Version 1

We drive the best outcomes for clients when conversations start with the raw reality. Simon Gerhardt IT Operations Director at Hays, described his first reaction to being told the business had to migrate its entire estate, with no disruption, against a deadline it hadn't chosen, in one word: "Terror." That's not a quote you'll find in many case studies! But in my experience, it's exactly that kind of honesty that creates the foundations for success. The story is captured in the images below, and you can read how it ends in the full case study here: https://lnkd.in/eMxK4QzN

Nice little weekend project to keep me out of trouble. Mostly built so I'd stop being surprised by my Claude Code usage limits, and to squeeze the most out of each window before it resets. A 7.3" colour e-ink panel on a Raspberry Pi, running the open-source InkyPi project with a set of screens I wrote for it. It cycles every ten minutes through the weather for the village, BBC headlines with a paragraph for each story, the house's solar and battery state, the health of the homelab, and a page showing how much the local LLM box has saved me against cloud API pricing. There's a Bitcoin page too, mostly to remind me the odds of my little miner finding a block are about one in two million a day. It also shows the time in rotation, which Kate Marlor points out a clock could have also done. Ah well, one side project finished. Roughly a hundred still on the desk in pieces.

How does a market-leading financial services organisation build the right AI foundations? That was one of the big questions facing Key Group, later life lending experts, as it looked to simplify its technology estate and consolidate on Microsoft. Working with Key Group, we've helped reduce incident volumes by 5x, accelerated provisioning, and freed up more time for their technology teams to focus on customer outcomes rather than infrastructure issues. As Jack Neary, Group Technology Director, puts it: "It's all about strong foundations for the AI era and Version 1 have knocked that out of the park." Read the full story.

How does a market-leading financial services organisation build the right AI foundations? That was one of the big questions facing Key Group, later life lending experts, as it looked to simplify its technology estate and consolidate on Microsoft. Working with Key Group, we've helped reduce incident volumes by 5x, accelerated provisioning, and freed up more time for their technology teams to focus on customer outcomes rather than infrastructure issues. As Jack Neary, Group Technology Director, puts it: "It's all about strong foundations for the AI era and Version 1 have knocked that out of the park." Read the full story to see how Key Group approached its managed services transition, and prepared for its next phase of AI and Microsoft investment: https://lnkd.in/eqF9DiZb Tim Parker | Brad Mallard | Marcus Cox | Barry McIvor | Tegan Williams | Conor McColgan | Rosie Whyte | Paul Chawke | James Lees | Alison Say

Tokenomics is moving up the agenda for technology and finance leaders. As AI adoption scales, the conversation is shifting from experimentation to ownership, governance, model choice and ROI. On 7th October, we're bringing together senior leaders at our AI Studio in Dublin for a closed-door discussion on managing AI costs, balancing risk and proving business value. Hosted by Brad Mallard, our CTO, the session will feature insights from Stuart Halford, CIO at Goodbody and Patrick Cooney, our CFO. We'll then open the discussion to explore what's working, what's not, and what's next in AI spend and governance, led by Jason Pepper, Software Asset Management Practice Lead. If you’re responsible for AI spend and finance governance at an enterprise organisation, we’d love for you to join us. Register here: https://lnkd.in/eRhs7QHP

33 months. Then 18. Then 12, 11, and now under 9. That's the gap, generation by generation, between a frontier AI release and an open-weights model of similar capability you can run on a gaming PC. The best model in the world from spring 2025 already fits on one. I charted it below, going back to GPT-3. Everyone watches the ceiling. Which lab is ahead this month, which benchmark just fell. This chart is about the floor: roughly a year behind the frontier there's a second line rising just as fast, and everything under it is free, private and yours to keep. The question that matters for real work is how much of what you use AI for actually needs the ceiling. Drafting, summarising, extraction, internal search, most day-to-day coding help. Honestly? A big share of it runs fine on the floor. And the floor of 2026 is now the frontier of 2025. So the practical skill isn't picking the right lab. It's knowing which of your workloads are floor workloads, because those are heading towards zero marginal cost and on-prem privacy on a schedule you can now roughly predict. The frontier will keep the headlines, the floor is where the economics live.