Chatting GPT

What actually happens when your team uses AI to write a blog post, a tender or a report, and does the law now care? Alan Thornton, founder of the AI governance platform Tipcan Solutions, joins Maryrose Lyons to make the EU AI Act genuinely usable for built environment professionals.

Alan walks us through the the latest requirement in EU AI Act compliance that is labelling. If you're using AI and sharing certain types of text content now you must label it, but not all content. A general design blog is fine. A glowing piece about a local councillor is not, unless a named human takes responsibility for it. That idea, that every firm must have real accountability runs through the whole conversation.

From there: shadow AI and where the risk really hides, how to properly inventory your AI use cases, and the fines that now apply, including a sneaky second one for handing a regulator incomplete information in a panic. Alan and Maryrose also get into what 'high risk' means once CE marks and national sandboxes enter the picture, and why Ireland's oddly fragmented approach to enforcement is worth watching.

Practical, occasionally very funny, and refreshingly free of scaremongering.

Show notes:

Shadow AI, and why the biggest risk is often the staff member you least suspect
The four risk levels, and why most firms sit in the "limited risk" transparency band
The public-interest test: one blog post needs an AI label, a near-identical one does not
The release valve, how genuine human oversight makes the labelling duty ease
How to properly inventory your AI use cases, and why a spreadsheet will not cut it
The fines that now apply, including a second, sneakier one for panicking under pressure
What "high risk" really means, once CE marks and national sandboxes enter the picture
Why Ireland's fragmented approach to enforcement is worth watching
The one thing every firm can do this week to get on the right side of the Act
Alan Thornton on LinkedIn: https://www.linkedin.com/in/alan-thornton/
Tipcan Solutions
EU AI Act · Article 50, transparency · Article 99, penalties
ISO/IEC 42001 (AI management systems)
CE marking (EU)
RICS AI standard (in force 9 March 2026)
UK Data (Use and Access) Act 2025
Ireland's EU AI Act implementation
We are AI Institute
Join us o 17 September for a webinar on: The EU AI Act, your staff, your risk: ignorance is no defense

Register now
Hosted on Acast. See acast.com/privacy for more information.

Andreessen Horowitz has a thesis that every building is still designed by software from the 1990s. Niklas Lindgren agrees, and he is building the company to change it. His claim is blunt: AI is about to change how 70 to 80% of the hours on an engineering project get done.

Niklas is the founder and CEO of Endra, a Stockholm-based AI platform built specifically for mechanical, electrical and plumbing (MEP) engineers. While we don't normally have vendors on the podcast, we made an exception in this case because Endra looks like the Revit killer everyone's been waiting for after all these years. Andreessen Horowitz certainly seems to think so, having invested in the Swedish based AI start up.

In this episode, Niklas walks Maryrose through why MEP has been so slow to change, why jobs are not disappearing even as the work transforms, how firms can build entirely new services on the back of AI, and why outsourced teams are the most exposed. He closes with the one practical step every engineering firm should take this week.

A sharp, grounded conversation about the AI rebuild of the built environment, from someone doing it inside the largest firms in the industry.

In this episode

Why MEP engineering has barely changed since the 1990s, and the three layers of complexity that kept it stuck
Why Autodesk is a partner, not a competitor
What Endra has learned from working with world-class engineering teams
Why AI will change 70 to 80% of project hours, and why jobs are going nowhere
The new products and pricing firms can build on the back of AI
Why outsourced, commoditised work is the most exposed
The one practical step every engineering firm should take this week
Why this is a people game that takes real time, not a quick pilot

Links

Endra Website
Niklas Lindgren on LinkedIn
Hosted on Acast. See acast.com/privacy for more information.

Karen Lewis Enright is a change management professional with over 20 years of transformation experience across global organisations including Johnson & Johnson and Wipro. She holds a Masters in organisational psychology, but what makes her stand out is two decades spent helping leaders turn strategy into meaningful change. She specialises in AI strategy and is the first change management professional to join ChattingGPT.

In this conversation, Maryrose and Karen get into the human side of AI adoption: how to lead when you're uncertain yourself, using the image of a slipping cup of coffee to explain where people's attention goes under stress, why the sceptics belong in the room alongside the enthusiasts, and how to build trust by being honest with people about what AI changes in how they work. Karen tells the story of a self-parking car that never gets used, a miniature version of every stalled AI rollout, and makes the case for celebrating the curious champions most organisations try to shut down.

She talks through the change management tools she reaches for, from change agent networks on-site to reverse mentoring between graduate engineers and senior partners, to knowing when to pause a BIM rollout or a new tender process and let a stretched team catch up. She closes on the advice she keeps returning to: start with the why, and get the data, the governance and the change management right before you scale the technology across the practice.

Connect with Karen Lewis Enright on LinkedIn.

Hosted on Acast. See acast.com/privacy for more information.

Most of the AI conversation is a numbers game: licences bought, people "using AI", whether the firm has a strategy yet. James Garner thinks that is the wrong number to watch. The money is flowing and the strategies are written, but the thing almost no one is funding is capability — whether ordinary professionals can actually use the tools well. That gap, he argues, is widening, and it is where the real risk now sits.

James is Head of AI at Gleeds, a Fellow of RICS, and co-founder of the Project Flux podcast. He has been a quantity surveyor since 1998, a profession he jokes you could pull someone into from 1890, give a day's training, and they would manage fine today — which is exactly what makes this moment different. He sat on the RICS working group behind its new responsible-AI standard, in force since March, which for the first time makes baseline AI literacy a requirement for every chartered surveyor, not a nice-to-have.

James walks Maryrose through why an "AI strategy" that is not a business strategy powered by AI will fail, the difference between transparency and explainability (and why he prefers "bookending" to "human in the loop"), the model-cost reckoning coming for anyone on a flat monthly plan, and the eight Claude files every professional should set up to get real value from Claude Code. Along the way, the habit that keeps him ahead: half an hour, every Friday, to do nothing but experiment.

A practical, clear-eyed and genuinely useful conversation about closing the gap between owning AI and actually using it, before it widens any further.

James Linkdln
Hosted on Acast. See acast.com/privacy for more information.

Most of the AI conversation is fixated on AGI: when machines will match human intelligence, and which lab gets there first. Dr. Craig Kaplan thinks that is the wrong thing to watch. The real story is what comes seconds later artificial super intelligence, systems a thousand times, or even a trillion times, smarter than the smartest human alive. And he believes it is arriving faster than almost anyone is prepared for.

Craig has been working on this since the 1980s. He earned his doctorate at Carnegie Mellon alongside Nobel laureate Herbert A. Simon, one of the scientists who named the field of AI, and he has spent decades building collective intelligence systems — including ones that traded billions of dollars on Wall Street and beat the best hedge funds. That track record is the foundation of his core argument in this episode: the widely held belief that safer AI must mean slower AI is simply wrong. You do not need to slow down, he says. You need a smarter design.

That design is a democratic "society" of cooperating AI agents — millions of them, each carrying the values of a separate human checking one another in the open, rather than a single monolithic black box no one can see inside. Craig walks Maryrose through why this is both more powerful and far safer, what Pope Leo's first encyclical gets right and wrong about treating AI as a mere tool, and why the values we hand these systems now, in their "childhood," will shape everything that follows. He also makes the case for sovereign AI and the Global South, where cultures like Nigeria's Yoruba risk being erased by models trained overwhelmingly on Western data. And the pace is staggering: some AI capabilities, he notes, are now doubling in as little as 1.3 months.

A sharp, grounded, and surprisingly optimistic conversation about getting the most important technology in human history right before it arrives.

Find Craig's white papers and free designs at superintelligence.com

Here's you can see Craig's Linkdln
Hosted on Acast. See acast.com/privacy for more information.