Future Processing

The media industry still cannot agree on how to measure what it buys.   Several measurement currencies run side by side. Linear advertising has unique ad identifiers, digital does not. So media buyers end up depending on platforms and channels for the figures they trade on.   This was one of the themes at Future of Media in Manchester last month, which Jarek Feith and Pawel Panowicz attended. The discussions also covered how AI is changing media planning, where creativity belongs in that process, and how reach is built when audiences no longer behave as one group.   This matters to us because the technology is the smaller problem here. Whoever defines what counts as a view also decides where the money goes, which makes this a question about rules before it is a question about tools.   Whatever standard the industry agrees on will still have to work inside systems that already run. Ad identity has to match across linear and digital advertising. Campaign reporting needs to stay consistent while the measurement rules change. That is engineering work, and it decides how fast any new standard becomes usable.

On the first day of Marine Claims International: Dublin 2026 in Malahide, Anna Łukaszczyk, our Strategic Initiatives Lead for Insurance Practice, and Karol Urbańczyk, our Insurance Solutions Architect, presented a session titled "The Inbox is the System". It asked whether the gap between when claims information first appears and when it is formally captured is something the market has simply accepted, or something worth designing for. Thank you to everyone who took part in the discussions.

"We bought everyone a licence and told them to get on with it." That is the answer Daria Polończyk, Head of Delivery at Future Processing, hears most often when she asks other companies how they approach AI transformation.   Tools speed up individual tasks, but as more people, handoffs and decisions get involved, coordination becomes the bottleneck. Without redesigning the process, the early speed-up does not hold.   Our own starting point was the same kind of gap. In summer 2025, 81% of respondents already used AI tools, while team-level AI practice stood at 34 out of 100.   Daria explains the decisions behind it in her article: https://lnkd.in/epVieuaP   Our case study on how we closed that gap, with methodology, measurements and results, is coming soon.

Company announcements are full of AI successes. Almost no one shares the story of how a transformation was run, with the methodology, measurements and results. We want to make that easier for other organisations, so our case study shares the numbers behind our own. We surveyed our employees twice: in summer 2025 and again in April 2026. Team-level practice went from 34 to 52.7 out of 100. The same data showed a clear link. People who rated their team's AI practice higher also gave AI more credit for their projects' success (r = 0.42). How often each person uses AI matters too, but team practice adds something on top of it: the link remains when we take individual use into account. For a company, that changes what is worth managing: licence counts and usage rates describe individuals, and team practice adds something on top. It matches the route our own transformation took: through governance, roles and the delivery model before it reached the commercial model. We publish the full case study soon, with the methodology and measurements behind these figures. Sign up for the event and you will get a reminder when it goes live.

How do you tell the difference between a process that looks good on paper and one that works, in a way you can measure? That is a question insurers will be working through at FutureTech Leaders: Operations, data, reality on 27th October in Munich, where the panel looks at what to fix first, what to standardise and where differentiation should stay a competitive advantage. Sebastian Risse joins the panel with exactly that experience behind him. As Head of Operational Excellence, Property & Casualty at Swiss Re, he built a maturity model spanning the full reinsurance value chain, connecting process and data maturity to measurable business outcomes rather than an audit score. He now runs Risse Advisory GmbH, helping insurers get their operating models ready for AI-enabled ways of working. Come and hear it for yourself. Registration is open on our website.

Any delivery partner can say its team knows how to put Claude into production. Verifying that claim takes more, which is why the first level of the Claude Partner Network is built around a certified delivery team.   Future Processing is now a Certified Partner of the Claude Partner Network, the programme through which Anthropic works with firms that put Claude into production for customers. The status follows the Claude certifications our people completed, which we announced in August.   Arkadiusz Dymek, Managing Director of our Cloud Vertical, explains why we treat technology-provider certifications as working tools for delivery teams, and what each of them has to demonstrate before it is granted.   His full view, together with the details of the announcement, is in our press room: https://lnkd.in/d_s_gz2X

What does 'client' mean within a company's own systems? Even without a clear definition to work from, a large language model will still produce an answer. Future Processing was a Gold Sponsor of SQLDay Lite in Gdańsk, run by Data Community Poland, with our colleague Grzegorz Brodny, who leads the Tricity branch, as its main organiser. Alongside him, three more of our people spoke on this very topic: Krzysztof Paś, Remigiusz Tunowski and Krzysztof Nykiel, who leads our Data Solutions team. Krzysztof P. showed what the same question looks like in a migration. The source system marks a client as active, the data platform marks the same client as inactive, and the business expert says that is not how it works. Unless someone decides which version is right, both move to the cloud and the legacy goes with them. Grzegorz Brodny and Remigiusz Tunowski presented the other side: in an ontology, business objects and the rules for changing them are defined once, and a language model works on exactly those definitions. Teams get the flexibility of the model without giving up the reliability of hard data. The work behind that sits earlier than most AI conversations go: in preparing the data and definitions a model depends on. It's the kind of work our Data Solutions team does with clients across industries, and one of the reasons AI projects succeed or stall long before anyone talks about the model itself.

Every board now has a number for AI use, and in most companies that number is high. Evidence that the company has changed because of it is harder to find.   We measured that distance in our own organisation before trying to close it. In summer 2025, 81% of our people said they used AI tools. In the same survey, the same people scored the company 34 out of 100 for using AI as a team.   The first number describes individual habits. The second describes how our teams used AI together at that point.   Access to the tools turned out to be the smallest part of closing that gap. What took the time was everything the tools do not touch: deciding how AI may be used on a given piece of work and who answers for the risk, reshaping a team once producing code is no longer the expensive step, and pricing a project that takes a fraction of the hours it used to.   In the coming weeks we publish the case study: our path from AI adoption to AI transformation. It comes with best practices from our methodology, measurements and results.

Companies invest in AI and use it in business functions, yet few can point to a measurable change in the P&L. McKinsey's State of AI report puts the share of companies drawing more than five per cent of EBIT from AI at around six per cent.   Adam Gaca, Dawid Przespolewski and Daniel Jaros spoke on the role of AI in modern M&A at the Andersen Corporate and M&A Summit in Warsaw on 24 September. A short exercise with the audience of dealmakers and transaction lawyers matched that picture.   For deal teams, this changes how "AI in the target" should be read. It can mean an operating model built around AI or licences for tools people use to chat with a large language model, and telling the two apart is a due diligence question that has a direct effect on price.   M&A has decades of practice in assessing value and risk. AI plays two roles in a transaction: it is a tool for the legal and deal teams running it, and part of what is being bought. Applying the same discipline to both, with shared standards for what counts as evidence of value, would give acquirers a clearer view of what they are paying for.

Claims transformation can begin with a single, well-chosen change. The more useful question is what is slowing the claims process down, and which change would make the biggest difference. Future Processing's research points to three stages of maturity: process optimisation, expert augmentation and ecosystem integration. Rather than attempting everything at once, the goal is to identify the capability with the highest impact and start there. That is also the approach behind futureClaims: maturity assessment, discovery workshops and solution exploration, designed to help claims organisations choose the right next step. Not sure where to begin? Start by assessing your claims maturity and defining the right path forward. https://lnkd.in/dnhjQCsr