gone one step further - ensuring you are not missing something critical. Social media is the home for exchanging ideas and insights, but more and more has become a place for radical ideas, hate and crime. Everyone has a right to privacy, but that doesn't mean as a business customer you need to do business with firms or people that do not align to your core values. Our AI-assisted Screening solution can pick up on the things you want to know before you are committed through a contract. This is targeted to: - Asset Managers - Wealth Managers - Banks - Professional Services - Broker Dealers - Payment institutions - Other financial services 👉 Want to learn more, speak to us here: https://lnkd.in/eFzHBj9g

🚀 Et si la Bourse devenait votre prochain levier de croissance ? D.A.T.E. est une PME de la BITD qui développe des solutions de gestion thermique de haute performance pour des secteurs stratégiques tels que la Défense, le spatial, le quantique et les technologies critiques. 🔔 La société vient de s'introduire sur Euronext Growth Paris, où elle a levé 12,6 M€ ! 🎓 Il y a un an et demi, Philippe BOTTU participait à #IPOready, notre programme de préparation à une potentielle introduction en Bourse. Aujourd'hui, il revient sur le parcours de D.A.T.E, son IPO, et les prochaines étapes. RDV en commentaires pour en savoir plus sur IPOready ! Adam Bohill, Alexis JANIN, Cécile Rarivozanany, Delphine d'Amarzit, Mathieu Caron, Eric Tossah, Jerome HERVE, Alain Baetens, Ismail El khalloufi, Frédéric Boiffin, Aurelie Girod, Héloïse Delon, Xabier Laforestrie, Gaultier de Tarade

This week in Frankfurt, I am feeling a particular sense of pride. Today, we are launching Crédit Agricole Deutschland, a decisive milestone in our Group’s European journey. Our #ACT2028 strategic plan is clear: we are committed to becoming a leading bank and insurer across Europe, driving sustainable growth and deepening our impact on the continent. Germany naturally occupies a central place in this vision. As Europe’s largest economy, a major savings market and the continent’s industrial crossroads, Germany is essential to the future of Europe and our Group. We are not newcomers here. We have been investing in the country for more than 70 years and already serve close to one million customers through our businesses. Last spring, the successful launch of CA Savings, our digital savings platform, marked the first stage of our development in the German retail market. Today, with Crédit Agricole Deutschland, we are opening a new chapter by deploying our universal banking model. With Crédit Agricole Deutschland, we want to put Crédit Agricole’s financial strength, deep expertise, proven innovation and collective power at the service of German households and businesses. By combining the best of digital technology, artificial intelligence and human relationships, we aim to build a bank that is useful, sustainable and future-oriented. Our history has taught us that lasting success is built over time. This is how we have grown across Europe. And this is with the same long-term commitment that we are investing in Germany today. Because I am convinced that there can be no truly leading European bank without a strong presence in Germany. And because there can be no stronger Europe without institutions capable of supporting its economy, its businesses and its citizens over the long term.

Building an AI-native product is not the same thing as building AI people trust, and when working with users who are experts in their field, keeping a human at the center of the work is a top priority. Since I joined Hypha in May, I've been leading design for the Deal Analyzer and Covenant Compliance products. Harnessing the power of AI to make private lenders faster and more accurate in their work has remained priority #1 for me. It's tempting to measure this kind of work by how much the software can finish on its own. But the lenders I design for aren't really asking for more completed work. They already know how to underwrite a deal. What they own that the software doesn't is context: why a borrower got an exception last time, which covenant their team actually watches, what a number means in the shape of a particular deal. None of that lives in a document. So, two things take up most of my attention: The first is surfacing reasoning and providing citations. If the product hands someone an answer and asks them to believe it, it's asking for a kind of trust it hasn't earned. Being able to see where a number came from, and check it, is the difference. The second is keeping judgment inside the workflow rather than around it. The review and override moments aren't cleanup steps for when the model gets it wrong. They're where the expertise actually is, and the design should treat them that way. The more interesting opportunity isn't technology that completes more work for you. It's technology that becomes an extension of your expertise, one that preserves your context, your institutional knowledge, and your judgment instead of routing around it. On October 22 we're hosting our first Leverage webinar. Active bank and private lenders will talk through how they're navigating origination, underwriting, and portfolio management, and you'll see the products I've been designing at Hypha working across those same workflows. If you're designing for experts in any field, I think this is the most interesting version of the problem right now: not how much you can take off someone's plate, but how much of their thinking you can keep on it. Join the webinar on October 22 to preview this design thinking and see how Hypha is solving complex workflows for private lenders. https://lnkd.in/gmp34kSW PS a shoutout to my amazing product and engineering partners on these products Mark Conley Brightbill Jr, Pedro Colon Hernandez, Ph.D. and Michael Dessoulavy!

We've been heads down building Eisen Tax Center. Today, its generally available to all customers. End-to-end tax reporting for financial institutions and corporate AP teams, from the first W-9 to backup withholding. Collect the payee information. Check the TIN. Prepare the form. Get it to the IRS and the recipient. Handle what comes back. Those steps belong together. Tax Center connects them through Payee Manager, Form Production and Backup Withholding. One popular feature: FIRE-to-IRIS conversion. Keep the FIRE export your team already produces. Eisen handles the conversion, checks for filing errors and manages filing. A new filing destination shouldnt force your team to start over. This is another part of account lifecycle operations that were bringing into Eisen. We keep money with its owners. If you run tax and information reporting at a financial institution or a corporate AP team, see Tax Center here: https://lnkd.in/dEGCnKs9

Same AI model. Same agent. 51 real private equity questions. Changing only the data layer lifted the score from 86.6 to 95.1, cut tokens by 41%, and reduced tool errors 3.4× (13.5× on structured workflows). We ran this because in private capital, the edge isn't the model or the workflow. It's your data, and whether AI can understand it. So we connected the same model to two backends through MCP: Metal and a leading enterprise file platform. Here's what we found. The widest gap, up to 50 points, came on four kinds of questions: • Version: which EBITDA, draft or return table controls. A file index treats every version as an equal match. • Time: which period a number belongs to, and when a CIM was actually received rather than when the file was created. • Identity: what a business actually is, across codenames, aliases and duplicate copies. • State: where a deal or screening stands in the pipeline, which no document records. None of these answers sits in a single file. A context layer holds that structure; a file index makes the model rebuild it on every query. The file platform kept pace whenever the answer lived in one set of documents. The gap opened when it didn't. Models will keep getting better. The firms that get the most from them will be the ones whose data is ready. Quick read on the key findings: https://lnkd.in/e-SXzW7A How we ran the test: https://lnkd.in/eivsUuaR

I'm hosting two webinars on crypto compliance, and I'd love for you to join. On 𝟮𝟭 𝗢𝗰𝘁𝗼𝗯𝗲𝗿, we'll look at two sides of crypto compliance: getting the fundamentals right and navigating more complex challenges around sanctions and jurisdictions. 𝟭𝟰:𝟬𝟬–𝟭𝟱:𝟬𝟬 | 𝗚𝗲𝘁𝘁𝗶𝗻𝗴 𝘀𝘁𝗮𝗿𝘁𝗲𝗱 𝘄𝗶𝘁𝗵 𝗰𝗿𝘆𝗽𝘁𝗼 𝗰𝗼𝗺𝗽𝗹𝗶𝗮𝗻𝗰𝗲 We'll cover the essentials of digital assets, regulation and what this means for compliance processes in practice. 𝟭𝟱:𝟯𝟬–𝟭𝟲:𝟯𝟬 | 𝗔 𝗱𝗲𝗲𝗽 𝗱𝗶𝘃𝗲 𝗶𝗻𝘁𝗼 𝘀𝗮𝗻𝗰𝘁𝗶𝗼𝗻𝘀 𝗮𝗻𝗱 𝗷𝘂𝗿𝗶𝘀𝗱𝗶𝗰𝘁𝗶𝗼𝗻𝘀 Together with crypto and OSINT investigator Kieran Kaal, we'll take a closer look at sanctions and jurisdiction-related challenges in digital assets. My goal? To move beyond theory and focus on what actually matters when dealing with crypto in financial institutions. If you're working in compliance, banking or digital assets, I hope to see you there. Register here: https://lnkd.in/epC7yx-v

“We’re all here to make money.” That’s what I told the audience on my TOKEN2049 panel. Technology Innovation is great. But if you’re not helping your clients make money, what’s the point? L1s, L2s, wallets, DeFi, stablecoins, RWAs, tokenization, and the next big protocol. There’s no shortage of innovation in crypto. But at the end of the day, traders want access to more markets, better execution, and more opportunities to make money. That’s exactly why we built CoinRoutes, and why we’re taking it a step further with CRXTrade.com. Now our clients can trade crypto, stocks, futures, and other asset classes from one platform, powered by the same institutional-grade CoinRoutes algorithms used by professional traders around the world. More markets. Better execution. Less friction. But there’s something even better than making money: building a business with people you genuinely enjoy working with. Great spending time in Singapore with our CoinRoutes team from the USA, Asia, London, and Dubai. We’re building something special, and having this crew along for the ride makes it that much better. #TOKEN2049 #CoinRoutes #CRXTrade #Trading Michael Holstein Jennifer Shaughnessy Michelle Herman, MBA Mia Y. Sitora Mashrapova Raed Saade

Triaging breaks, checking file deliveries, swivel-chair work, double checking systems, writing emails: that way of working won't exist in Operations in five years ⬇️ This is already a reality now for some firms working with us. In the most advanced cases, Claude talks to Duco over MCP and the two systems do all the exception management work. See here for some ROI intel: https://lnkd.in/ekV_X9ZM There are major implications here. The best people in Operations become supervisors of agents, and they get 10x more work done. Their job is to keep the agents busy and to make mission critical decisions when they pause. One of the consequences here is that a gap starts to open between leaders and slower adopters. Leaders have flat cost with volume rising. Slower adopters need to keep hiring, take longer to get into new markets and products because the middle and back office slow them down, and ultimately have bad share price fundamentals in a few years. So the question isn't about tech at all... what is your team going to look like in 2 to 3 years?

Standing still means falling behind. I was traveling again last week. Although it can be exhausting at times, I found myself reflecting on the journey home that I always return with plenty of exciting insights and experiences. On one of the last journeys for example, I visited a country where the EOS Group might want to invest. I had interesting discussions with the central bank, business leaders, and various ministries, such as the Ministry of Finance. Approaching a new market, getting to know the culture, the economy, and the key players is one of the highlights of my job! Observing and analyzing how stable a market is, what values are upheld, and how the competitive landscape looks like – there are so many different layers and levels to focus on. What I find most fulfilling is connecting these different perspectives, determining whether they align with EOS's principles, and ultimately assessing whether we can build a sustainable long-term presence while staying true to who we are and what we stand for. Let’s see how this one turns out. I thought it looked very promising…

It all started with a frustrated bank CEO telling me, “I can’t tell compliance vendors apart anymore. They all make the same claims. I just want to know which AI tools will actually deliver.” Today, we answer that question with the FinCrime Agent Benchmark, the only benchmark that tests how AI, including Claude Fable 5.1 and GPT 6 Astra, perform on real financial crime screening and adjudication work. Most AI benchmarks score models performance on generic tasks. But AML/KYC does not run on one set of rules. Every institution has their own SOPs, risk appetite and screening thresholds. To make a real comparison, you need the same cases, the same procedures and the same scoring criteria. That had never been done for AML/KYC. FCA Bench asks the questions compliance teams actually need answered. Can the AI apply an institution’s policies, identify risk and reach the right disposition consistently? We evaluated nine LLMs on their own and with Castellum.AI's agentic harness. The results speak for themselves. ➡️ Unharnessed LLMs missed up to 72% of true positives, meaning risk is either never surfaced or wrongly cleared. ➡️ Without a harness, even the latest models failed at basic compliance work, where regulators expect minimum accuracy in the high nineties. GPT 6 Astra achieved 71.1% accuracy standalone, but 99.3% with Castellum.AI’s harness. Claude Fable 5.1 standalone had 64.3% accuracy, but 99.5% with Castellum.AI’s harness. ➡️ Standalone model accuracy was as low as 30.8% (Meta Muse Spark 1.2 — worse than a coin flip). ➡️ Standalone LLMs achieved only 45.2% consistency when tested on the same case three times. Imagine explaining to your auditors and regulators that your screening system is as good as a coin flip… Castellum.AI’s harness raised LLM consistency to 97.2%. A model going from 71% to 99% accuracy isn’t just a huge leap, it's the difference between unacceptable and acceptable to regulators. It also makes one thing very clear. You cannot prompt your way to reliable AML/KYC decisions. LLMs need guardrails, infrastructure and specialized data to be accurate, consistent and defensible. The full research is linked in the comments Read the full benchmark: https://lnkd.in/g5Rq6ey7