The Digital Transformation Playbook

We track how Microsoft Copilot evolves from a drafting assistant into a system that can carry out multi-step work, build simple apps from plain language, and eventually run persistent agents while we are offline. We focus on what that shift forces leaders to decide on governance, approvals, ownership, and value, rather than counting licences and hoping for impact.

TL;DR/At a Glance

• The shift from Copilot as helper to Copilot as doer and why it lands on the leadership agenda 
• Home and Today as the new starting point for work and what that demands from processes and data quality 
• Co-work as task execution across email, calendar, documents, and systems with designed approval points 
• Usage based charging and the move from licence counting to cost per result thinking 
• Copilot Code for building apps in plain language plus the ownership and support risks it creates 

Copilot is no longer just a helpful sidebar that drafts emails and summarises meetings. Microsoft’s latest updates point to something bigger: AI that can carry out multi-step tasks, build lightweight apps from plain English, and soon keep working towards a goal while we are offline. That leap changes the conversation from training and adoption to operating decisions leaders cannot dodge: what work do we hand over, where do we keep human judgement, and who is accountable when an agent gets it wrong?

We walk through the major capabilities in plain terms, starting with Home becoming the single place to begin work, then moving into Co-work where Copilot takes actions across Microsoft 365 with explicit sign-off points. From there we explore Copilot Code, why “anyone can build an app” is both a gift and a governance risk, and what managed runtime means for IT oversight. We also unpack Autopilot, the most ambitious idea of all: persistent AI agents with their own identity, memory, and workspace, plus what independent evidence suggests about where agents are reliable and where they still overclaim.

Behind the headline features, we dig into the parts that make AI delegation real in an organisation: data access (including Power BI and business system connectors), identity and audit through Entra, and the shift to usage-based billing that forces a new metric: value per result. 

The episode closes with a five-question readiness check leadership teams can answer in a single meeting to decide what to try first and what needs fixing before more autonomous tools arrive.

You can also read this article on my website and LinkedIn pages, where I share regular insights on AI, strategy, and emerging technologies..

Support the show

If you are leading your businesses strategic transformation and need greater clarity, stronger execution and measurable results, let’s connect.
 
🌎 Website: www.KieranGilmurray.com
📅 Book a call: https://calendly.com/kierangilmurray/catch-up
📘 Kieran Gilmurray | LinkedIn
🌐 Substack: https://kierangilmurray.substack.com
📕 Amazon https://tinyurl.com/MyBooksOnAmazonUK 

AI Transparency Notice:  This podcast uses a hybrid format. When an episode features one of Kieran Gilmurray’s written articles, the narration is generated using a synthetic clone of his voice via ElevenLabs AI (the underlying article text is entirely human-authored). Episode descriptions and summaries are assisted by AI and should be considered unedited by a human unless specified. 

This episode features an abridged adaptation of Chapter 10, Talent and Skills for the AI Era: Building the AI Workforce of an Intelligent Organization.

Discover why AI transformation is not simply a technology or hiring challenge. It requires leaders to redesign skills, roles, workflows and career pathways so intelligence can become part of everyday work.

The episode explores AI’s rapidly advancing capabilities, its impact on knowledge work and the responsibility leaders have to prepare their organizations for a changing workforce.

Explore The Executive’s Guide to Strategic Intelligence and access the free preview:

https://kierangilmurray.com/strategic-intelligence/

Support the show

If you are leading your businesses strategic transformation and need greater clarity, stronger execution and measurable results, let’s connect.
 
🌎 Website: www.KieranGilmurray.com
📅 Book a call: https://calendly.com/kierangilmurray/catch-up
📘 Kieran Gilmurray | LinkedIn
🌐 Substack: https://kierangilmurray.substack.com
📕 Amazon https://tinyurl.com/MyBooksOnAmazonUK 

AI Transparency Notice:  This podcast uses a hybrid format. When an episode features one of Kieran Gilmurray’s written articles, the narration is generated using a synthetic clone of his voice via ElevenLabs AI (the underlying article text is entirely human-authored). Episode descriptions and summaries are assisted by AI and should be considered unedited by a human unless specified. 

Vibe coding has shifted from experimentation to an operational capability embedded in modern development workflows. Senior leaders now face a practical question of where it can safely accelerate internal tool delivery without increasing risk.

This episode explores how vibe coding can deliver value when treated as a governed system for rapidly building internal tools with clear ownership and measurable outcomes.  It explains how to deploy vibe coding for controlled, high-impact internal use.

TLDR / At a Glance

• Issue to pull request agent workflows
• Low risk internal tooling focus
• Governance through permissions and review gates
• Integration via Model Context Protocol
• Organisational bottlenecks over model limits
• 30 day controlled pilot approach

Vibe coding is moving from curiosity to an operational advantage, but only for leaders who treat it like a governed capability rather than a shortcut. We dig into what shifted in March 2026 and why that month marks a practical turning point: AI coding agents are no longer sitting beside work, they are being embedded into workflow systems and enterprise development tools where tasks can move from an issue to a draft pull request with far less friction.

We also unpack the three developments that change the risk profile and the opportunity at the same time. First, tighter integration with work management makes agent output easier to review in familiar change-control paths. Second, autonomy controls are now explicit: plan modes, permissions, and autonomous execution settings become configuration choices that need policy, not vibes. Third, broader integration layers let agents connect to more tools and data sources, which boosts usefulness but increases the blast radius if access is too broad or ownership is unclear.

From there, we get concrete about safe starting points for AI coding agents and internal tool development: internal reporting and summarisation, workflow wrappers around ticketing and service desks, knowledge navigation across policies and procedures, engineering-adjacent helpers, and security or review support that keeps speed from becoming a bottleneck. We also draw a clear boundary around what to avoid by default, including unsupervised customer-facing outputs and regulated decision contexts that demand stronger governance, accountability, and tested review mechanisms.

Finally, we lay out a controlled 30-day pilot plan and the metrics that actually prove value: time from issue to pull request, time to merge, defect escape rate, rework frequency, portfolio hygiene, and whether every tool has an owner and remains in use after 30 to 60 days. If you found this useful, subscribe, share it with a leader who owns delivery risk, and leave a review with the first internal use case you would pilot.

Support the show

If you are leading your businesses strategic transformation and need greater clarity, stronger execution and measurable results, let’s connect.
 
🌎 Website: www.KieranGilmurray.com
📅 Book a call: https://calendly.com/kierangilmurray/catch-up
📘 Kieran Gilmurray | LinkedIn
🌐 Substack: https://kierangilmurray.substack.com
📕 Amazon https://tinyurl.com/MyBooksOnAmazonUK 

AI Transparency Notice:  This podcast uses a hybrid format. When an episode features one of Kieran Gilmurray’s written articles, the narration is generated using a synthetic clone of his voice via ElevenLabs AI (the underlying article text is entirely human-authored). Episode descriptions and summaries are assisted by AI and should be considered unedited by a human unless specified. 

Responsible intelligence depends on treating governance, ethics and trust as continuous operating disciplines throughout the AI lifecycle.

As AI moves from experimentation into core operations, governance becomes essential to scaling intelligence without scaling hidden risk. Effective oversight creates the accountability, consistency and trust needed for intelligent systems to influence consequential decisions safely.

This episode explores how governance protects Decision Aperture as AI becomes embedded across the enterprise.

TLDR / At a Glance

• Governance as infrastructure for scalable intelligence
• Strong oversight reducing operational friction
• Trust through explainability and accountability
• Fairness controls for historical bias
• Continuous monitoring across the model lifecycle
• Human oversight, ownership and intervention mechanisms 

AI that works in a demo can still fail spectacularly in the real world, not because the model is “bad”, but because the organisation never built the guardrails to scale it. We explore why governance is the foundation of responsible intelligence, especially once automated decision-making starts touching thousands of outcomes across credit, pricing, hiring, clinical triage, logistics, and compliance. When decisions become embedded in workflows, small inconsistencies don’t stay small. They propagate, interact with feedback loops, and can turn performance gains into systemic risk.

We challenge the assumption that governance slows innovation. From our perspective, weak or retrofitted oversight creates the longest delays: pauses, rework, and endless renegotiation after a model is already live. Strong AI governance speeds teams up by making expectations predictable, approvals repeatable, and incidents less chaotic. We then get practical about trust: it isn’t built through messaging, it shows up in how people behave when recommendations are probabilistic, accountability is fuzzy, and the cost of being wrong is personal. That’s where algorithm aversion thrives and where clear escalation paths, explainable reasoning, and fair review processes make adoption possible.

Fairness and bias risk sit at the centre of scalable AI. Historical data can encode exclusion and turn it into automated policy unless you manage disparate impact as rigorously as any other operational risk. 

We also dig into explainability and GDPR realities, then widen the lens to lifecycle governance: monitoring, validation under change, and treating AI as a living system. 

For high-impact decisions, we make the case for human oversight, clear boundaries, and kill switches, backed by strong documentation and named ownership.

If you found this useful, subscribe, share the episode with a colleague, and leave a review. 

What part of AI governance feels most underbuilt where you work?

Support the show

If you are leading your businesses strategic transformation and need greater clarity, stronger execution and measurable results, let’s connect.
 
🌎 Website: www.KieranGilmurray.com
📅 Book a call: https://calendly.com/kierangilmurray/catch-up
📘 Kieran Gilmurray | LinkedIn
🌐 Substack: https://kierangilmurray.substack.com
📕 Amazon https://tinyurl.com/MyBooksOnAmazonUK 

AI Transparency Notice:  This podcast uses a hybrid format. When an episode features one of Kieran Gilmurray’s written articles, the narration is generated using a synthetic clone of his voice via ElevenLabs AI (the underlying article text is entirely human-authored). Episode descriptions and summaries are assisted by AI and should be considered unedited by a human unless specified. 

AI is scaling fast, but governance is struggling to keep pace. As AI systems move into core decisions, the risks grow faster than the controls designed to manage them.

This episode explores how organisations can turn governance into a strategic advantage.

TLDR / At a Glance

• AI risk scaling faster than oversight
 • Governance as executive responsibility
 • Regulatory pressure across EU, UK, US
 • Converging principles across frameworks
 • Audit logs, oversight, escalation controls
 • Governance enabling trust and scale

Strong governance transforms AI from a source of uncertainty into a system that can scale with confidence, accountability, and control.

Support the show

If you are leading your businesses strategic transformation and need greater clarity, stronger execution and measurable results, let’s connect.
 
🌎 Website: www.KieranGilmurray.com
📅 Book a call: https://calendly.com/kierangilmurray/catch-up
📘 Kieran Gilmurray | LinkedIn
🌐 Substack: https://kierangilmurray.substack.com
📕 Amazon https://tinyurl.com/MyBooksOnAmazonUK 

AI Transparency Notice:  This podcast uses a hybrid format. When an episode features one of Kieran Gilmurray’s written articles, the narration is generated using a synthetic clone of his voice via ElevenLabs AI (the underlying article text is entirely human-authored). Episode descriptions and summaries are assisted by AI and should be considered unedited by a human unless specified. 

AI is turning up in every legal team’s inbox, but senior leaders are still stuck on the same dilemma: what’s genuinely usable in law right now, and what’s risky theatre that could blow up on confidentiality, accuracy, or ethics? 

We sit down with Colin Levy, general counsel at legal tech company Malbek and an adjunct law professor, to get a grounded view of legal AI that goes beyond hype. 

His headline is simple and sharp: AI is a powerful but imperfect tool, and lawyers need to work with that reality rather than pretend it will disappear.

TL;DR / At A Glance

• why AI in law is “powerful but imperfect” and why that matters for risk
• what hallucinations look like in legal work and how to reduce them with context and grounded sources
• where AI is genuinely useful today such as first drafts, review, summarisation and knowledge creation
• why the real win is strategic productivity rather than raw efficiency
• how AI pressures the billable hour and what “value” can look like instead
• how to think about junior lawyer hiring and onboarding when AI can draft

We dig into the problem everyone whispers about and some people ignore: hallucinations. Colin explains why fabricated answers are a feature of how generative AI works, then walks through practical ways to reduce the risk, from giving better context and instructions to using tools trained on legal data for legal research. 

We also talk about what “collaboration” with AI looks like in real workflows: first drafts, document review, summarising large datasets, and creating usable knowledge, with humans still responsible for checking the work.

From there, we get concrete on leadership decisions: how to evaluate AI vendors without becoming a technical expert. We cover data handling and data sharing, security credentials such as SOC 2, privacy expectations including GDPR, and what to ask about model updates, transparency, and liability. 

We also unpack AI governance that actually works, training and permissioning, a sensible 90-day plan, and why measuring success by logins is meaningless. Finally, we tackle the future of junior lawyers, onboarding in the era of AI, and where practice areas like litigation analytics and contract work may shift next.

If you want a clear, practical approach to legal technology and AI governance that protects clients while improving output quality, press play, then subscribe, share with a colleague, and leave us a review. 

What’s the one AI rule your team refuses to compromise on?

Support the show

If you are leading your businesses strategic transformation and need greater clarity, stronger execution and measurable results, let’s connect.
 
🌎 Website: www.KieranGilmurray.com
📅 Book a call: https://calendly.com/kierangilmurray/catch-up
📘 Kieran Gilmurray | LinkedIn
🌐 Substack: https://kierangilmurray.substack.com
📕 Amazon https://tinyurl.com/MyBooksOnAmazonUK 

AI Transparency Notice:  This podcast uses a hybrid format. When an episode features one of Kieran Gilmurray’s written articles, the narration is generated using a synthetic clone of his voice via ElevenLabs AI (the underlying article text is entirely human-authored). Episode descriptions and summaries are assisted by AI and should be considered unedited by a human unless specified. 

AI capability only creates value when people have the confidence, authority, and incentives to act on intelligent signals. This episode examines the organizational conditions that determine whether insight becomes action or stalls inside existing decision structures.

It explores how culture, governance, workforce design, and accountability shape intelligent organizations.

TLDR / At a Glance

• Creators, translators, and operators
• Psychological safety under uncertainty
• Clear authority for human and model decisions
• Reskilling focused on judgment
• Incentives aligned with decision quality
• Decision-centric talent and expertise models 

The key takeaway is that effective AI adoption depends on designing human systems that support informed judgment, timely action, and continuous learning.

Support the show

If you are leading your businesses strategic transformation and need greater clarity, stronger execution and measurable results, let’s connect.
 
🌎 Website: www.KieranGilmurray.com
📅 Book a call: https://calendly.com/kierangilmurray/catch-up
📘 Kieran Gilmurray | LinkedIn
🌐 Substack: https://kierangilmurray.substack.com
📕 Amazon https://tinyurl.com/MyBooksOnAmazonUK 

AI Transparency Notice:  This podcast uses a hybrid format. When an episode features one of Kieran Gilmurray’s written articles, the narration is generated using a synthetic clone of his voice via ElevenLabs AI (the underlying article text is entirely human-authored). Episode descriptions and summaries are assisted by AI and should be considered unedited by a human unless specified. 

An intelligent enterprise depends on more than advanced analytics, models, and data platforms. Value emerges when signals, decisions, authority, and actions operate coherently across organizational boundaries.

This episode explores how integration turns local intelligence into coordinated enterprise capability.

TLDR / At a Glance

• System intelligence over isolated capabilities
• Connected signals, decisions, and actions
• Customer journeys as tests of coherence
• Shared authority and accountability
• Incentives aligned to enterprise outcomes
• Integration across data, interpretation, and execution

Smart models can still produce a stupid organisation. When signals sit in one function, interpretation happens in another, and action lands somewhere else entirely, the strategic intelligence loop slows at every handoff. We explore what it really takes to become an intelligent enterprise where intelligence operates as a system, not a set of isolated analytics wins.

We walk through concrete examples from banking, industrial operations, telecommunications, insurance, and supply chain. You’ll hear why banks can simultaneously treat the same customer as high risk, high value, and marginal, and how predictive maintenance can improve plant uptime while the wider network still misses delivery commitments. The thread running through every case is coherence: shared signals are not enough if the logic, authority, and accountability to act remain fragmented.

We also flip the lens to the customer experience, because customers become the first systems integrator. They feel the seams in broken journeys, missing context, and inconsistent decisions long before executives see the pattern in dashboards. From there, we get practical about integration: why platform consolidation and enterprise data programmes plateau without operating model change, and how control towers only deliver resilience once cross-functional decision rights are real.

If you care about AI strategy, decision intelligence, governance, and building a joined up organisation that can learn and adapt, this is for you. Subscribe, share with a colleague who owns “the handoffs”, and leave a review with the one integration barrier you want to fix next.

Enterprise intelligence scales when technology, governance, decision rights, and incentives work together as one connected system.

Support the show

If you are leading your businesses strategic transformation and need greater clarity, stronger execution and measurable results, let’s connect.
 
🌎 Website: www.KieranGilmurray.com
📅 Book a call: https://calendly.com/kierangilmurray/catch-up
📘 Kieran Gilmurray | LinkedIn
🌐 Substack: https://kierangilmurray.substack.com
📕 Amazon https://tinyurl.com/MyBooksOnAmazonUK 

AI Transparency Notice:  This podcast uses a hybrid format. When an episode features one of Kieran Gilmurray’s written articles, the narration is generated using a synthetic clone of his voice via ElevenLabs AI (the underlying article text is entirely human-authored). Episode descriptions and summaries are assisted by AI and should be considered unedited by a human unless specified. 

56% of chief executives say AI has brought them neither higher revenue nor lower cost over the past year. That stat sounds like a model problem, but we think it is usually an operating model problem: the organisation around the technology cannot reliably turn a capable output into a business outcome.

We break down why “better models” so often fail to change the numbers. A model can draft, predict, classify, and recommend, but it cannot fix inaccessible data, broken workflows, serial approvals, manual re-entry, or unclear decision rights. We dig into the most common value leaks that quietly kill enterprise AI ROI: data that cannot be retrieved in the right form at the right moment, processes that stay unchanged around a faster step, and recommendations that nobody is authorised to act on. We also cover adoption, incentives, governance, and measurement, because without clear metrics and accountability, “time saved” turns into anecdotes instead of profit.

A key tool we use is the adequacy threshold: define what “good enough” means for a specific use case, choose the cheapest model that clears that line, then invest management effort where the real constraint sits. We make it concrete with examples from insurance claims, procurement cycles, and professional services pricing, where AI can speed up document work while value still fails to reach the P&L.

We finish with the agent trap. As AI agents move from answering to acting, the ceiling rises, but so do the requirements for permissions, monitoring, cost control, exception handling, and governance. If you want AI value that lasts, the surrounding organisation has to be trusted to let a machine act. 

Subscribe, share this with a leader who owns outcomes, and leave a review: what is the biggest blocker to AI ROI where you work?

Support the show

If you are leading your businesses strategic transformation and need greater clarity, stronger execution and measurable results, let’s connect.
 
🌎 Website: www.KieranGilmurray.com
📅 Book a call: https://calendly.com/kierangilmurray/catch-up
📘 Kieran Gilmurray | LinkedIn
🌐 Substack: https://kierangilmurray.substack.com
📕 Amazon https://tinyurl.com/MyBooksOnAmazonUK 

AI Transparency Notice:  This podcast uses a hybrid format. When an episode features one of Kieran Gilmurray’s written articles, the narration is generated using a synthetic clone of his voice via ElevenLabs AI (the underlying article text is entirely human-authored). Episode descriptions and summaries are assisted by AI and should be considered unedited by a human unless specified. 

The fastest way to slow an organisation down is to keep “fixing” complexity by adding more. More meetings. More KPIs. More tools. More approvals. More governance. It feels decisive and responsible, but over time the accumulation creates execution drag: decisions take longer, ownership gets murky, and teams spend more energy coordinating than solving real problems.

TLDR / At a Glance

• Addition as a leadership signal
• Cognitive bias against subtraction
• Organisational systems that reward launches
• Busywork, meetings, and execution drag
• Strategic subtraction across meetings and governance
• Protecting capability while removing waste

We lay out the case for strategic subtraction: the leadership discipline of removing work, process and activity that no longer creates enough value. Along the way, we unpack why subtractive solutions are so easy to miss, especially under pressure, and why most organisations are built to launch things but not to stop them. We also borrow from Peter Drucker’s planned abandonment to ask a blunt question that should sit at the heart of strategy: knowing what we know now, would we start this again today?

From there, we connect the dots to measurable realities like busywork and coordination overhead, decision latency driven by unclear decision rights, and the way constant fragmentation undermines deep work, innovation and wellbeing. We then bring it to life with practical examples of subtraction at different levels, from cutting recurring meetings to reducing bureaucracy and even using product constraint as a strategic advantage.

If you’re trying to lead in the AI era, this is a sharper, kinder way to create focus without burning people out. Subscribe, share the episode with a colleague, and leave a review, then tell us what you would subtract first.

The key takeaway is that stronger leadership often means removing work that no longer creates enough value.

Support the show

If you are leading your businesses strategic transformation and need greater clarity, stronger execution and measurable results, let’s connect.
 
🌎 Website: www.KieranGilmurray.com
📅 Book a call: https://calendly.com/kierangilmurray/catch-up
📘 Kieran Gilmurray | LinkedIn
🌐 Substack: https://kierangilmurray.substack.com
📕 Amazon https://tinyurl.com/MyBooksOnAmazonUK 

AI Transparency Notice:  This podcast uses a hybrid format. When an episode features one of Kieran Gilmurray’s written articles, the narration is generated using a synthetic clone of his voice via ElevenLabs AI (the underlying article text is entirely human-authored). Episode descriptions and summaries are assisted by AI and should be considered unedited by a human unless specified.