Kieran Gilmurray

Cost cutting starts with budgets, expense lines and headcount targets.

• Subtraction starts with the work itself, asking which activities consume more coordination, time, risk or complexity than they return in value.
• A business can remove people while leaving a broken operating process intact, or cut spending while preserving duplicated reporting and unclear decision rights — transferring pressure into the system rather than removing it.
• They identify six ways to subtract: elimination, substitution, consolidation, hiding complexity behind a simpler interface, pausing without killing, and abstracting complex back-end operations.
• Each mode carries its own risk: consolidation concentrates risk, hiding can reduce transparency, and pausing can become avoidance.

IKEA's decision to discontinue its printed catalogue is Gilmurray's example of elimination: the catalogue was iconic, but customer behaviour had shifted, so ending print cut production and distribution burdens and freed resources for digital activity. Zipline's medical drone delivery in Rwanda is their example of substitution, replacing slower road delivery of urgent blood products with a faster route that reduced wastage.

Read original

Efficiency asks whether a change reduces unnecessary time, cost, handoffs or duplication.

• Resilience asks whether it protects the organisation's ability to adapt, recover and avoid brittleness.
• Prominence asks whether it improves visibility, trust, customer experience or strategic distinctiveness.
• Their core argument is that efficiency measured alone is a false economy: leaders must distinguish waste from slack.
• Waste absorbs resources without purpose; slack provides adaptability, quality assurance or safety.
• They set this tension against conventional cost-cutting logic, which they say is attractive precisely because it is easy to measure.

They point to thin, just-in-time supply chains that looked efficient until Covid-era disruption exposed how little buffer they had, and to the 2024 amended enforcement action against Citibank over data quality management and regulatory reporting — overhead that looks removable until it fails. They also cite third-party findings in support of the efficiency argument: Asana on time spent on "work about work", Atlassian on executives estimating that only a quarter of team effort is mission critical, and Microsoft research on the fragmented working day.

Read original

When the criteria for sufficiency are clear, approval cycles shorten; when governance is improvised, every deployment becomes a negotiation and every incident requires a bespoke response.

• They describe trust as an operating property, not a communications strategy: it depends on whether decisions can be explained, questioned and traced to a named owner.
• Fairness is an operational risk, not merely a moral aspiration: models learn from historical data, so scaling without fairness controls automates past exclusion.
• Explainability, stress testing, documentation and the ability to halt a system are control categories that belong in the decision record.

They cite research showing that people often prefer human judgment to more accurate algorithmic advice when they cannot see the basis of a recommendation, GDPR safeguards against solely automated decisions with significant effects, Basel Committee requirements for ongoing validation and stress testing, and the case for kill switches and human override in high-impact decisions.

Read original

Functions are rewarded for different outcomes, tolerate different levels of risk and act on different time horizons, so signals are filtered and decisions negotiated at every handoff.

• They call the customer the organisation's "first systems integrator", because customer journeys expose the seams before executive dashboards do.
• Their conclusion: shared information without shared accountability produces more debate, not faster decisions.
• Integrating data is not enough unless authority, incentives and escalation paths are aligned too.

Large banks built fraud, credit and marketing models in parallel, so the same customer could be treated as high risk by fraud and high value by marketing — cards blocked while retention offers went out. Supply chain control towers, they note, delivered end-to-end visibility but little improvement until cross-functional teams gained real authority to reprioritise inventory and reroute shipments.

Read original

Progress has always depended on more than the quality of an idea: people need knowledge, skills, funding and support to turn it into something useful.

• In their account, AI narrows the distance between spotting a problem and testing a response.
• The scarce resources then become the insight to spot a worthwhile problem, the judgement to shape a response and the willingness to act.
• Leaders must create conditions where people can experiment.
• They are clear about the limits: AI will not equalise talent or guarantee that good ideas succeed, and scaling still typically needs funding and organisational support.

A social worker who understands a human problem but lacks the skills to build a solution, or a teacher who recognises a learning difficulty but has no practical way to create a support tool, can now build prototypes, analyse patterns and make the case to decision makers. They apply the same idea to a farmer interpreting soil, weather and crop data, and a small business owner testing scenarios that once required specialist consultants.

This is the shortest and least framework-driven of the reviewed pieces: a LinkedIn essay built from short paragraphs and rhetorical questions, unlike the chapter adaptations with their "at a glance" summaries and named research. The material is consistently open about AI's limits.

Read original

It is the clearest standalone starting point: a complete argument with a named test, worked examples and practical closing questions. Newcomers get a concise introduction to Kieran Gilmurray's thinking on executive trade-offs.

The article distinguishes ordinary cost cutting, which starts with budgets and expense lines, from strategic subtraction, which starts with the work itself and asks what should no longer exist because it slows execution or absorbs scarce attention.

• It offers a three-part test — efficiency, resilience and prominence — and warns that efficiency alone can be a false economy.
• Thin Covid-era supply chains and the 2024 OCC enforcement action against Citibank show how overhead can look removable until it fails.
• The article names six ways to subtract: elimination (IKEA discontinuing its printed catalogue), substitution (Zipline delivering blood products by drone in Rwanda), consolidation (Estonia's X-Road once-only data principle), hiding complexity, pausing (Netflix's pause membership option) and abstraction (AWS infrastructure services).
• It cites Asana on time spent on "work about work", Atlassian on executives estimating only 24 per cent of team effort as mission critical, and Microsoft on the "infinite workday".
• It draws a firm line between waste and slack, then closes with practical questions for leaders: which controls protect real risk, which duplicate others, and which work creates more coordination burden than value.

For brands planning a campaign

The strongest single source for the requested thought-leadership piece: the triple test gives a consulting firm a ready editorial spine, while the six subtraction types give executive clients a vocabulary for their own operating reviews.

It is the systems argument behind the subtraction and governance material, and a useful second read for anyone seeking to understand why the author keeps returning to organisational design rather than tools.

The chapter argues that intelligence creates enterprise value only when it works as a connected system, rather than as isolated capabilities within functions.

• Its clearest example is large banks that built strong fraud, credit and marketing models in parallel: the same customer could be treated as high risk by fraud and high value by marketing, so cards were blocked while retention offers went out.
• In industrial operations, predictive maintenance models confined to plant teams optimise uptime locally while delivery commitments slip globally.
• Supply chain control towers delivered end-to-end visibility but little improvement until cross-functional teams gained authority to reprioritise inventory and reroute shipments.
• Aviation and freight recovery times fell only after decision authority and escalation paths were consolidated.
• The recurring claim is that integration is an organisational problem before it is a technical one: shared information without shared accountability produces debate rather than faster decisions.

For brands planning a campaign

Useful for clients running parallel analytics teams or platform programmes: the bank and control-tower examples lead directly to a diagnostic about decision rights, not just data plumbing.

It is the best starting point for a risk, compliance or audit audience because it frames governance as what lets intelligence scale, rather than what slows it down.

The chapter argues that governance allows intelligence to scale safely, because the same systems that amplify performance also amplify errors, bias and systemic risk.

• Its counter-intuitive claim is that weak or retrofitted governance slows innovation more than strong oversight: it forces rework and repeated negotiation after systems are embedded, while standardised expectations shorten approval cycles.
• It treats trust as an operating property rather than a communication strategy, citing algorithm aversion — people preferring human judgment to more accurate algorithmic advice when they cannot see the basis of a recommendation.
• On fairness, it argues that models inherit historical patterns and that scaling without fairness controls effectively automates past exclusion.
• On explainability, it points to GDPR safeguards against solely automated decisions with significant effects.
• The chapter also covers stress testing under the Basel Committee's requirements, kill switches and human override in high-impact decisions, and documentation and ownership across the model lifecycle.

For brands planning a campaign

A natural pair with the integration chapter, these pieces support a governance-and-operating-model proposal for regulated clients, where explainability, ownership and intervention rights are key buying concerns.

It is the shortest and most accessible entry point, presented as a short argument rather than a chapter, and works well as a first read before the longer Substack pieces.

The post sets out three shifts.

• First, personalisation: the printing press gave everyone the same book and the internet gave everyone the same pool of information, but AI can adapt to an individual's understanding, circumstances and goals.
• Second, wider, cheaper access to expertise, illustrated by a farmer interpreting soil, weather and crop data and a small business owner analysing markets that once required specialist consultants.
• Third, and in the author's words most important, is the democratisation of execution: a social worker who understands a human problem but cannot build a solution, or a teacher who recognises a learning difficulty but cannot create a support tool, can now prototype, analyse patterns and communicate ideas to decision makers.
• The post acknowledges that AI will not make talent equal or remove every barrier.
• The scarce resource may instead be the insight to recognise a worthwhile problem, the judgement to shape a response and the willingness to act — with leaders responsible for creating conditions where people can experiment.

For brands planning a campaign

A strong human-interest opener for leadership audiences before the more technical chapters, and a useful theme for clients whose AI messaging has become feature-led rather than people-led.

The test is whether activities consume more coordination, time, risk or complexity than they return in value.

• They argue that cutting a team while leaving broken workflows, duplicated reporting and unclear decision rights intact simply transfers pressure into the system.
• The article names six ways to subtract — elimination, substitution, consolidation, hiding, pausing and abstraction — and illustrates each with a reference case, including IKEA dropping its printed catalogue, Zipline’s medical drone delivery in Rwanda, Estonia’s X-Road once-only data principle, Netflix’s pause membership option and AWS infrastructure abstraction.
• They also warn that each mode carries its own risk, such as consolidation creating concentration risk or hiding reducing transparency.

Each mode is briefly defined and anchored to a named case, making the framework easy to reuse in a workshop or client briefing.

For brands planning a campaign

This is the clearest commission opportunity in the posts: Gilmurray has published it as the second in a three-part series. A brand can slot into the existing arc or commission the closing piece on making subtraction a repeatable leadership discipline.

Examples: Source 1

Efficiency asks whether it reduces unnecessary time, cost, handoffs or duplication.

• Resilience asks whether it protects the organisation’s ability to adapt, recover and avoid brittleness.
• Prominence asks whether it improves visibility, trust, customer experience or strategic distinctiveness.
• The core warning is that efficiency alone is a false economy: a supply chain with no buffer looks efficient until disruption arrives, and a control process looks slow until a regulatory failure shows why it existed.
• They support the efficiency argument with Asana’s finding that knowledge workers spend around 60 per cent of their time on “work about work,” Atlassian’s estimate that executives see only 24 per cent of team effort as mission critical, and Microsoft research on the fragmented “infinite workday.”
• They ground the resilience argument in Covid-era supply chain disruption and the OCC’s 2024 amended enforcement action against Citibank over data quality and regulatory reporting.

The three tests form a decision checklist, followed by questions leaders can ask, such as which controls duplicate others and what should be paused rather than killed.

For brands planning a campaign

A brand selling operating-model, resilience or risk tooling can use this frame for a co-authored piece or executive webinar. It gives clients a named way to evaluate trade-offs, rather than another cost-cutting argument.

Examples: Source 1

The recurring framework is a Strategic Intelligence Loop spanning sensing, interpretation, decision, execution and learning, with Decision Aperture showing how leaders connect intelligence to judgement and judgement to action.

• Gilmurray argues that the loop compounds only when it runs end to end; when sensing happens in one function, interpretation in another and action in a third, handoffs slow the process, filter signals and turn decisions into negotiations.
• They position the book as practical rather than theoretical, with frameworks that connect judgement to measurable organisational performance.

Each chapter adaptation opens with a short “at a glance” summary, develops the argument through named third-party research and case examples, then closes with cross-links to other chapters in the series and subscription, article-library and podcast prompts.

For brands planning a campaign

The book gives a campaign a ready-made narrative anchor and serialised structure. Brands seeking a chapter-adjacent piece can ask Gilmurray to cover a specific stage of the loop they serve — sensing, decisioning or learning — rather than AI in general.

Examples: Source 1 · Source 2 · Source 3

The same systems that amplify performance also amplify errors, bias and systemic risk.

• Their counterintuitive claim is that weak or retrofitted governance slows innovation more than strong oversight, forcing organisations to pause, rework and renegotiate foundations after systems are embedded.
• Trust, in this framing, is an operating property rather than a communications strategy: it depends on whether decisions are explainable and challengeable and whether accountability is clear.
• They cite algorithm aversion research, GDPR safeguards against solely automated decisions with significant effects, Basel Committee requirements for stress testing and ongoing validation, and the case for kill switches and human override in high-impact decisions.
• They also treat fairness as an operational risk: models inherit bias from historical data, and scaling without fairness controls effectively industrialises yesterday’s prejudices.

Gilmurray frames each risk structurally rather than morally and pairs it with a control expectation: documentation, ownership, lifecycle accountability, monitoring and defined authority for intervention.

The piece covers principles and control categories, not specific vendors, tools or implementation approaches. Brands seeking product-level comparison should supply those details.

For brands planning a campaign

This is Gilmurray's strongest territory for regulated sectors such as banking, insurance and healthcare. A brand selling monitoring, model-risk or assurance tooling can commission a governance-adjacent piece that builds on the established argument and adds the practical control layer.

Examples: Source 1

Gilmurray illustrates the point with large banks that built sophisticated fraud, credit and marketing models in parallel, so the same customer could be treated as high risk by one function, high value by another and marginal by a third — cards blocked while retention offers were issued.

• They extend the pattern to industrial predictive maintenance that optimises plant uptime while missing network-level delivery commitments.
• The limiting factor, they argue, is organisational rather than technical: functions are rewarded for different outcomes, operate under different risk tolerances and hold different time horizons.
• Gilmurray calls the customer the organisation’s “first systems integrator,” because customer journeys expose seams before executive dashboards do.
• The argument draws on McKinsey work on end-to-end decisioning, Bain research on omnichannel transitions, BCG on churn and on supply chain control towers, Deloitte on claims transformation, Accenture on network orchestration, and HBR observations on ERP and core-system modernisation.

Gilmurray builds the case through paired industry examples, then draws a single conclusion: integration requires aligning authority, accountability and incentives, not just connecting data.

For brands planning a campaign

This topic suits consultancies and platform vendors working on operating models, decision rights and cross-functional data programmes. A brand can commission a piece on one vertical — banking, telco, insurance or logistics — using Gilmurray's argument as the frame and supplying sector-specific detail.

Examples: Source 1

Gilmurray sets out three shifts enabled by generative AI — personalisation, wider and cheaper access to expertise, and the democratisation of execution.

• They argue that progress has always depended not only on the quality of ideas but on access to knowledge, skills, funding and support, and that AI narrows the distance between having an idea and testing whether it could work.
• The examples are concrete: a farmer interpreting soil, weather and crop data; a small business owner testing scenarios that once required consultants; a social worker who understands a human problem but lacks the skills to build a solution; and a teacher who recognises a learning difficulty but has no practical way to create a support tool.
• Gilmurray closes by arguing that technical capability may no longer be the scarce resource; insight, judgement and the willingness to act matter more — but only if leaders create conditions for people to experiment.

Short paragraphs, rhetorical questions and a closing question to readers, with a hashtag-style call around unlocking talent.

Gilmurray cautions that AI will not equalise talent, remove every barrier or guarantee success; people may still need funding, expertise and organisational support to scale. This is an opportunity thesis, not a how-to. Brands seeking implementation steps should provide them.

For brands planning a campaign

This is the most LinkedIn-native piece in the posts and shows Gilmurray can make an opportunity case to a broad professional audience. Brands in education, workforce upskilling or accessible AI tooling can commission a companion piece on the leadership conditions that turn access into execution.

Examples: Source 1

The article links to the book on Amazon and to a landing page offering a free two-chapter preview, and a LinkedIn post refers to "my three books". Ask who published the book; the posts don't say whether it was self-published or issued through a publisher.

View example

That echoes Gilmurray's argument that integration is organisational before it is technical.

McSheehy's evidence comes from humanitarian and emergency-response environments and frontline change; Gilmurray writes from an executive strategy and AI-governance perspective, drawing on book-derived frameworks and named consulting research.

Read original · Original 2

That overlaps with Gilmurray's strategic-subtraction argument and the protect, improve, combine or retire discipline.

Kieran Gilmurray focuses on the learning-and-development function and leadership development, with practitioner questions about courses, resources and planning. Their frame is broader executive strategy, covering cost, resilience, governance and integration.

Read original · Original 2

Both treat resilience as a reason to simplify, not just to buy more.

Haagman's examples and audience centre on cybersecurity and CISO practice, including a zero-budget test and an academic research programme; Gilmurray writes about AI strategy, governance and operating models beyond security.

Read original · Original 2

Her two posts start from that situation.

• One describes transport booking systems redesigned for an emergency response that disrupted longer-term programmes, and local procurement or staffing arrangements that solved one problem while creating governance or equity problems elsewhere.
• The other describes overlapping teams in a large emergency response, where duplication blurred accountability, and sets out what clarity requires: purpose, interfaces, decision rights and collective accountability.
• A consulting firm selling to public bodies, humanitarian organisations or matrix-structured businesses would get a more credible draft from her than from a general strategy writer because she starts with the operating mess, not the framework.

She has not published a codified subtraction method.

• The material is two short LinkedIn posts, with no three-part test, list of subtraction modes or long-form article to point clients to.
• A firm seeking a named, citable framework to license or build a workshop around would need to supply the structure and use McSheehy for the lived examples.
• Her frame is about aligning local change with wider direction, not deciding what to remove.
• If the buyer wants to know what should no longer exist, Kieran Gilmurray’s published article answers directly; McSheehy addresses the adjacent question of how to stop local fixes from fragmenting the whole.

Examples: Source 1 · Source 2

Her October post is a direct fit: it names the accumulation problem, describes the tail every new resource carries (communicate, maintain, update, support, evaluate, retire), and offers four decisions a team can apply: protect, improve, combine and retire.

• She reframes the annual planning question as what the team has earned the right to stop doing, rather than what it should add next year.
• A consulting firm with an L&D practice, or one pitching a workshop to a chief learning officer, would find her more immediately useful than a general strategy voice because the examples and decisions are already in the buyer’s language.
• Her June post adds another angle: AI may make leadership development more important by exposing what is not leadership, such as polished decks, fast summaries and confident decisions, and shifting value to judgement, listening and developing people.
• That’s a strong thought-leadership hook for an HR audience.

Her scope is L&D, not the enterprise operating model.

• Her posts do not address duplicated controls, approval layers, governance or supply-chain buffer, so a client asking how to subtract across a whole business would get a narrower answer.
• The material comprises two short LinkedIn posts, with no long-form article, third-party research base or named framework beyond the four decisions in the October post.
• A firm seeking a substantial authored piece with citations and cases would need to build one around her, or choose Gilmurray, whose subtraction article already includes named research and company examples.
• If the buyer is a chief operating officer or strategy director rather than a people-function leader, that tradeoff is likely to outweigh the fit.

Examples: Source 1 · Source 2

His August post describes a CISO Singapore fireside built around a Zero-Budget Test: if the cyber budget became zero tomorrow, what would you still do first?

• He turns that into questions about which temporary exceptions have become permanent architecture, where detection and response compensate for complexity or unnecessary access, and what should no longer be true.
• He argues that resilience can come from subtraction: remove unnecessary privilege, eliminate dependencies, challenge exceptions and simplify operating models.
• That’s a sharper security-specific version of the subtraction argument than a general strategist would produce.
• His September post adds research credibility: he describes an invitation to become a Research Fellow at the University of Warwick and work expanding the CISO Global Study into a wider global research programme.
• For a consulting firm selling to CISOs or a cyber-risk practice, he would bring the piece a credible field voice and an existing community connection.

The material is cybersecurity-specific.

• It does not offer a general framework for subtracting work, controls or complexity across a whole enterprise, and it does not address the cost-led versus value-led distinction that a general management audience needs.
• The posts are also two short posts: the fireside format, the zero-budget questions and the research-fellow announcement, but no long-form article setting out a repeatable method.
• A firm would have to supply the structure and the non-security examples.
• Gilmurray's subtraction article already does that work across functions; Haagman is the better choice only when the buyer's decision is specifically about security architecture, privilege, dependencies or cyber resilience, and the firm is willing to build the thought-leadership asset around a shorter, more specialist base.

Examples: Source 1 · Source 2