Enterprise and government deployment experience, in plain language for executives and decision-makers. Every article has an audio version.
16 articles co-written with Maria Singh, Dr. Christina Chua and Chris der Kinderen. Meet them ↓
I write with practitioners when the subject is their field, not mine. They check the facts I would get wrong and put their name to the result.
Vice President, The AI Collective; Founder, AISecurityLab.ai
At T-Mobile she helped cut critical vulnerabilities by 90% across more than 75,000 endpoints.
Latest togetherFollow the Actions: Detecting Rogue AI Agents Before the Damage Spreads
Director, Institutional Sales, Moomoo Financial Singapore
More than ten years in financial services across private wealth, key accounts and institutional clients.
Latest togetherInference Cost Accounting for Finance Teams
Project Delivery Manager APAC, Vontier
Has delivered forecourt technology programmes worth more than US$70 million in Malaysia and India.
Latest togetherThe Future of Payments and AI at the Fuel StationThe fear sitting under AI decisions, from the boardroom to the shop floor. New part weekly.
5 of 7 parts published
Latest, part 5Governance Theatre Is What Fear Looks Like on a SlideWith Dr. Christina Chua
Baselines, attribution, inference cost and stage-gates.
4 of 6 parts published
Latest, part 4Inference Cost Accounting for Finance TeamsWith Maria Singh
How websites become targets, how to detect a rogue agent and how to bound its autonomy.
3 of 3 parts published
Latest, part 3Put Boundaries Around Autonomy: Practical Controls for Rogue Agent RiskThe AI takeoff already under way, the endgames researchers have mapped, and the narrow path that keeps people in charge. New part weekly.
3 of 10 parts published
Latest, part 3Agent Swarms and the Closing LoopNamed organisations and what their own filings, regulators and the press show about how they built, used or failed with AI. Each one has a talk deck and an audio version.
Agents now work in teams for days, and the labs use them on AI research itself. Cursor ran a swarm for close to a week and OpenAI reports 3.1 agent-workdays for every human workday. What happens as the human step in the loop gets shorter, and which checkpoints to keep. Part 3 of a ten-part series on the future of humanity.
Read article →In September 2026 Claude wrote a computer-checked proof of Fermat's Last Theorem in Lean, about 13 million lines long, in 11 days. In July the same checker was fooled into accepting a false proof. A formal proof language lets a machine certify a result without a human referee, and it changes what truth and expertise mean. Part 2 of a ten-part series on the future of humanity.
Read article →Vernor Vinge and Ray Kurzweil taught us to expect the singularity as a single moment, around 2045. In the last eighteen months the people building frontier AI started saying it has already begun, and the best mathematician alive told his field to prepare for more proofs than it can digest. No alarm went off. Part 1 of a ten-part series on the future of humanity.
Read article →At a Federal Reserve conference in July 2025, Sam Altman said some banks still accept a voiceprint to authorise large transfers, and that AI has defeated it. The same weakness reaches the family phone call and the video meeting. This post sets out what has broken, what has been lost to it, and how to verify identity when a voice or a face is no longer enough.
Read article →With Dr. Christina Chua
The price per token falls every quarter while the AI bill rises, because volume, tokens per task and model choice grow faster than prices fall. How to tag every model call to the use case that ran it, build a cost-per-outcome line that sits beside the benefit, and run a monthly AI cost review on three numbers. Part 4 of the AI Benefit Realisation series.
Read article →In 2018 Jeff Bezos said that 90 percent of the work people do is never seen, and that only professional pride keeps the standard high on that part. Most AI products today are a screen and a prompt on top of someone else's model. The value, the safety and the staying power sit in the part underneath. This is a case for building it, and for being proud of it.
Read article →With Maria Singh
A rogue AI agent says reasonable things while it does unreasonable ones. The first AI-run espionage campaign was caught by its tempo and volume, not its words. Maria Singh and I set out twelve action signals a security team can alert on, how damage spreads between agents, and the logs and clocks regulators now expect. Part 2 of 3 on rogue AI agents.
Read article →With Maria Singh
If an AI agent can always be fooled some of the time, the controls that matter limit what it can do when it is. Maria Singh and I set out five autonomy tiers, the hard limits that hold outside the model, human approval that is more than a rubber stamp, kill switches that have been tested, and a 90-day plan for banks and government agencies. Part 3 of 3 on rogue AI agents.
Read article →With Maria Singh
More than half of web traffic is now automated, and a growing share of it is AI agents acting for someone. Agents attack websites at machine speed, and websites can hijack the agents that read them. Maria Singh and I set out both routes, the cases that prove them, and why banks and government agencies carry the most risk. Part 1 of 3 on rogue AI agents.
Read article →Marc Andreessen says the gap between what AI-native startups and established companies believe AI can do is the widest of his career. The evidence mostly supports him. Here is why the gap exists inside incumbents, where his argument goes too far, and what a board can do about it this quarter.
Read article →In October 2026 one developer released seven open-source Adobe alternatives written with Claude, and both large AI labs shipped their own documents, slides and animation tools inside the chat window. The software most exposed is the kind most people use only lightly. For enterprises, the bigger question is who governs the software a model builds on demand.
Read article →With Maria Singh
Brain-inspired neuromorphic chips run AI as bursts of events inside sensors, cameras and controllers, and some of them keep learning after they ship. That changes cybersecurity in three ways: attackers can work through light, timing and power as well as code; the usual scanners and patches do not reach the chip; and the device you approved may not be the device in the field a year later. Maria Singh and I set out what changes, what the research already shows, and how to govern AI that adapts inside physical devices.
Read article →I often see people produce impressive work with AI that they cannot explain, and cannot fix when it breaks. As an employer, that work is of limited use to me. The studies point the same way: people who let the tool do the thinking learn less, debug worse and misjudge their own speed. We should teach the process, because the process is where the skill comes from.
Read article →With Dr. Christina Chua
An AI pilot rarely runs alone. Routing changes, campaigns and new hires land in the same quarter, and a before-and-after comparison credits all of it to the model. How holdouts, staggered rollouts, switchbacks, synthetic controls and trend-adjusted comparisons separate the AI's effect, what each costs, and which to use when no control group is possible. Part 3 of the AI Benefit Realisation series.
Read article →With Maria Singh
A vulnerability score describes a flaw. It says nothing about the AI agent standing behind it: how freely it acts, what it is allowed to touch, which tools it can call, what it remembers and which business process it runs. Agent-aware vulnerability management adds those five things to exploitability before deciding what to fix first. ISO/IEC 27090:2026, published in October, gives the threat side of the same picture. Here is how the two fit together and how to run it.
Read article →Three in four organisations say they have an AI usage policy. Fewer than half monitor their production AI for accuracy, drift or misuse, and only a third of those with a policy regularly check for unsanctioned AI. Policies written to produce the feeling of control survive a board meeting and fail an audit. Here are three tests, drawn from ISO/IEC 42001 audit work, that separate a control from a performance.
Read article →With Dr. Christina Chua
A baseline taken after an AI pilot starts is a reconstruction, and reconstructions flatter the pilot. What to measure, for how long, how much the number moves on its own, and who signs the record before the first production model call. Part 2 of the AI Benefit Realisation series.
Read article →With Maria Singh
Most organisations have spent years hardening their software, cloud and vendor supply chains. AI adds a dependency those programmes were not built for: the model, and the training data, fine-tunes, retrieval corpus and tools that shape what it does. It is probabilistic, hard to inspect and usually sourced from outside. Here is where the model supply chain breaks, what the public record shows, and how to design integrity, provenance and behavioural limits to the same standard as the rest of the chain.
Read article →What I learned building an empathic AI companion: why reading emotion is harder than it looks, how emotionally aware AI changes the human-machine interface, the laws already in force, and the guardrails that keep these conversations safe.
Read article →With Dr. Christina Chua
Most AI programmes cannot prove what they returned because nobody recorded the starting point, nobody agreed how to separate the AI's effect from everything else, and the running cost sits in a cloud bill no one maps to a use case. A framework the CFO and the Chief AI Officer run together: baselines, attribution, inference cost accounting and stage-gates that stop funding for pilots that miss their targets.
Read article →Where AI can safely take load off doctors and nurses today, how it can help train them, and the risks hospital leaders need to control: automation bias, errors carried into the record, deskilling, unsafe drafts and patient data leaving the building.
Read article →With Maria Singh
Security reviews score findings one at a time and assign each to the team that owns the asset. An AI agent connects those assets by design, so three Low findings can add up to one critical path that no team owns. FIRST's CVSS guidance already says how to score chains, and OWASP's agentic guidance says how to break them. Here is how to govern the path as well as the parts.
Read article →With Dr. Christina Chua
Brokers are letting AI agents place orders, compliance teams are handing e-communications review to language models, and Project Guardian has put tokenised bonds into production. Each move raises a control question that boards in Singapore have not yet answered: who is liable when an agent trades outside its mandate, how to prove a quieter alert queue is a better one, and whether live asset data can be trusted to value a tokenised infrastructure bond.
Read article →Being late is the dominant fear in the mid-market right now. Two in three CEOs say the risk of falling behind has pushed them to buy technology before they understood its value, and half say the pace has left them with piecemeal systems. Here is what being late costs, and an order of operations that protects a budget from its owner's anxiety.
Read article →With Maria Singh
Most organisations defer some patches and record the decision as an accepted risk. Those decisions assumed that turning a patch into a working exploit took skill and weeks. Anthropic's 2026 evaluations show a model doing it in under an hour for about $2,000 per exploit, and its September threat report describes criminals running exploit research around the clock. Here is how to rewrite the risk register.
Read article →With Chris der Kinderen
The fuel station is one of the first places in retail where AI, the connected car and new payment rails meet at scale. What changes at the pump and in the shop, where the risks sit in pricing, fraud and customer data, and what operators should do in the next twelve months.
Read article →Your manager shuts down every AI proposal, and you can see competitors moving. A practical guide to reading the fear behind the blocking, building a case your boss can say yes to, protecting yourself when the behaviour turns toxic, and knowing when the honest answer is to leave.
Read article →Too early to retire and feeling too late to start again. A note for the generation that has adapted to every wave since the fax machine, on the tiredness underneath the anxiety, how much working life is left, and what our experience is worth now that AI writes the first draft.
Read article →With Dr. Christina Chua
Banks, exchanges and central banks are moving money and securities onto the same programmable rails that DeFi built. By 2031 the two systems will share settlement, collateral and market mechanisms, with regulated access on top. AI arrives at the same time as customer, attacker, target and supervisor. Here is how the integration will unfold, where AI changes the risk, and what boards and regulators should do now.
Read article →With Maria Singh
No quantum computer can break today's public-key encryption yet, but the estimates of what it would take fell twentyfold in a year, and Google now targets 2029 for its own migration. AI systems rest on the same RSA and elliptic-curve cryptography: the links that carry training data, the signatures that vouch for models and the tokens that identify agents. Here is what breaks, what does not, and what boards should do now.
Read article →DBS reported about S$1 billion in economic value from data analytics and AI in 2025. The figure blends revenue, cost avoidance and losses averted, is measured with control groups, and sits outside the audited accounts. Here is what it counts, what it leaves out, and what a board should demand before quoting its own number.
Read article →Between November 2021 and October 2023 DBS suffered a run of digital outages. MAS responded with about S$1.6 billion in extra capital and a six-month pause on non-essential IT changes. The failures came from the same modernisation that made DBS a leader, and the lesson for boards is to govern technology risk before the pace of change outruns it.
Read article →In 2018 DBS reorganised into 33 platforms, each led jointly by a business head and a technology head on one scorecard, funded as a standing unit and reviewed on live dashboards. Here is how the model works, what it added in 2021, and how a smaller organisation can adapt it.
Read article →Stalled adoption gets diagnosed as a change management problem and treated with more training. The behaviour is rational. People are protecting their standing, their expertise and their income, and one written commitment moves uptake further than any training budget.
Read article →Students, developers, knowledge workers and specialist physicians all lose measurable skill when AI does the thinking for them, and the loss shows up within minutes to months. The studies are early and uneven, one headline study has already been retracted, but the direction is consistent. The same studies also show what protects the skill, and none of it requires giving the tools up.
Read article →Regulators publish more than two hundred changes a day and most compliance teams still track them in spreadsheets. AI can read, sort and map that volume to your obligations and controls. It cannot be trusted to interpret the rule, and the regulator will still hold the institution to what it said.
Read article →Most agents have adopted AI and half of them report no difference to the business. The gain sits in the first hour after an enquiry and in scoring leads on behaviour, and every AI output a buyer sees still needs a licensed human to sign it off.
Read article →The mandate arrives written as strategy and starts as the fear of being seen as behind. Three questions tell you which one you have, and there is a way to answer that costs nobody their standing.
Read article →A viral podcast argues artificial superintelligence will kill us all. The actual research behind that claim is narrower and more useful. No one has proven a smarter-than-human system can be controlled, and almost no one outside a handful of labs and governments has a say in whether we build one anyway.
Read article →Most AI programmes fail the same way products used to fail, by betting big before anyone tested the smallest version of the idea. Here is how Lean's build-measure-learn discipline makes AI adoption safer.
Read article →NVIDIA's DGX Station runs frontier models of up to a trillion parameters locally, and open frameworks like OpenClaw and Hermes give those models long-running agent bodies. The technology question is largely settled. The adoption questions, people, economics and operating model, are not.
Read article →The fear I had was not that the technology would fail. It was being the person who is supposed to know. Here is the test I now run before I let a feeling about AI drive a decision.
Read article →A project can score 9 on five dimensions and 2 on autonomy readiness and still win the ranking on a straight average. That is the flaw in treating all six scoring dimensions as equal inputs, and the fix is a gate, not another weighting tweak.
Read article →A six-dimension score only settles an argument if two different people scoring the same project land on roughly the same number. This is the rubric that makes that possible: what a 2, a 5, and an 8 look like on each dimension.
Read article →Most companies pick their first agentic AI project on instinct or whoever spoke loudest in the room. Here is a six-dimension scoring method, built for agentic AI's biggest wildcard: autonomy, that gives you a ranked answer instead of a guess.
Read article →Agentic AI is not just a chatbot with a new name. This guide explains the six parts that make an AI system 'agentic,' in plain words, for leaders who do not have time for jargon.
Read article →The ninety-day AI plan tells you when. This is the companion piece that tells you who. Exactly what the executive sponsor, data owner, frontline user, governance owner, and technical lead each need to do, at every one of the seven stages.
Read article →There is too much to read about AI. This guide cuts through the noise. It gives you six simple questions to answer, in order, so you know exactly how to start using AI at your company.
Read article →Six concerns show up consistently across the 2026 CHRO surveys: reskilling, governance exposure, employee trust, unproven ROI, unredesigned jobs, and the CHRO's own AI fluency. A practical, sourced response for each one, drawn from Gartner, Mercer, Deloitte, McKinsey, SHRM, and the Conference Board.
Read article →METR's data shows the length of tasks AI agents can complete alone doubling roughly every seven months, and companies are already cutting the management layer that 'you'll orchestrate the agents' was meant to fill. What's left for entry-level, mid-level, and senior knowledge workers, and why the honest answer is a design choice, not a reskilling roadmap.
Read article →IAB, Emplifi, and NIM data from 2026 show a widening gap between how positively marketers believe audiences feel about AI-generated creative and how audiences feel. What's driving the distrust, and what specifically closes the gap without giving up AI's production advantages.
Read article →An AIMS is the clause-4-to-10 system ISO/IEC 42001 certifies, not a written AI policy: risk assessments, impact assessments, roles, internal audits, and a statement of applicability an auditor can test. What gets checked first, and how to generate that evidence from day one instead of reconstructing it before the audit.
Read article →94% of security leaders say AI is the biggest driver of change in cybersecurity this year, and it is reshaping both sides of the fight at once. What the 2026 data shows, what defenders can deploy today, and the skills gap standing between here and a defensible position.
Read article →A trustworthy AI agent isn't a model you deploy once. It's treated like a new hire: skills assessed against a job description, KPIs checked against actual intent, re-benchmarked on a schedule, reviewed by a human who knows what good looks like, and rollback-ready the day it starts making real decisions.
Read article →A personal note to anyone leading their first real AI project: the wobble is normal, the disappointment is normal, and the way through is one small step, one honest team, and one task nobody else wants to do.
Read article →A corrected scaling law, real data on correlated model error, and the economics of compute scarcity all point the same way: near-total AI capability does not translate into near-total substitution. The five numbers that set the ceiling, and the one most enterprise business cases leave out.
Read article →Not cloud infrastructure, not copilots bolted onto existing software. A benchmark of the AI-native companies whose entire business is the AI product itself, what they charge for, and why the pricing model is the real signal to watch.
Read article →17,600 recorded actions, a self-respawning fleet across eleven Kubernetes nodes, and zero human operators. OpenAI and Hugging Face's own technical disclosures reconstruct the July 2026 incident in detail. Here's the evidence, and the concrete controls that would have stopped it.
Read article →Productivity and pay have moved apart for 45 years, and 2026's data shows AI accelerating the split, not closing it. What that means for workers who already built the skills the market asked for, and what businesses need to rethink about how wealth gets created.
Read article →Big Law and Wall Street are restructuring around AI-assisted triage, cutting the entry-level work that built the judgment their accountable humans still need.
Read article →Andrej Karpathy renamed his own term within a year of coining it. Google's code is now 75% AI-generated. The data shows unstructured AI assistance doesn't reliably save time, but orchestrating agent loops and graphs produces results no solo prompt ever could. What the director's job looks like, with real examples and a way to prepare.
Read article →The World Economic Forum expects 92 million jobs to disappear and 170 million to appear by 2030. The debate over whether that's good news or bad news is a distraction from the only question that matters: what you do about your own exposure, and where human judgement still holds value AI can't touch.
Read article →Opacity in machine learning is not one problem but three, and only one of them, opacity by construction, cannot be fixed by disclosure. Where it comes from, what it costs when it goes wrong, and the architecture that lets an organisation ship opaque components without depending on explaining them.
Read article →Global AI governance is fragmenting exactly as frontier capability crosses new risk thresholds. What the Financial Stability Board, the IAEA, and the Montreal Protocol got right that the current run of voluntary AI declarations does not.
Read article →Outcome-based AI pricing is sold as a full transfer of delivery risk to the vendor. Sixty years of performance-contracting history, from Rolls-Royce's Power by the Hour to a £467 million probation contract failure, says otherwise. What moves, what stays with the buyer, and what the contract has to specify.
Read article →Fluent, competent exposition now costs approximately zero to produce, and the market has repriced it to match. Here's what the field research shows about who gains, who narrows, and what still carries information when anyone can sound articulate on demand.
Read article →Analysis and research now cost nothing to run yourself. Here are the 10 categories of consulting work, from executive alignment to crisis judgment, where premium fees still hold in the AI era, and the four capabilities every firm delivering them shares.
Read article →Elon Musk told SpaceX employees they'd 'effectively be the parents' of Grok. Take that seriously and sixty years of parenting research becomes strangely relevant to AI governance. What decades of studying real children tell us about instilling values in something that will eventually outgrow every rule we write.
Read article →Human-in-the-loop assumes someone sees the bad action in time. The 2026 evidence, including Anthropic's own research, says that assumption fails. What containment looks like when detection isn't the plan.
Read article →Not the sci-fi fear. The mundane one: an agent that trusts the wrong instruction, a credential nobody's watching, a tool description that lies to the model reading it. Four real 2025 incidents, and what the security model has to look like once an agent gets write access.
Read article →The question sounds the same from a graduate, a manager, a business owner, and a CEO, but the anxiety underneath it is different for each. A close look at what's driving it, and what seems to steady people while the picture stays unclear.
Read article →A simple guide for Singapore businesses: what to check before an AI agent goes live, how to score it fairly, and the safety rules to set first.
Read article →Regulation gets the headlines, but it isn't what stalls most enterprise AI programmes. Security, unit economics, data infrastructure, and reliability are, and each one compounds if left unaddressed.
Read article →Reliability in agentic systems comes from the engineering around the model, not the model's capability. Ten requirements, from bounded task specification to trajectory-level observability, that determine whether an agentic deployment holds up in production.
Read article →For decades, more years on the job meant fewer mistakes and a bigger paycheck. AI now does in seconds what used to take a career to learn. Here is what earns you value now, and the entry-level problem nobody has fixed yet.
Read article →McKinsey found only 23% of companies are scaling agentic AI, and adoption within any single function still lags well behind that headline number. Here's what's running in production today, with named examples from Singapore banks and government.
Read article →So many teams ask how to start with agentic AI. Gartner says over 40% of agentic AI projects will be cancelled by 2027, and 88% of pilots never reach production. Here is the simple order that works, in plain English.
Read article →Fixing fragmented data is only half the job. Once that data is combined and fine-tuned into a model, the access controls that used to protect it disappear. Here's how to put them back.
Read article →Gartner forecasts 60% of AI projects will be abandoned through 2026 for lack of AI-ready data, and fragmented data scattered across old systems is the most-cited reason why. Here's how to fix your data before you buy the AI.
Read article →McKinsey found 88% of companies use AI regularly, but only about a third have grown it company-wide. MIT NANDA found only 5% of AI test projects lead to lasting profit. Three studies agree: the problem is how work is set up, not how smart the AI is.
Read article →A 2025 government report shows AI use by Singapore small businesses tripling to 14.5%. But the gap with big companies grew to 48 points. Five easy categories explain who is catching up, and one you should not trust yet.
Read article →A 2026 survey found 55.7% of governments use AI, but only 42.9% have a written policy for it, a 12.8-point gap. Here's which places are closing that gap, and why copying an existing policy works better than writing one from scratch.
Read article →Gartner says only 48% of AI projects ever make it into production. Five key decisions, made before you sign anything, decide whether your company's AI grows or gets stuck.
Read article →Training your staff to use AI tools is not the same as having an AI plan. Data from McKinsey, BCG, Deloitte, and PwC all shows the same gap: it's top leaders and boards who need to understand AI.
Read article →The governed cognitive layer's theory is one thing. Here's what breaks in production: contract boundaries, routing thresholds, memory architecture, and the failure modes your existing test suite won't catch.
Read article →Corporate technology has run on a seven-layer OSI model for forty years. Generative AI needs a governed cognitive layer, what I call Layer 8, that turns unpredictable reasoning into governed, auditable enterprise action.
Read article →Formal curricula move on approval cycles measured in years. Frontier AI moves weekly. A framework for closing that gap, for graduates entering the field and for engineers already in it.
Read article →We can generate reports, deploy code, and synthesise data faster than ever. What's harder to see is what that speed costs: professionals accountable for work they can no longer explain how they built.
Read article →AI is going into production faster than most organisations can build guardrails for it. AI Assurance is how you close that gap, with evidence, not good intentions.
Read article →The benchmark conversation is fading. What matters now is whether an agent finishes a real task reliably, and whether you can prove it. For Singapore's regulated and public-sector deployments, that makes traceability the real currency.
Read article →Entry-level hiring really is down, and AI is part of why. But the same shift that closed one door for new graduates has opened another, and I'm watching the founders I mentor walk through it.
Read article →Companies are building high-tech AI infrastructure before they've upskilled a single employee to use it. That sequencing error is why the tech spend is massive and the adoption is near zero.
Read article →The AI honeymoon phase is ending. What's left is a less glamorous question: can this run at scale, on a budget, without falling over. Five architectural moves for teams past the pilot stage.
Read article →AI accelerates the first 80% of a project and then traps you in an endless revision cycle. Here are the three psychological mechanisms behind that loop, and the boundaries I now set to get out of it.
Read article →For SMEs, AI results depend less on which tools they pick than on how well leaders agree on goals, how clearly processes are defined, and how firmly the work is phased.
Read article →Businesses built on 1990s assumptions now face serious operating risk. As geopolitics, cybersecurity, and AI converge, the leaders who thrive by 2030 will be the ones who questioned every assumption in 2026.
Read article →Enterprise AI transformation fails not because the technology is flawed, but because organisations suspend discipline in response to ambition. The fix is structural, not motivational.
Read article →As AI moves from pilot to production, a CAIO's role shifts from championing model capability to demonstrating return on investment, operational resilience, and risk containment. These four dashboards make that possible.
Read article →Standard SaaS contract templates were not designed for AI. Here are the ten clauses that protect you, and why most organisations only discover what's missing after something goes wrong.
Read article →Workers at more than 90% of companies are already using personal AI tools for work tasks. The governance risk is not the AI. It is your data leaving your systems without any controls in place.
Read article →MIT research is explicit: most AI pilot failures come from poor use case selection, not model quality. Here is the six-criterion framework for choosing right the first time.
Read article →Before buying any AI tool, you need to know whether your business can use one. This diagnostic draws on the same readiness framework used in enterprise deployments, reduced to five questions any business owner can answer.
Read article →Organisations are moving AI from pilots into live use, and failing at the first step. The TRACE framework gives leaders a clear way to test which tasks can be handed to an autonomous agent.
Read article →The software AI market is crowded and margins are compressing. The next large commercial opportunity in AI is embedded intelligence: AI built directly into hardware, appliances, and physical infrastructure.
Read article →Better prompts are table stakes. The real productivity gains come from structured AI loops: chained, recursive, self-improving workflows where Claude operates as an execution engine.
Read article →Agent frameworks are engineering infrastructure, and deploying them without a validated process baseline transfers accountability to a system that carries none. The correct sequence is tool fluency before agent deployment.
Read article →Disruption initiated from the outside removes agency. Disruption initiated from within is a controlled demolition: the organisation retains decision authority over sequence, scope, and pace.
Read article →Hosting large language models yourself is right only under certain conditions. The break-even point depends on more conditions than most buying conversations admit.
Read article →Enterprise AI programmes in Asia are speeding up. The models are running. But the way they are measured is still tied to the wrong layer of the technology stack, so decisions get made on incomplete information.
Read article →Organisations deploying AI at scale are accumulating token debt at pace. Without deliberate cost architecture, inference spend becomes the single largest barrier to production ROI.
Read article →'Rolls-Royce has charged airlines per flying hour (not per engine) since 1962. What is new in 2026 is that the same logic can now be applied to knowledge...
Read article →The conversation around AI in enterprise has largely been shaped by two competing anxieties: the breathless pace of adoption, and the fear of being left...
Read article →A bad process, accelerated by AI, produces bad results faster and at scale.
Read article →During InteracTech Asia 2026, HR leaders gathered to examine what it really takes to stay relevant in an AI‑driven workplace. I would like to share pivotal...
Read article →Most professionals adopting AI aren't failing because the technology is bad. They're failing because they're making the same predictable mistakes, entirely...
Read article →In 2026, the real barrier to AI is not access. It is the decision to begin.
Read article →The current AI market is characterised by rapid iteration, short product lifecycles, and an overwhelming number of seemingly similar offerings. The barrier...
Read article →Public and private organisations are both spending more on artificial intelligence, and spending it faster. Many are held back by the way they are set up.
Read article →For most of my career, the unspoken rule was clear: accumulate expertise, build a deep knowledge base, and stay ahead through technical competence. The more...
Read article →White-collar jobs are being lost faster, and from several directions at once. The technology sector recorded over 150,000 job cuts in 2025, mostly...
Read article →AI now moves on a cadence measured in months.
Read article →In many operational domains, AI systems can now perform oversight and checking more effectively than human reviewers, especially at scale and over long time...
Read article →Many SMEs are being pushed to “adopt AI” through grants, vendor pitches, and digitalisation roadmaps. They deploy a chatbot on their website, bolt an AI...
Read article →Most organisations have experimented with retrieval‑augmented generation (RAG), but very few have a production‑grade RAG that can withstand regulatory...
Read article →A Zero to One AI business sets out to create something new instead of making an existing solution a little better. The...
Read article →By 2025, roughly 78–89% of enterprises reported using AI in at least one business function. Yet only approximately 1% of those same organisations describe...
Read article →In AI strategy, 'disruptive' and 'destructive' are not buzzwords.
Read article →People assume that because I run AI and smart city programmes across Asia and the Middle East, I must lie awake at night worrying about rogue superintelligence.
Read article →An effective AI team is not a monolithic engineering squad. It is a multidisciplinary unit spanning strategy, engineering, data, operations, and governance.
Read article →The question of what it takes to achieve and sustain growth is not theoretical. It is a daily operating concern across multi‑billion‑dollar infrastructure...
Read article →Enterprise AI has shifted decisively. Gartner projects that 40% of enterprise applications will incorporate task-specific agents by the end of...
Read article →As generative AI tools become embedded in day‑to‑day work, many professionals are reporting a subtle but persistent strain: they feel mentally exhausted...
Read article →As organisations move from single large language model (LLM) chatbots to full agentic systems, the discussion is shifting from “which model?” to “what kinds...
Read article →In many organisations, AI leaders are encountering a subtle but significant form of resistance. It is not open objection to the technology, nor a lack of...
Read article →We're witnessing a fundamental shift that most organisations have not yet fully grasped.
Read article →For eighteen months, we watched enterprise AI initiatives proliferate. Millions flowed into vendor products, consultancy engagements, proof-of-concepts, and...
Read article →Life at work is faster and more efficient after AI, but for many leaders, it is not easier. Stakeholders expect you to “use AI on everything” and deliver...
Read article →AI systems fail less often because of model limits than because of unclear, inconsistent language. A precise, shared vocabulary works as technical infrastructure, and decides whether AI investments scale or stall.
Read article →The question for technology leaders, public sector executives, and infrastructure operators has shifted. It's no longer a question of whether to deploy...
Read article →The dominant fallacy of artificial intelligence is not that it 'does not work', but that it is routinely treated as autonomous, intelligent and dependable...
Read article →The value of an organisation will increasingly be determined by how effectively it adopts and scales new technologies, not by whether it uses technology at...
Read article →Voice AI is therefore best understood as a foundational capability in the next wave of digital infrastructure. For high-growth regions in Asia and the...
Read article →Large Language Models have dominated the artificial intelligence discourse since 2022, yet accumulating evidence from technical research, enterprise...
Read article →Your AI prototype is crushing it. Impressive outputs, smooth demos, and stakeholders are thrilled. Then you deploy it to production and... it starts falling...
Read article →The most dangerous moment in the deployment of artificial intelligence is not when systems fail catastrophically. It is when they deliver partial truths...
Read article →The era of AI experimentation is ending.
Read article →Except it's never done. And the biggest line item is something that rarely makes it into the budget: people.
Read article →Knowledge used to be power. AI changed that. As AI gets better at generating expertise instantly and practically for...
Read article →Most AI projects don't work out. We're talking 95% of enterprise AI pilots delivering basically zero ROI. And in 2025, 42% of companies threw in the towel...
Read article →A prediction error is not merely a statistical anomaly. It is an operational liability. A recommendation engine that misclassifies a movie causes mild...
Read article →We're at a critical moment in how we build technology. AI is moving from isolated innovation pilots to core infrastructure: the kind that runs cities,...
Read article →Here's something wild: while you're busy using AI to write your reports, your client is probably using their own AI to read them, before any actual human...
Read article →When I first heard about 'vibe coding' a few months ago, I was skeptical. The idea of letting AI write production code based on casual...
Read article →Last quarter, I watched our team burn through $8,000 in API costs because nobody bothered to write good prompts. We were getting useless outputs, running...
Read article →I’ve been speaking to a lot of you recently, and I can sense the anxiety. You’ve spent four years studying computer science, engineering, or data science,...
Read article →Last month, our team was contemplating about upgrading to an AI model with a 1 million token context window. 'We can throw our entire knowledge base at...
Read article →Last month, I walked through our department area and noticed something that's been keeping me up at night. On one side, I saw team members using AI tools to...
Read article →I spent three weeks fine-tuning prompts for our RAG system. Tweaking instructions, adjusting temperatures, testing different models. The results? Marginal...
Read article →Just not too long ago, we almost killed a landmark AI project. The technology was brilliant: our data science team had built a system that could process...
Read article →Three weeks ago, a colleague, someone I've worked with for 15 years, came into my office and quietly asked if I thought his job was safe. He's a senior...
Read article →No spam, no drip sequence. Just an email when a new article or tool goes live.
Most of what's on this page started as a note to myself after a project went a way I didn't expect, written mostly so I wouldn't forget the lesson before the next one. I publish the ones that are wrong sometimes too, because pretending every call I've made was right would make this page useless to you. If something here saves you a mistake I already made, that's the whole point of writing it down. Read the one that annoys you most.
