This Week in Leading AI

Leading AI

Imagine two mates at the bar. Thirty years of business between them. And all they want to talk about is AI. That's "This Week in Leading AI". The podcast where Kieron and Neil cut through the hype, share what's really working in the world of Generative AI, and helping people figure out this AI thing without the techno-babble. Just honest conversation, real stories from the AI coalface, and the kind of straight-talking advice you'd only get from people who've worked together for 30+ years, been there, done that, broken things, gone "Oh S***!, fixed it, and lived to tell the tale. They claim Leading AI is the best job they've ever had and are having a blast doing it. It shows.  Warning: may cause you to actually enjoy learning about AI  Pull up a stool. We'll get the beers in.

  1. 5d ago

    Yer Granny's breeches

    Week 15. Kieron is back from six days on the beige boat of joy — a 1980s river cruiser that he describes as "full on retro". Neil describes Kieron as the “Magnum P.I. on the Thames”. If you’re lucky enough to be under 50, look it up. Kieron said his six-year-old son's favourite thing about the whole trip was watching YouTube on the iPad. The joy of parenting in 2026. "Neil, you're the biggest problem in the company" 😬 Both Kieron and Neil independently asked Claude to give them brutal, honest feedback on their business — and told it not to worry about their feelings. Neil went first. Claude told him he was the biggest blocker in the company. Then Kieron ran the same exercise. Same result. They're now wondering if Claude just tells everyone they're the biggest problem — which, frankly, might be the kindest way to land a difficult truth. The more useful output: stop comparing yourselves to Copilot, focus on the context layer, security, compliance and control — that's where the real value is. And get a non-exec director. Claude was quite insistent about that last one. Scaling — what would we do if we launched today? 🚀 Kieron's river-based reflection led to a genuinely important strategic question: what would Leading AI do differently if they were starting from scratch today? Not abandoning anyone — that's not who they are — but drawing a line in the sand and asking what a cleaner, simpler, better-priced version of the business looks like. The pricing team session from last week fed into this, with Claude doing a brilliant job of synthesising everyone's thoughts and — as Neil noted — flattering Kieron just enough to make his ideas feel more persuasive than they perhaps were. Behavioural science at work. Three customer stories 📋 Neil had a busy week of conversations. A private sector organisation selling personalised services to global clients wants to partner with Leading AI to scale their communications — consistent messaging across multiple regions, multiple languages, explainability and accuracy baked in. A NOC/SOC security company (Network Operations Centre and Security Operations Centre, for those who didn't know — Neil didn't) wants to use KnowledgeFlow to get the right information to the right people in real time during incidents. And a large organisation doing lots of public sector bidding got in touch off the back of the 100/100 bid story — they want to explore BidWriter. Meeting on Monday. BidWriter impresses a fellow boater 🛶 Kieron set up BidWriter for an organisation at speed last week — they got their documents in the day before the tender deadline, used it straight out of the box on a big bid, and the feedback from the team was excellent. Kieron bumped into the CEO on the river over the weekend. The CEO was delighted with the result. Neil noted that the CEO still hasn't signed the contract. His suggestion: contracts should only be signed, in person, on the beige boat of joy from now on. Claude Opus 4.8 is out 🤖 Kieron flags the new Opus 4.8 launch. Key improvement: it's reportedly better at knowing when it doesn't know something — which is genuinely hard to achieve. An AI that doesn't know what it doesn't know is a hallucination machine. An AI that flags uncertainty is a trust machine. The question is how Anthropic is achieving it, because the model itself doesn't know what context it's missing. Meanwhile, Microsoft has apparently cut their internal Claude licences because it's too expensive. You can't run Opus at scale on a thin margin. Sonnet is the sensible choice for KnowledgeFlow — and even then, token costs are a real consideration. Model switching — the Hoover analogy 🌀 Why would you turn the suction down on a Hoover? Kieron doesn't understand that setting. Neil does — you wouldn't you’re your Granny's teapot flying up the tube. The analogy maps perfectly to model switching: customers say they want it, but what they really want is the best model for the job without having to think about it. The answer is automated intelligent switching — a governing agent that picks the right model for the task. Leading AI monitors models constantly. Most customers don't — and AI marketing on LinkedIn isn't exactly an unbiased source of guidance. Neil is heading to Scotland. He says his friend there is lovely, but after two beers he can't understand a word he says. He just nods sagely and reverses his position if the man looks grumpy. To be fair, Neil does that with pretty much everyone when he’s had a drink. 🤣 Two mates. A bar. Thirty years of business between them. And all they want to talk about is AI. Pull up a stool — we'll get the beers in. 🍺

    29 min

  2. May 26

    "Can Me Mam Use This?"

    Episode 14: Steam Engines, 100 Out of 100 & Can Me Mam Use This? 🍺 Week 14. Veterans now, apparently. Kieron is off to pootle on the Thames on the beige boat of joy after this. Neil is enjoying the last few days of summer in the Lake District before the cold winds of June blow through. The pantomime horse of a podcast rolls on. The steam engine analogy — and why it changes everything 🏭 Kieron opens with a cracking insight from behavioural scientist Rory Sutherland of Ogilvy. When electricity replaced steam power, early adopters simply swapped their giant steam engine for a giant electric motor — and changed nothing else. The efficiency gains were marginal. The real prize only arrived when someone realised electric motors could be miniaturised and built into each individual machine, allowing the entire factory layout to be redesigned from scratch. The assembly line followed. The analogy for AI is exact: most organisations are still swapping the steam engine. They're bolting AI onto existing workflows and getting 10 minutes here and there. The transformational gains come when you rethink the factory entirely. Not many organisations are there yet — but the ones that get there first will be very hard to catch. AI curious, AI opportunistic, AI first 🎯 Neil's three categories of organisation, drawn from Gartner. Most are curious — trying one thing, dipping a toe in, not yet convinced. Some are opportunistic — using AI where it obviously helps. Very few are AI first — genuinely rethinking how they operate with AI at the centre. The gap between AI curious and AI first is widening every week. And the AI deniers? Still out there. One of them is close to Neil. The doghouse awaits. Moving to GPT 5.4 — the nuances nobody warns you about 🔧 Leading AI has been migrating customers to GPT 5.4 mini — cheaper on tokens, better at reasoning, faster. But the move from 4.1 hasn't been frictionless. The same prompts behave differently. Tiny changes in how a model interprets an instruction can cause it to pull slightly different data, produce slightly different structures, introduce subtle errors that are genuinely hard to spot. Kieron has spent most of the week in the weeds of it. The quality of the writing and reasoning is better — but getting there takes real expertise, meticulous testing, and a lot of prompt rewriting. This is exactly why "our IT team can build this" is a dangerous sentence. Prompt engineering is so 2025 — Nate B. Jones 💬 Neil references Nate B. Jones' latest Substack: in 2025, you prompted AI like a smart intern — detailed, structured, prescriptive. In 2026, with the newer reasoning models, treat it like a strategic advisor. Give it your hypothesis, your context, your data — and ask it to challenge you, push back, and find things you haven't thought of. Kieron accidentally proved this the day before: his pricing strategy prompt included "I don't know what I don't know, come back with other ideas" — and the output was brilliant. Neil's verdict: you were already doing 2026 prompting without knowing it. 100 out of 100 🏆 Neil's BidWriter story of the week. A customer used KnowledgeFlow BidWriter to produce a first draft, then used the built-in evaluation scoring to identify weak sections. They refined the weak areas, ran it back through evaluation, consistently hit 5 out of 5 for every question — and then the actual bid evaluation team came back with 100 out of 100. A perfect score. On a competitive tender. Neil has never heard of it before. Neither have we. Healthcare and the move from single tool to platform 🏥 The public sector customer conversations are shifting. It's no longer "can I have Policy Buddy?" It's "how do I give a whole cohort of people — across primary care, secondary care, physiotherapy, any healthcare provider — consistent knowledge and consistent messaging?" Linked to the work Leading AI did with North East Lincolnshire and the Care Plus Group on multilingual cancer information. The private sector version: how does an organisation become the trusted hub of knowledge for their customers — explainable, reliable, verifiable? Context layer. Data in order. Answers that can be traced back to source. That's the differentiator. The future of education assessment 🎓 Kieron thinks big. AI detectors don't work. Handing in written work is already mistrusted. But nobody's talking about the real prize: AI can assess on everything, over a whole career — every essay, every contribution, every presentation, every rugby match. A data file on each individual that tells employers far more than a grade in an exam taken on one morning when they weren't feeling great. No more failing people who happen not to suit academic-style assessment. There are vested interests in the way. But it needn't be like this anymore. "Can me mam use this?" — Chris Quickfall from Cognassist 🏆 Kieron gives a well-deserved shoutout to Chris, CEO of Cognassist, who was speaking at the FE Tech conference this week about the upcoming SEND white paper and the growing neurodiversity challenge in colleges. His product design test: his mum is a teacher, not great with technology. If she can use it, they ship it. Four words. Best product design philosophy of the year. Kieron is pootling on the Thames. If you see the slow beige boat of joy, give him a wave. Two mates. A bar. Thirty years of business between them. And all they want to talk about is AI. Pull up a stool — we'll get the beers in. 🍺

    37 min

  3. May 19

    Gartner - The Glastonbury of AI

    Episode 13: Live from the Glastonbury of AI — Our Gartner Debrief 🍺 Week 13. Unlucky for some — but not for two people who've just spent three days at the Gartner Data and Analytics Summit, AKA the Glastonbury of AI. Neil says he was nearly as exhausted after three days sitting down as after five days at the actual Glastonbury. What goes on at Glastonbury stays at Glastonbury. But what goes on at Gartner comes out on this podcast! The emotional rollercoaster 🎢 Monday evening beer: both of them felt reassured. Everything they'd heard confirmed Leading AI was on the right track. Tuesday: Kieron doubted everything — tech bro language, acronym soup, imposter syndrome at full volume. By Wednesday: unpack the jargon, and it turns out they already knew most of it, and in some cases were ahead of it. Neil's summary: two separate meetings with Gartner specialists, both said we were on the right track. Donald's response to Neil's email was along the lines of: "I'm on it, stop hassling me"... only ruder. Context is king 👑 Kieron's biggest takeaway. The context layer — telling your AI what your organisation is, what your data means, and how different teams need to use it — is the difference between good retrieval and bad retrieval. The example: ask an AI "how many sales did we have this quarter?" Without context, it doesn't know what "sales" means (invoiced? agreed? handshake?), what your financial quarter is, or which column in your MIS system to look at. KnowledgeFlow already builds data dictionaries automatically when it loads data — but there's more to do. Knowledge graphs are the next step: storing context so the AI can pick up the right layer depending on who's asking. Ontology, knowledge graphs and semantic layers — explained for humans 🧩 Neil was confused by the three terms being used interchangeably at Gartner. He asked Perplexity to explain the difference like a 15-year-old. The answer: think of a school. The ontology is the rule book (what is a teacher, what is a pupil). The knowledge graph is the directory (Bob is a teacher, Alice is a student). The semantic layer is the notice board (how many pupils are in Year 10?). Get all three in place and your retrieval gets dramatically better. Turns our they're already doing a lot of it — they just didn't know it had a name. Feedback loops — the missing piece 🔄 Kieron's second big theme. The agentic email system works — it reads inboxes, triages, drafts responses, handles routine inquiries automatically. The next challenge: capturing what happens when a human looks at the draft. Did they send it unchanged? Edit it slightly? Rewrite it entirely? That data, captured over hundreds of interactions, tells you which types of email to fully automate and which ones still need a human. For a Housing Association, if 297 out of 300 pet policy inquiries sent unchanged are sent unchanged, automate your pet policy. The challenge: you only capture that feedback if the human stays in the platform rather than copying and pasting out of it. Which leads neatly to... How do you make KnowledgeFlow so good people feel stupid going anywhere else? 💡 Neil's challenge to the team. Inspired partly by Gartner's focus on designing solutions that disappear — like the GP recording consultation tool that lets doctors look at patients instead of screens. And inspired partly by the stat that doctors interrupt patients after an average of 18 seconds. If the technology is invisible, the human interaction improves. Ibby and Donald are already building something. Watch this space. Human in the lead, not human in the loop 🧠 One of Gartner's sharpest lines. Don't just put humans in the loop to click okay, okay, okay — they'll stop paying attention and let everything through. Use humans where human judgment actually matters. Pet inquiry? Automate it. Mould report or smell of gas? Human in the lead, immediately. The Gartner stats (cos Neil's a stato at heart) 📊 Only 6% of AI leaders surveyed believed their organisations and people were AI ready. Only 12% felt their data was properly secured and governed. And fewer than 50% of organisations currently track their AI costs. Gartner's framework: are you AI cautious, AI plus, or AI first? Because if you're cautious while your competitors are AI first, you're already losing ground you may not get back. The bonkers corner 🤪 The futurist with purple shoes was very entertaining. Neural prosthetics already exist that let you move a hand by thought. But would you take a cheaper version if it played ads? (There's a Black Mirror episode about this.) Would you let your employer connect to your neural network and pay you for time spent thinking about work? A woman in the audience laughed so hard she got the microphone. Her response: "If that happens, I'm screwed. I don't think about work all that much." Final thought from purple-shoes: would you want your wife to connect to your neural network? Neil's verdict: the doghouse would be permanent. Product of the week 🎵 (hum the jingle) KnowledgeFlow now has memory. Kieron tested it by telling it to start every response with "Hey dude." Forgot he'd done it. Later asked it to analyse some data. It said: "Hey dude, here's the analysis." The serious version: memory means KnowledgeFlow can remember your role, your preferences, your output formats — securely, inside your own Azure tenancy. Something Claude can do publicly, but not securely. KnowledgeFlow now can. Neil is in Scotland in the sunshine. He's knocking the top off a beer and going into the garden. Neil's wife Helen should probably (definitely) not connect to his neural network. Two mates. A bar. Thirty years of business between them. And all they want to talk about is AI. Pull up a stool — we'll get the beers in. 🍺

    38 min

  4. May 12

    The Billion Pound Bid

    Buzzsprout Description Episode 12: Sir David Attenborough, Magic Wands, and prep for the Glasto of AI 🍺 Week 12. The pantomime horse of a podcast is back. Kieron is heading to David Attenborough's 100th birthday picnic after this. Sort of. It's at his son's school, 100 metres from Sir David's house. There is a non-zero chance of Sir David making an appearance. Neil wants photographic evidence if he does. Next week: both of them are off to the Glastonbury of AI — the Gartner Data and Analytics Summit in London. The Leading AI team back is bracing itself for what whacky ideas they come back with. Neil's uncle — 100 years, a minesweeper and a ZX Spectrum 🎖️ Neil lost his uncle two weeks ago — one month short of his 100th birthday. A man who served on a minesweeper in the Pacific in the Second World War and saw huge technological change in his lifetime. . Neil reflects on the pace of that change: TV took 13 years to reach 50 million users. The internet took four years. ChatGPT reached 100 million users in two months. AI isn‘t slowing down.  Shadow AI — the problem that won't go away 👤 If your organisation is forcing people to use the “default AI” (yes, you Copilot) and it keeps stopping halfway through a task, don't be surprised when they quietly spend $20 a month on Claude and get the job done properly. The shadow AI problem is real, and it's growing because the default can’t do the job. Microsoft has been quietly removing Copilot from places where nobody's using it. A deleted social media post says it all. Safeguarding — its important!🔒 Neil flags two major developments this week. Meta has just lost a case in New Mexico over making their tools addictive for young people. And Google and Character.AI settled out of court in January with families suing them over the role AI played in the deaths of young people. Neil recommends a piece by Nate B. Jones (now on his fourth mention — no, still not on commission) on why parents need to talk to their children about AI relationships. It's not a chatbot. It doesn't understand you. And if you don't explain that clearly, the consequences can be devastating.    Product of the week — the Agentic Bid Writer goes live 🎉 Donald created a working prototype this week which Kieron and Neil have both been demoing… but ‘forgot’ to tell Donald. I guess they didn’t want to bother him… 😂 Kieron said: "What if the first time you even saw a new opportunity, you already had an 80% complete bid ready to go?" Which led to... The billion-pound consortium bid 💷 A consortium of seven organisations has approached Neil about using KnowledgeFlow BidWriter to write a £1 billion bid together. Seven separate KnowldgeFlow assistants, each loaded with their own technical docs, marketing material, insurance certificates, and case studies. Press the button. First draft in ten minutes. For anyone who's ever spent three weeks chasing consortium partners for content they should have sent on day one, this is the dream. ROI — if you switched off your AI today, would anyone notice? 📊 Neil poses a question someone put to him this week: if you cancelled all your AI subscriptions tomorrow, would your business notice? The answer for most people is: immediately and painfully. That's your ROI right there. And it links back to the social workers turning down jobs at councils without AI tools. The gap is real. It's widening. And the organisations still asking "but what's the return on investment?" are the ones who'll find out the hard way. Kieron may or may not meet David Attenborough at a school picnic this afternoon. Neil wants a photo. We all want a photo. Two mates. A bar. Thirty years of business between them. And all they want to talk about is AI. Pull up a stool — we'll get the beers in. 🍺

    45 min

  5. May 5

    AI Code with your fries?

    Episode 11: McKinsey Got Hacked, McDonald's Writes Python & We're Big in Uzbekistan 🍺 Week 11. Eleven weeks of consistent podcasting — a personal consistency record for both of them. 🥳  The podcast now has listeners in Venezuela, Malaysia, Kenya, Ukraine, Vietnam and Uzbekistan. Almost certainly the same person with a very well-travelled VPN. Hello to all our world listeners. Kieron stopped off at Neil's northern castle on the way back from a Housing gig in Glasgow. He took the sunshine with him when he left.  McKinsey got hacked — and it's a warning for everyone running RAG AI 🔐 Donald WhatsApped Kieron the news at 7am. McKinsey's internal RAG system, Lilly, was breached in March — 100,000 documents, 57,000 user account details, and the prompts, all exposed through 22 open endpoints. Probably an AI tool that spotted their JSON file formats and quietly helped itself. The lesson: if you're running RAG AI without regular penetration testing, you're hoping for the best. Leading AI runs pen tests constantly. Donald spotted and locked down a minor exposed endpoint the same morning. That's what vigilance actually looks like. Prompt injection — and the McDonald's Python developer 🍟 Prompt injection is the art of slipping instructions into an AI to make it do things it shouldn't. The McKinsey version is terrifying. The McDonald's version is brilliant: a customer asked their support AI to help him finish a Python script before ordering chicken nuggets. It obliged. He announced he was cancelling his Claude subscription. £20 a month versus unlimited nuggets. With large fries and a milkshake. Pricing — transparency, tokens and not getting ripped off 💷 The strategy session in Penrith produced a really important conversation. Token pricing is confusing, opaque, and vaguely terrifying (see the £150k overnight bill from Episode 10). Neil's take: be radically transparent. Fixed costs, consumption costs, kill switches, and using mini models that are 10 times cheaper. It's the right thing to do so customers understand what they're buying. Even if his mates it the pub call him a "soft lefty". New wins 🎉 A new council social care customer. And a trade body confirmed on Wednesday that KnowledgeFlow's Policy Buddy is being deployed across 10 member organisations. Shared knowledge, shared learning, shared insights across a geographically dispersed group. The kind of national change infrastructure Kieron and Neil have spent careers building, now with AI baked in. Kieron at the Share Annual Conference 🏴󠁧󠁢󠁳󠁣󠁴󠁿 Kieron spoke at the Share Annual Conference and Awards in Glasgow. A packed room and an honest conversation about what Copilot can and can't do. Most people in the room were using AI to summarise documents. Which is fine, but it's also the worst way to let AI bias creep in unchecked, because Copilot doesn't know who you are, what you care about, or what a good summary looks like for your organisation. KnowledgeFlow does. AI observability — the Glastonbury of the AI world 🔬 Kieron and Neil are heading to the Gartner Data Analytics Summit in a couple of weeks.  They describe it as the Glastonbury of the AI world. Kieron is on a mission to track down Wilco Van Ginkel, Senior VP at Gartner, who he mat last year. Wilco's latest research is on live AI observability and evaluation, which is exactly what Leading AI is currently building. The key insight: you can't test AI output like a calculator. You mark it like an essay. And you need to keep testing live because model drift happens. Wilco, brace yourself. The Cumbria Clock Company, repairers of Big Ben, get a heartfelt farewell shoutout. Keith the pirate clockmaker may yet appear as a guest. Two mates. A bar. Thirty years of business between them. And all they want to talk about is AI. Pull up a stool, we'll get the beers in. 🍺

    41 min

  6. Apr 28

    From pilots to practice

    Episode 10: TechUK Recognition, You Get What You Pay For & The Very First Cyber Attack 🍺 Ten weeks. The longest either of them has been consistent at anything. Neil's briefly out of the doghouse. Kieron's on squash instead of beer. And Leading AI has just had two case studies published in one of the most important AI reports of the year. Pull up a stool. TechUK's "From Pilots to Practice" report — and we're in it. Twice. 🏆 TechUK — the UK's leading technology trade association representing over 1,100 member companies — published their landmark report From Pilots to Practice: Using AI in the Public Sector with 19 real-world case studies showing how AI is genuinely transforming public services. It features not one but two KnowledgeFlow implementations: the FE college quality assistant and, in housing, Taff Housing. Being independently selected by TechUK as an example of AI that actually works in practice — not just in pilots — is a significant validation.  Agentic AI for housing — getting serious with Taff Housing Kieron was with the Taff Housing CEO, director of technology and senior team this week — a monthly meeting they hold to track progress and plan what's next. On the roadmap: a fully agentic tenant inquiry system that triages and instantly answers routine emails, freeing frontline teams for complex cases. And an AI-powered repairs checker that reads job descriptions, checks them against schedule of rates codes, and flags when a contractor's invoice looks a little... creative. Yes, Kieron finally remembered to show the photograph this time. First draft of a  £500 million bid written in 4.5 hours Neil used KnowledgeFlow's BidWriter to produce the first draft of a 7,500-word tender response for a £500m contract in four and a half hours. What would have taken a week landed at 80% complete before Tuesday lunch.  McKinsey now tests candidates on AI prompting Not whether they know about AI — whether they can prompt well, challenge outputs, and think critically alongside it. Social workers are already turning down job offers at councils without AI tools. Now the world's top consulting firm is screening people out if they can't work with AI. The direction of travel has never been clearer. The IQ of your AI depends on what you're paying for Anthropic's Opus 4.7 has an estimated IQ equivalent of around 140 — top percentile of humans. The free tools? Closer to 100. It's like hiring a £20k accountant versus a £100k one. If you tried AI and thought it wasn't impressive, you were probably using the wrong model. You get what you pay for. The $150,000 overnight token bill A company set an AI agent running overnight. By morning, Google presented them with a $150,000 token bill. KnowledgeFlow is now building automatic kill switches for all client deployments. And a brilliant tip from Nate B. Jones (third plug this series, still not on commission): convert your PDFs to markdown before loading them into AI and you'll save up to 87.5% of your token costs. Most people don't bother. Most people will when it starts hitting them in the wallet. The first ever cyber attack — 1834 Neil drops a history bomb: the first recorded technology attack happened in France in 1834, when two men hacked telegraph wires to manipulate financial markets. Kieron's response: what's actually changed?  Ten weeks in. Still going. Still on squash instead of beer. Apparently Kieron has "aura" though, so things are looking up.  Two mates. A bar. Thirty years of business between them. And all they want to talk about is AI. Pull up a stool — we'll get the beers in (even if we're not drinking them ourselves). 🍺

    42 min

  7. Apr 21

    Neil's in the doghouse

    Episode 9: AI Deniers, AI Slop & KnowledgeFlow Cracks Salesforce 🍺 Neil's on a non-alcoholic beer again — this time because he's in the doghouse with Mrs Watkins and needs to drive her to a romantic weekend away to patch things up. It's that kind of Friday. Welcome to Episode 9. Mrs Watkins is an AI denier — and she's not alone Neil tried to convince his wife of the wonders of AI. She said "it's great but it's not for me." Sound familiar? Neil points out this is almost word for word from Richard Susskind's How to Think About AI — a whole category of people who can see the value but won't change their process to fit around it. Then again, Staples' share price apparently collapsed when Mrs Watkins switched from post-it notes to spreadsheets, so perhaps there's hope. The future of management — courtesy of Nate B. Jones Neil recommends a brilliant piece by Nate B. Jones (his second plug this series, and no, he's not on commission) on what management is actually for. Three roles: routing information to the right place, sense-making in the noise, and accountability. How much of each is AI-able? More than most managers would like to admit. Trust in a world of AI slop Can you trust a video anymore? A photo? A LinkedIn post? Kieron raises the uncomfortable reality that AI-generated content is everywhere — including on Instagram (those animal rescue videos? Mostly fake). The organisations that will win are those with genuinely trusted brands and curated data sets — like the King's Fund or Stripe. Being a trusted source is now a competitive advantage. Anthropic's Claude 4 Opus (Mythos) — the model you're not allowed to have Kieron digs into the buzz around Anthropic's most powerful model, apparently so capable it performed zero-day attacks on every major operating system in its first outing. Is it genuinely that dangerous? Or brilliant pre-IPO marketing? Five security questionnaires, 20 hours, and a lot of AI slop Neil spent most of the week answering overlapping, partially nonsensical security questionnaires from a prospective customer, most of them clearly generated by ChatGPT (the M-dashes are a giveaway). The cobbler's children moment: Leading AI built a KnowledgeFlow security RAG for themselves mid-episode and were answering questions live before the call ended. Twenty hours of pain, sorted in minutes. Outcomes-based pricing — the next frontier Goldman Sachs says AI companies are moving away from per-seat licensing toward outcomes-based pricing. Kieron has already had the first conversation about it: a college currently pays £20 per student application document check to an outsourced company. KnowledgeFlow can do the same thing for about 30p of AI processing. The maths are not subtle. Product of the week 🎵 (build to a crescendo) KnowledgeFlow now connects directly to Salesforce via live API calls. No more exporting data, no more waiting for reports, no more paying tens of thousands for bespoke dashboards. Any staff member can now ask questions of their Salesforce data in plain English and get instant answers — across multiple data tables, in real time. A personal moment — Kieron's dad's care plan Kieron's father moved into a care home in January. The family received a 16-page care plan full of jargon and boxes nobody understood. Kieron put it into KnowledgeFlow, asked "what can I do to help?", and got five clear bullet points back. He sent it to the family WhatsApp. Everyone said thank you. That's what this is actually for. Plus: house manuals loaded into a RAG, and the ongoing mystery of whether Neil will successfully escape the doghouse by Monday. Two mates. A bar. Thirty years of business between them. And all they want to talk about is AI. Pull up a stool — we'll get the beers in.  Even non-alcoholic ones for those who want them.🍺

    39 min

  8. Apr 14

    Copilot is for entertainment purposes only

    No beer again this week — just water, Coke, and the usual dose of brilliant conversation (Ed. - Seriously? Who writes this stuff?).  AI liability — who's actually responsible when it goes wrong? Kieron sat down with Peter Lee of Simmons & Simmons, head of AI governance, to ask the question nobody has a clean answer to yet: when AI gives someone wrong information that affects their life, whose problem is it? The short answer? There's no case law yet in England. The longer answer involves multi-layer liability chains, emerging insurance products, the EU AI Act, and why Leading AI's obsession with privacy, accuracy and monitoring means they're already further ahead than most. Peter's comment? It's really interesting that you're thinking about this — because most aren't. Copilot is for entertainment purposes only. Verbatim. Kieron found it in the terms and conditions. Microsoft's own Copilot licence states — and this is a direct quote — "Copilot is for entertainment purposes only." It also confirms Microsoft makes no warranty that responses won't infringe copyright, defame anyone, or actually work as intended. And if you share the output? Entirely your problem. This sits beautifully against everything they said about AI liability ten minutes earlier. ISO certifications, the EU AI Act and why it keeps Kieron awake at night Leading AI holds both ISO 42001 and 27001 — among a low hundreds of UK organisations to have done so when they got them. The EU AI Act defines "high risk" as tools that affect people's lives. Some of KnowledgeFlow's tools clearly fall there. Being worried about it, they agree, is probably the right response. Product of the week 🎵 (your jingle here) Sentiment analysis is now live in the KnowledgeFlow admin console. The system flags when users push back on responses — when someone says "no, that's not what I meant" or "that's great." Early warning signals before problems get reported. Combined with the ongoing work on client impact reports, this is all part of the push to measure real-world outcomes, not just prompts and tokens. Smart targets, weekly parent reports and the 25% problem Up to a quarter of teachers leave within their first year. Neil raises the question: what if better tools could change that? Smart targets written weekly instead of termly. Parent reports sent regularly instead of once a term. Personalised, data-driven, done in minutes. The conversation about what this could mean for teacher retention — and student outcomes — is a genuinely important one. AI agents having an argument Oscar (Kieron's 19-year-old son) is building a multi-agent system — a project manager running five AI agents, the clever ones on cheaper models. He set a $3 budget. Two of the agents started arguing with each other and burned all the money. His solution: build firewalls between them so they can only communicate via the project manager. As Neil points out: that's why project managers exist. Vendor lock-in, the end of Salesforce, and helium Neil raises a real-world case: a company used AI to replace its risk management software entirely by hoovering up Teams transcripts, loading them into an LLM, and getting daily priorities out the other side. No Salesforce needed. Then things get geopolitical — it turns out making AI chips requires helium, a third of the world's helium comes from Qatar and can't currently get out of the Strait of Hormuz, and after 40 days on a ship it starts to deteriorate. Token costs going up. Chip costs going up. Energy costs going up. The Large Hadron Collider once had a tonne of helium leak. The scientists sounded hilarious on the radio. Neil wishes they'd had beer. Two mates. A bar. Thirty years of business between them. And all they want to talk about is AI. Pull up a stool — we'll get the beers in. 🍺

    40 min
See All (15)

About

Imagine two mates at the bar. Thirty years of business between them. And all they want to talk about is AI. That's "This Week in Leading AI". The podcast where Kieron and Neil cut through the hype, share what's really working in the world of Generative AI, and helping people figure out this AI thing without the techno-babble. Just honest conversation, real stories from the AI coalface, and the kind of straight-talking advice you'd only get from people who've worked together for 30+ years, been there, done that, broken things, gone "Oh S***!, fixed it, and lived to tell the tale. They claim Leading AI is the best job they've ever had and are having a blast doing it. It shows.  Warning: may cause you to actually enjoy learning about AI  Pull up a stool. We'll get the beers in.

Information

  • Creator
    Leading AI
  • Years Active
    2K
  • Episodes
    15
  • Rating
    Clean
  • Copyright
    © 2026 This Week in Leading AI
  • Show Website
    This Week in Leading AI