Richard Foster-Fletcher
The Writing
This is where I set out my commentary on what artificial intelligence is doing inside organisations. Each piece interprets my research, tests a developing finding, or follows a question the evidence has raised.
The work rests on controlled experiments, corporate disclosures, contracts, court filings, observed model behaviour, and documented cases, rather than on prediction. Read it in order, move through it by area of interest, or put a direct question to it below.
Alongside this sit the Research and the Structural Dynamics.
Ask the archive
For example: who carries the blame for AI decisions? · what do companies actually report about AI? · humanoid robots in the home · loss of skill
51 articles, newest first
2026
Jul 2026
The End of the AI Premium
A case study in the economics of saying 'AI' in public
It reads the frequency of the word 'AI' in a company's disclosures as a priced signal that inflates and deflates with the scarcity of the claim rather than tracking engineering progress.
Evidence & DisclosureCorporate disclosure & reportingMarkets, moats & competitionProductivity & value measurement
Jul 2026
Discoverable Doubt
Why Leaders Won't Think Aloud with AI
Adds the problem of discoverability to the debate on executive AI use, showing why the most senior and highest-value application of AI is the one leaders rationally avoid.
Executive Judgement & AccessExecutive access & decision recordsAccountability & liability
Jun 2026
Who decides what the CEO can ask AI?
A study of visibility, control and AI capability at the top of the organisation
Provides an empirical test showing that approved enterprise tools refuse exactly the crisis moments executives most need, and that a local model offers an unused alternative.
Executive Judgement & AccessRefusal & suppressed answersExecutive access & decision recordsSystem prompts & instructions
Jun 2026
Blamefall
When mandated AI leaves the signatory carrying the fall
Names blamefall and isolates the split inside weak output, extending existing accounts of automation blame that stop at cases where the tool was actively wrong.
Responsibility & LiabilityAccountability & liabilityAdoption mandates & strategic non-useGovernance & oversight
Jun 2026
The AI Predictability Trap
What if mandatory AI adoption made your company predictable and easier to outmanoeuvre?
It identifies a strategic downside of mandated AI adoption, that it makes a company's pattern-based decisions predictable to rivals, and argues for deliberate, selective use and intentional friction.
Vendor & Market DynamicsAdoption mandates & strategic non-use
Jun 2026
The Subtraction Study
What corporate rules do to AI answers
Supplies controlled evidence that corporate compliance prompts, not model ability, strip decisiveness from AI answers, quantifying a substitution firms rarely notice.
Model Behaviour & ReliabilitySystem prompts & instructionsRefusal & suppressed answersGovernance & oversight
Jun 2026
700 AI Statements, 2 Measured Outcomes
A Review of Corporate Disclosure
Turns the volume of corporate AI talk into a measured coding, showing that disclosure records deployment and intent but almost never a measured outcome.
Evidence & DisclosureCorporate disclosure & reporting
May 2026
Recursive self-improvement
are the new billion-dollar AI labs building something beyond the capacity of businesses?
Connects frontier-lab ambition to organisational readiness, showing that firms lack both the reporting and the governance to use autonomous machine reasoning.
Memory, Continuity & Frontier AIFrontier AI & self-improvementProductivity & value measurement
May 2026
Has the Board Read the System Prompt?
The hidden instruction shaping what enterprise AI will answer, refuse, soften and omit
Identifies the system prompt as an ungoverned, board-invisible document and argues it belongs inside director oversight.
Model Behaviour & ReliabilitySystem prompts & instructionsGovernance & oversightRefusal & suppressed answers
May 2026
The Groupon Experiment
What Happens When AI Becomes a CEO's Thinking Partner?
Uses Groupon as a live case for how embedding AI in the revenue core can concentrate commercial judgement in one executive, and shows what disclosure cannot capture.
Executive Judgement & AccessJudgement, strategy & option-spaceAutonomous agents & agentic systemsCorporate disclosure & reporting
May 2026
The Reverse Turing Test
What Gets Lost When AI Makes Words Cleaner
Reframes AI editing as information loss, naming the reverse Turing test and showing which organisational communication depends on the signal editing removes.
AI-Shaped Language
Apr 2026
This Looks LLM-Shaped
When 150 SEC filings shift in ways researchers associate with AI in writing
It supplies the inferential bridge for Foster-Fletcher's SEC-filings work, showing how controlled research on feedback-tuned models licenses a corpus-level claim of AI influence without any disclosure.
Evidence & DisclosureAI-shaped language & homogenisationAccountability & liabilityCorporate disclosure & reporting
Apr 2026
Banning Chinese Humanoid Robots
The export market may not work the way Washington thinks.
It reframes humanoid export policy from a security-threat story into a supply-allocation problem driven by the producer countries' own demographic and fiscal pressures.
Robotics & EmbodimentChina & geopoliticsHumanoid robots & embodiment
Mar 2026
The First Annual Reports of The LLM Era
My new study examining the impact of generative AI on 150 SEC filings
Documents a measurable shift in the character of SEC filing language coinciding with enterprise LLMs, while declining to claim causation.
AI-Shaped LanguageAI-shaped language & homogenisationCorporate disclosure & reporting
Mar 2026
The Undesigned Company
The Artificial Evolution of the Modern Firm
It names a structural pattern in which a firm's reasoning is shaped by retention rather than intent, leaving institutional memory emergent and untraceable rather than authored.
AI-Shaped LanguageAuthorship, framing & organisational reasoningMachine memory & forgetting
Mar 2026
The Governance Contradiction
Was AI Oversight just declared impractical?
It generalises IMDA's admission about agents into a condition of every AI approval process at scale, and points from better oversight toward constraining what AI does to the organisation it enters.
Responsibility & LiabilityGovernance & oversightAccountability & liability
Mar 2026
The Awareness Trap
The bias the regulation cannot fix
It shows that Article 14's human-oversight requirement is an aspirational performance standard the psychological evidence says cannot be reliably met, distinguishing population-level correction from protection of the individual decision.
Responsibility & LiabilityRegulation & standardsGovernance & oversight
Mar 2026
The AI Reconstruction Illusion
Generation is an event, not an object
It distinguishes a model that was accurate on one occasion from one that is accurate, and shows why audit files built to retrieve AI reasoning capture fabrication rather than reconstruction.
Evidence & DisclosureReasoning, explanation & audit
Feb 2026
The Signature Fiction
Liability and the AI-drafted document
It shows that AI-assisted drafting leaves the meaning of a signature unchanged while hollowing out the defensible reasoning it is supposed to certify, connecting the governance bargain to documented professional failures.
Responsibility & LiabilityAccountability & liability
Jan 2026
AI's Single-Path Illusion
Where strategy disappears
It names a specific mechanism, the Single-Path Illusion, by which AI removes the visible alternatives that strategic judgement and board oversight depend on.
Executive Judgement & AccessJudgement, strategy & option-spaceGovernance & oversight
Jan 2026
Gravity of the Generic: Why Distinctiveness is Now a Cost
Augmentation is the common description. But it is misleading.
It names and evidences the 'gravity of the generic', reframing AI augmentation as a measurable pull toward sameness that most approval processes have no language to resist.
AI-Shaped LanguageAI-shaped language & homogenisation
Jan 2026
Human Middleware
The first draft is now cheap. The second draft is the cost.
It names and accounts for the hidden human labour that makes AI output usable, and frames the durability of AI adoption as depending on the sustainability of that oversight layer rather than on technical reliability.
Capability & Its ErosionHuman oversight & middleware labourAI-shaped language & homogenisation
Jan 2026
Residual Logic
Why the First Move Determines the Game
It names residual logic as inherited model framing that persists through editing, and reframes editing an AI draft as selection rather than authorship, with framing decisions functioning as unstated policy.
AI-Shaped LanguageAuthorship, framing & organisational reasoning
Jan 2026
Brittlement
The erosion of human capacity through routine offloading
It names the hidden atrophy of human cognitive capacity under routine AI offloading and locates the risk in a generational transfer gap and in boundary cases where eroded judgement fails.
Capability & Its ErosionHuman skill & its erosionHuman oversight & middleware labour
Jan 2026
Model Volatility
When the tool changes without telling you
It names the silent, unannounced change in hosted models as a distinct structural risk and shows why reproducibility and audit fail when behaviour, not access, is what organisations depend on.
Model Behaviour & ReliabilityModel reliability & driftCopyright, contracts & litigationRegulation & standards
Jan 2026
AI is a One-Player Game
Why LLM capability refuses to scale
It reframes enterprise AI as a non-scalable tacit skill, challenging the assumption that model capability can be captured and distributed like software.
Capability & Its ErosionTacit knowledge & prompting skill
Jan 2026
The Medium Is the Message
What the Data Says About Your Embedded AI Strategy
It uses market-share data to question the dominant embedded-AI strategy, arguing that treating a reasoning system as a feature inside legacy software may be an architectural mistake.
Vendor & Market DynamicsMarkets, moats & competitionAccountability & liabilityAdoption mandates & strategic non-use
Jan 2026
The Three Percent Revolution
What Banking's AI Projections Actually Tell Us
It reads banking's own AI workforce projections against the rhetoric to show incremental automation being relabelled, adding a numbers-first corrective to claims of sweeping change.
Workforce, Jobs & AutomationJobs, workforce & displacementProductivity & value measurement
2025
Dec 2025
You Cannot Negotiate with Code
Why Physics, Not Policies, Will Govern AI
It argues that governing autonomous agents is a matter of architecture rather than written policy, relocating authority over acceptable risk from legal and risk functions to system designers.
Responsibility & LiabilityGovernance & oversightAutonomous agents & agentic systemsReasoning, explanation & audit
Nov 2025
The December Uncertain
The Doubt at the Heart of the Generative AI Revolution
It treats the public uncertainty of a leading AI architect as a planning signal, arguing strategy should hold AI forecasts more lightly than pitch decks imply.
Memory, Continuity & Frontier AIFrontier AI & self-improvementJudgement, strategy & option-space
Nov 2025
Tesla's Three-Year Window
The $400 Billion Valuation Gap Between Tesla as Tech Company and Tesla as Automaker
It applies organisational and financial analysis to Tesla's robotics bet, arguing the valuation has priced in a reinvention that manufacturing, data and supply-chain realities may make impossible.
Vendor & Market DynamicsMarkets, moats & competitionHumanoid robots & embodiment
Nov 2025
Unready but Unstoppable
The Economic Forces Rushing Humanoid Robots Into Our Homes
It grounds humanoid home robotics in care economics and hidden teleoperation, showing adoption will proceed on cost logic despite the technology being unfinished.
Robotics & EmbodimentHumanoid robots & embodimentHuman oversight & middleware labour
Oct 2025
What Changes When AI Gets Real Memory
Inside the push to solve artificial intelligence's continuity problem
It maps memory as the next infrastructure layer in AI and identifies whoever controls it as the emerging locus of influence over how agents reason and decide.
Memory, Continuity & Frontier AIMachine memory & forgettingVendor lock-in & dependency
Oct 2025
Forget Forgetting
How AI Memory Systems Sidestep Biology
It draws a hard line between forgetting and configured deletion in AI memory systems, adding a mechanism-level critique to the wider account of memory infrastructure and lock-in.
Memory, Continuity & Frontier AIMachine memory & forgettingAI-shaped language & homogenisationReasoning, explanation & audit
Oct 2025
AI and the Instruction Delusion
Why it's not a machine to command, but a mirror to understand
It reframes AI unreliability from a technical fault into a strategic risk rooted in a mistaken mental model, positioning the human-machine interface, not the model, as the site of competitive advantage.
Model Behaviour & ReliabilitySystem prompts & instructionsTacit knowledge & prompting skillModel reliability & drift
Oct 2025
Deconstructing the Walmart-OpenAI Alliance
The race to become the new starting point for shopping
It reads a celebrated retail partnership as a contest over consumer intent and data, distinguishing the temporary feature Walmart gains from the permanent asset OpenAI acquires.
Vendor & Market DynamicsAutonomous agents & agentic systemsRetail, commerce & AI customersMarkets, moats & competition
Oct 2025
The 10,000-Hour Gamble
Is today's AI mastery only a temporary advantage?
It tests the durability of Foster-Fletcher's earlier mastery argument against the possibility of an architectural breakthrough, isolating computational scepticism as the transferable skill that survives a paradigm shift.
Capability & Its ErosionTacit knowledge & prompting skillHuman skill & its erosion
Sep 2025
The 10,000 hour sprint
Why AI mastery looks like obsession, not efficiency
It defines what genuine AI mastery is and why organisations cannot see or measure it, extending the gravity-of-the-generic argument into a claim about hidden capability gaps.
Capability & Its ErosionTacit knowledge & prompting skillAI-shaped language & homogenisation
Sep 2025
The Gravity of the Generic: Resisting the Pull
Resisting the Pull of AI Mediocrity
It introduces the gravity of the generic as a structural account of AI-driven linguistic and cognitive convergence, grounding it in corpus evidence and the model's own self-description.
AI-Shaped LanguageAI-shaped language & homogenisationAI companionship & product identity
Sep 2025
Beyond the Generalist Chatbot
Why 2026 is the Year of the Specialist AI
It makes the operational and governance case that capability does not scale from a single general model, predicting a shift to specialist, multi-model toolsets and the fragmentation that follows.
Vendor & Market DynamicsMarkets, moats & competitionGovernance & oversight
Sep 2025
The McLean-Walton Divergence
A Forty-Year Lesson in Who Will Actually Profit from AI
It uses a forty-year business-history parallel to relocate AI advantage from the models themselves to non-replicable supporting assets: feedback loops, trust and codified judgement.
Vendor & Market DynamicsMarkets, moats & competitionTacit knowledge & prompting skillJudgement, strategy & option-space
Aug 2025
The Grind vs. The Forge
Is AI Removing the Struggle That Shapes Great Leaders?
It adds a distinction between repetitive apprenticeship work and judgement under consequence to the question of how AI changes the formation of future leaders.
Capability & Its ErosionHuman skill & its erosionJudgement, strategy & option-space
Aug 2025
Most organisations aren't adopting AI
Are Vendor Algorithms Defining Company Culture?
It reframes enterprise AI adoption as a loss of organisational control, arguing vendor algorithms quietly standardise a company's culture and reasoning.
Vendor & Market DynamicsCulture & vendor-shaped organisationsModel reliability & driftAuthorship, framing & organisational reasoning
Aug 2025
Challenger or Companion?
How OpenAI's Identity Crisis is Becoming a Strategic Threat
It frames a major AI vendor's product-personality tensions as a strategic vulnerability that focused competitors can exploit, connecting model optimisation choices to market positioning.
Vendor & Market DynamicsAI companionship & product identity
Aug 2025
Meta's Superintelligence Gamble
Can Zuckerberg's Ambition Outrun the Limits of Technology and Time?
It sets a large frontier-AI bet against concrete financial, competitive and technical constraints, and proposes measurable 24-month tests for whether the bet is working.
Vendor & Market DynamicsFrontier AI & self-improvementMarkets, moats & competition
Aug 2025
AI Is Not Just the Tool, It's the Test
Are Companies Using AI to Augment Their Workforce or to Filter It?
It reframes AI usage mandates as a diagnostic that exposes thin workforce capability, and offers a single test distinguishing augmentation from filtering.
Workforce, Jobs & AutomationAdoption mandates & strategic non-useJobs, workforce & displacementProductivity & value measurement
Jul 2025
Using AI Just for Productivity Gains
It argues that framing generative AI as a productivity tool misreads its significance, using OpenAI's own commissioning of economic research and dissenting economists to reframe the value question around business-model obsolescence.
Vendor & Market DynamicsProductivity & value measurement
Jul 2025
AI isn't just replacing human roles
AI reveals a strategic failure. Our first reflex should not be automation.
It traces today's automatable roles back to Taylorist scientific management and its office descendant, reframing AI analytics as a diagnostic of management choices rather than a licence to automate.
Workforce, Jobs & AutomationJobs, workforce & displacementCulture & vendor-shaped organisations
Jul 2025
Under Siege
Can Getty and Shutterstock Survive the Rise of Generative AI?
It documents how the incumbent stock-photo agencies responded to generative image models through litigation, a merger and data-licensing, and identifies provenance and indemnity as the market that may survive.
Vendor & Market DynamicsCopyright, contracts & litigationMarkets, moats & competition
Jun 2025
Serving AI Shopping Agents
It examines how the largest retailer is re-engineering marketing, product data and interoperability for autonomous shopping agents, offering a blueprint for retail adapting to machine customers.
Vendor & Market DynamicsRetail, commerce & AI customers
Mar 2025
Bill Gates Warns AI Will Take Most Jobs
Why It's a Wake-Up Call, Not a Prediction
It separates Gates's job-displacement warning as policymaker advocacy from the technological reality, enumerating the breakthroughs required and advising business leaders to focus on mastery and adoption cycles rather than imminent displacement.
Workforce, Jobs & AutomationJobs, workforce & displacementHumanoid robots & embodiment
_1781012389223-Bb_8V6Hx.png)