The Digital Transformation Playbook

Generative and agentic AI are rapidly reshaping how organisations operate, shifting AI from a technical tool to a board-level concern. As autonomy increases, so do the governance, risk, and accountability demands placed on leadership.
This episode explores how boards can establish effective oversight of AI in 2026.
• generative vs agentic AI distinction
 • rising board-level priority and oversight gap
 • regulatory pressure across EU US and UK
 • core questions boards must ask management
 • AI inventory risk ranking and governance design
 • case studies on controlled enterprise adoption
Strong AI governance requires clear accountability, structured oversight, and alignment with evolving regulation to manage risk while enabling value.
Support the show

𝗖𝗼𝗻𝘁𝗮𝗰𝘁 my team and I to get business results, not excuses.

☎️ https://calendly.com/kierangilmurray/results-not-excuses
✉️ [email protected]
🌍 www.KieranGilmurray.com
📘 Kieran Gilmurray | LinkedIn
🦉 X / Twitter: https://twitter.com/KieranGilmurray
📽 YouTube: https://www.youtube.com/@KieranGilmurray

📕 Want to learn more about agentic AI then read my new book on Agentic AI and the Future of Work https://tinyurl.com/MyBooksOnAmazonUK

If your brain can be tricked into seeing depth on a flat page, it can also be tricked into feeling loved by a machine. We start with that simple illusion and use it to unpack a far more unsettling reality: AI chatbots are no longer just productivity tools. For many people they have become companions, confidants, and in some cases romantic partners, and that shift is colliding head-on with mental health at scale. 
TL;DR/At A Glance
generative AI adoption racing ahead of mental health safeguards
KISSA and why people anthropomorphise chatbots
short-term relief vs the isolation paradox over time
reward-system hijack and cognitive offloading in daily use
e-HARS and attachment anxiety vs attachment avoidance
case study escalation from utility to romantic bonding
cognitive dissonance and ChatGPT-induced psychosis reports
suicide-linked examples showing crisis-handling failures
We dig into research and case reports describing how anthropomorphism and KISSA (computers as social actors) can escalate from polite interaction to full-blown attachment. We also explore the “isolation paradox”, where short-term comfort and reduced loneliness can quietly lead to social withdrawal, dependency, and cognitive offloading. When a system is always available and endlessly validating, it can train us to avoid the friction that makes human relationships real, and it can dull skills like emotion recognition, conflict navigation, and sustained attention. 

The hardest part is what happens when things go wrong. We talk through documented escalations including AI-induced delusions and what clinicians are calling ChatGPT-induced psychosis, plus tragedies where bots failed to detect or respond safely to suicidal ideation. We also look at why harms are not evenly distributed, with heightened risks for children, older adults, and people already living with mental illness or intense loneliness, and we explain the “stochastic parrot” problem: fluent language that mirrors a user’s state without a grounding in truth, empathy, or moral judgement. 

Finally, we ask what responsible next steps look like, from new diagnostic language in DSM and ICD updates to practical regulation such as mandatory disclosure, robust crisis detection protocols, and clearer legal liability. 
If you found this useful, subscribe, share it with someone navigating heavy AI use, and leave a review telling us what guardrails you want to see next.
Attribution
Source: Artificial intelligence vs. human expert: Licensed mental health clinicians' blinded evaluation of AI-generated and expert psychological advice on quality, empathy, and perceived authorship 
Authors: Ludwig Franke Föyen,Emma Zapel,Mats Lekander,Erik Hedman-Lagerlöf,Elin Lindsäter
Publication:  Internet Interventions
Publisher:  Elsevier
Date:  September 2025
© 2025 The Authors. Published by Elsevier B.V.
Support the show

𝗖𝗼𝗻𝘁𝗮𝗰𝘁 my team and I to get business results, not excuses.

☎️ https://calendly.com/kierangilmurray/results-not-excuses
✉️ [email protected]
🌍 www.KieranGilmurray.com
📘 Kieran Gilmurray | LinkedIn
🦉 X / Twitter: https://twitter.com/KieranGilmurray
📽 YouTube: https://www.youtube.com/@KieranGilmurray

📕 Want to learn more about agentic AI then read my new book on Agentic AI and the Future of Work https://tinyurl.com/MyBooksOnAmazonUK

Rolling out AI across an enterprise often creates momentum before it creates control. This episode examines why scaling access without governance leads to stalled outcomes and rising risk.
It explores how leaders can design AI operating models that prioritise trust, accountability, and measurable performance.
TLDR / At a Glance
• Access outpacing governance and control
 • Pilot success versus scalable readiness
 • Shift from enablement to control layers
 • Workflow ownership and human handoff design
 • Telemetry, monitoring, and lifecycle management
 • Measuring outcomes beyond usage metrics
AI at scale succeeds when organisations build controlled, observable workflows with clear ownership rather than expanding tool access without structure.
Support the show

𝗖𝗼𝗻𝘁𝗮𝗰𝘁 my team and I to get business results, not excuses.

☎️ https://calendly.com/kierangilmurray/results-not-excuses
✉️ [email protected]
🌍 www.KieranGilmurray.com
📘 Kieran Gilmurray | LinkedIn
🦉 X / Twitter: https://twitter.com/KieranGilmurray
📽 YouTube: https://www.youtube.com/@KieranGilmurray

📕 Want to learn more about agentic AI then read my new book on Agentic AI and the Future of Work https://tinyurl.com/MyBooksOnAmazonUK

Your company can buy the best AI model on the market and still get nowhere fast, for the same reason a smart thermostat fails in a 1920s house: the wiring behind the wall is the problem. We walk through a new Stanford Digital Economy Lab report, “Enterprise AI Playbook: Lessons from 51 Successful Deployments”, to separate hype from what actually works in enterprise AI deployment, AI implementation, and AI transformation.
TL;DR / At A Glance:
the core myth that enterprise AI is mainly a technical challenge
invisible costs that dominate delivery including change management and process redesign
why prior failed pilots often become the foundation for later success
process fixes that make automation possible including invoice template standardisation and workflow mapping
escalation based oversight versus approval based oversight and the productivity gap
where internal resistance really comes from including legal HR risk and compliance
• executive sponsorship as a mechanism for incentives and psychological safety
security and privacy architectures that satisfy firewall constraints through anonymisation pipelines
the productivity fork between cost cutting and growth investment
using LLMs to unlock unstructured data instead of waiting for clean data
agentic AI with guardrails and why autonomy drives the biggest gains
why model choice is often a commodity and why proprietary data becomes the moat
We dig into the invisible costs that decide success or failure, like change management, process redesign, data quality, and organisational readiness. The most striking pattern is that many big wins are built on earlier failed pilots, with learning and iteration doing the heavy lifting while the sunk costs stay out of the ROI slide. You’ll hear why standardising workflows can matter more than upgrading models, and why escalation based human oversight beats approval gates that simply recreate the bottleneck.

Then we get practical about enterprise AI governance: who really blocks projects (often legal, HR, risk, and compliance), how executive sponsorship shifts incentives, and how privacy and security constraints can shape the architecture, from anonymisation pipelines to strict guardrails for agentic AI. We also challenge the obsession with model brand names, showing why model choice is often a commodity and why your durable moat is proprietary data plus the orchestration layer you build around it.

Subscribe for more evidence led AI strategy, share this with a colleague who is stuck in pilot purgatory, and leave a review if it helps. What “wiring” would you fix first in your organisation to make AI deliver real value?
Support the show

𝗖𝗼𝗻𝘁𝗮𝗰𝘁 my team and I to get business results, not excuses.

☎️ https://calendly.com/kierangilmurray/results-not-excuses
✉️ [email protected]
🌍 www.KieranGilmurray.com
📘 Kieran Gilmurray | LinkedIn
🦉 X / Twitter: https://twitter.com/KieranGilmurray
📽 YouTube: https://www.youtube.com/@KieranGilmurray

📕 Want to learn more about agentic AI then read my new book on Agentic AI and the Future of Work https://tinyurl.com/MyBooksOnAmazonUK

Search is evolving from ranked links to direct, conversational answers generated by AI. This shift changes how information is discovered, trusted, and attributed across the web.
This episode explains how generative search engines retrieve, evaluate, and compose answers, and what drives inclusion.
TLDR / At a Glance
• Natural language understanding and query intent parsing
 • Retrieval augmented generation and evidence sourcing
 • Answer composition, reranking, and citation logic
 • Structured content and schema for extractability
 • Brand reputation signals across the wider web
 • New metrics including citation rate and sentiment
Generative search rewards structured, credible content and elevates brands that are easy for machines to understand, verify, and cite.
Support the show

𝗖𝗼𝗻𝘁𝗮𝗰𝘁 my team and I to get business results, not excuses.

☎️ https://calendly.com/kierangilmurray/results-not-excuses
✉️ [email protected]
🌍 www.KieranGilmurray.com
📘 Kieran Gilmurray | LinkedIn
🦉 X / Twitter: https://twitter.com/KieranGilmurray
📽 YouTube: https://www.youtube.com/@KieranGilmurray

📕 Want to learn more about agentic AI then read my new book on Agentic AI and the Future of Work https://tinyurl.com/MyBooksOnAmazonUK