Brightbeam

Your gran thickens her gravy with flour. Your aunt swears by cornflour. They’ve both made perfect gravy for forty years. Eventually, at Sunday lunch, you ask which way is right. And then you’re reminded that your uncle can’t eat wheat. Ah. Work has its own gravy moments. Two experienced reviewers rate the same report differently. Two planners put the same changeover on different shifts. When you’re training an agent to do that work, the reflex is to settle it: pick one answer and call it correct. Sometimes that’s exactly right. Some disagreement is just noise, and worth cleaning up. But some of it is knowledge. One planner knows the fitter is on days this week. One reviewer remembers the supplier changed. And some is genuine uncertainty, which should stay open until there’s evidence to close it. You can’t tell which until you ask why. An apprentice who learns from both experts works out, over time, when each approach fits. Mark an agent against a single answer, without the reason behind it, and you may teach it that a valid alternative is wrong. So before anyone writes the answer key, find out who’s coming to lunch. Happy Monday.

The labs are guarding their secrets, shopping for yours and arguing over who pinched what. And then OpenAI found a bug wrapped in toilet paper. It’s all gone a bit le Carré. The drama unfolds in this week’s SNI – along with all the AI news worth knowing across tech, biopharma, medtech, advanced manufacturing and insurance. 👇

Much of our Medtech playbook is about making knowledge easier to reach. But then we made it 278 pages long. So it’s now online in full. And your AI assistant can read it for you - and discuss its contents. Together you can ask questions and work out how the playbook can best help your business. Having registered for free access - including all 28 deep dives - download the adviser file and give it to Copilot, ChatGPT, Claude or whichever assistant you use at work.  Which means that today you can start where the playbook suggests: with a recurring problem of your own. Ask it where you’d begin with a CAPA backlog, say, and what would need to be in place before AI could help. And if you’d rather have it on paper, we’ll still post you a free printed copy anywhere in the EU. Thank you to Irish Medtech Skillnet, Irish Medtech and Ibec for helping bring the playbook to life. Links in the comments 👇🏼

Ireland, we are told, wants to be Europe's leading applied AI economy. Here at Brightbeam, we rather like the emphasis on applied. Around 1,500 people - all expected at the RDS in Dublin for today's International AI Summit - discussing future prospects from the AI economy. A year ago, we hypothesised that the optimal division of labour would be: - The US creates AI; - China reduces its cost; - The EU regulates it; and  - Ireland implements it.  So it's good to see implementation getting this kind of national attention. But implementation still moves at very different speeds. Some organisations go from proof of concept to production in months. Others stay in ‘prototype purgatory’ for years. In our experience, the model is rarely the sticking point. More often, it's whether an organisation understands how its own decisions get made - whose judgement matters, and who answers for the result when AI is involved. Get that clear, and you can give AI a useful job to do. And what you learn from one deployment can make the next one easier. So here's somewhere to start tomorrow morning: pick one decision your team makes every week, and ask the people who make it to walk you through it. Livestream and our earlier piece in the comments 👇🏼 #AISummitIE #Ireland2026EU

Ever so proud of Melissa Proxenos and Corey Duffy for being finalists in this year’s AI Awards! Melissa is shortlisted for AI Person of the Year, recognising technical excellence and leadership. Corey is up for Young AI Role Model of the Year, recognising the work and promise of the next generation of AI leaders. We know they’re top notch. But it’s lovely to see their brilliant work recognised more widely. Fair play to you both. Congratulations to everyone else shortlisted, and thank you to AI Ireland for spotting a couple of good’uns. Bring on 17 November. #AIAwardsIRL

Did you know there's an emerging protocol for conducting human-agent symphonies? We have MCP for connecting agents to tools and A2A for connecting them with each other. But through our work with regulated enterprises, it became clear that the agent stack needed a third protocol for how people and agents work together. Delegating work, reviewing results, overriding decisions. And so Brightbeam Applied Research developed CHAP, the Collaborative Human-Agent Protocol: a common language for that collaboration, with a shared record of who did what and why. So when a person intervenes, their judgement becomes part of the record. And a team can both reconstruct what happened and use those corrections to improve results. Now, the Agentic AI Foundation, part of The Linux Foundation, has published a guide to CHAP by our own Dr Arsalan Shahid. Arsalan uses musical notation to explain the idea, then follows a support ticket from delegation through to human review – showing CHAP working alongside MCP. CHAP is open source and remains a public draft, available to test and build on. Read it here: https://lnkd.in/erqu86-b

An attack is spreading through your network, and you want it stopped. But your agent sits waiting for your approval. The hypegeist missed what a small fish can teach us about the cost of waiting. On a reef in Curaçao, researchers filmed predators trying to catch small fish. The predators were bigger and faster, closing the gap until their prey was almost within reach. Then a last-moment turn would send the little fish out of their path. The predators’ nervous systems simply couldn’t respond in time. Their reactions took just 35 milliseconds, but even that was too slow for this fast food. Across the filmed attacks, the predators missed 97% of the time. And evolution seems to have little room to shorten that delay. Microsoft’s latest Digital Defense Report describes AI shortening parts of cyberattacks from days to minutes. It argues that defenders will need to give agents more authority to respond. In any case, keeping a human involved doesn’t necessarily mean they’re in control. At September’s congressional hearing on military AI, governance researcher Anna Mysyshyn asked asked what it means when one person supervises twelve autonomous systems. That person can remain responsible without the time to exercise that responsibility. So how do you decide whether an agent should act or wait? Getting it wrong has a cost, but so does waiting. So you need to weigh both. Imagine a future with a million agents at work. You can’t approve every routine decision. And yet authority has to be earned through demonstrable improvements in the work. The people who know the work must define what good decisions look like, set authority by decision type and establish enforceable limits. There has to be a record of what people and agents decide, why, and what happens next. That evidence should determine whether authority widens or narrows, within the rules and as the work changes. You remain responsible for the work. Where a decision needs your approval, make sure you have time to judge it.

So which is it: a bubble about to burst, or a boom with world trade forecast to grow 3.9%, despite a war? You might also like to know what Giorgia Meloni has done with four seconds of her own voice. And who Tim Harford called out for misremembering. The reveals, as ever, are in this week’s SNI – along with all the AI news worth knowing across tech, biopharma, medtech, advanced manufacturing and insurance. 👇

Did you hear? The 2027 Generative AI for Medtech playbook is out. Order yours today and we’ll send you a free printed copy, anywhere in the EU. 97 professionals across six Medtech companies talked us through their working days and the problems they wanted to solve. Those conversations helped identify and assess 92 use cases, of which 28 were then selected by participants for deeper exploration. Opportunities range from planning production and assessing regulatory changes to analysing clinical data. And the playbook also covers the data, validation and controls these applications need, so experts can check and challenge their outputs. We launched it yesterday with Irish Medtech, Irish Medtech Skillnet and Ibec, and we’re thrilled to have it out in the wild. Thank you to everyone who helped bring it to life, and to those who joined us for yesterday’s launch. It’s the kind of book that works best with dog ears, highlighters, and scribbles. So order your free copy via the link in the comments. 👇🏼

There’s a lot to be said for finding next month’s production problem this month. You have time to chase a missing component, adjust staffing or rethink the plan. All considerably easier before the work is due to start. That’s what we worked on with Meissner’s planners in Castlebar. And Meissner has now been shortlisted in two categories at the Pharma Industry Awards for this project. This follows their Digital Innovation & Transformation shortlist at the Lean Business Ireland Awards, which takes place on 22 October.  The platform we built together went from concept to production in 17 weeks. It brings labour, materials and sterilisation capacity into one view, replacing two to four hours of manual preparation each week and giving planners more time to act on what they see. Congratulations to John O'Shea, MBA, Dip IoD, Joseph Convey, Tom Kelly and the wider team at Meissner, and to our own Melissa Proxenos, Corey Duffy and Darren Moen. Plenty to celebrate. Good luck at the Pharma Industry Awards on 26 November. Case study in the comments. 👇