When software stops being a menu
Menus did two jobs and we only ever complained about one. As AI removes them, discoverability stops being a layout problem and becomes a teaching one.
48 essays
Menus did two jobs and we only ever complained about one. As AI removes them, discoverability stops being a layout problem and becomes a teaching one.
AI models are commoditising and human intelligence always was. Being smart was never scarce — the container you build around it is what creates value.
Your AI strategy is probably theatre, and your employees see straight through it. That visibility is a real cost, in trust, credibility, and adoption.
A brand-voice connector pipes your tone into every AI tool, but it can't judge when not to apply it. That judgement is policy — a layer you still own.
Google logged a 32% rise in prompt-injection attacks. If your AI agent reads the open web, it's in hostile territory now, and here are three controls.
AI has stopped behaving like software and started behaving like a raw material. Budget it as a variable cost, cap it and meter only what pays its way.
Buyers now ask an assistant who your alternatives are, and get a table you never wrote. Your competitive set is no longer argued for — it is assigned.
The 'AI replaces a $50k hire' lists see only tasks. A salary also buys judgement and accountability: the real work you can't hand to a stack of tools.
Big firms survive AI agent sprawl because they have a CIO; your small business doesn't. Here's a simple three-step protocol to contain it by yourself.
The agentic convergence trap: when every rival runs the same AI, the plans converge. Your real edge is judgement, taste and the data nobody else owns.
Six AI labs now ship the same agent. When the demos all look alike, choose on workflow fit, switching cost, data gravity and team habit, not features.
A third of B2B buyers now find vendors through AI chatbots. The real shift isn't AEO — it's that you've lost the ability to track your own visibility.
A team going quiet on AI isn't proof the tools failed. It's stage three of the doom cycle: a predictable adoption valley, and a management job to fix.
The AI subsidy era is ending and frontier-model bills are rising fast. A 90-day routing playbook to keep your SMB's AI costs predictable, not painful.
Most agentic software is assistive once it lands in your account. The questions to ask so you can tell roadmap from running product before you buy in.
Claude, ChatGPT and Gemini each wrap a model inside a product. Learn to tell the layers apart and you can lower your AI spend and buy software wisely.
The UK's £500M Sovereign AI Fund is the right call on speed, compute and visas — but it doesn't fix the reason British AI is worth more in California.
Gary Vee's 93% marketing waste figure is real, but the 20% reallocation fix won't happen — that wasted spend is funding the dashboard the board reads.
Per-seat pricing was a polite fiction. Four major vendors stopped pretending this fortnight. Here is what every SaaS buyer needs to renegotiate today.
AI's real return isn't automating tasks. It removes the coordination tax of meetings, briefings and approvals — the saving an ROI dashboard won't see.
AI is reshaping marketing on two fronts: the tools your team uses and the audience you reach. Most teams watch just one and miss the one that matters.
A wrongful-death suit over Google's Gemini shows AI safety is product architecture. What every team that ships an AI copilot must build in, right now.
NotebookLM — now Gemini Notebook — grounds every answer in your documents, not training data. Using it as a strategy partner that cannot invent facts.
Stanford research finds single agents usually beat multi-agent swarms on equal compute. The swarm tax is architecture inflation eating your AI margin.
Research shows heavy AI users are losing the thinking habits that made them valuable. The competitive edge goes to those who use it without surrender.
OpenAI proposes robot taxes and citizen wealth funds. The company most likely to displace workers is writing the rules. Whose side is the rulebook on?
Intercom is rebuilding its product so AI agents, not developers, are the primary user. What changes for your docs, errors, pricing, auth, and roadmap.
AI advice helps the best entrepreneurs and hurts the rest. A 2026 MIT Sloan Kenya field study, plus a five-step kit to second-check AI before you act.
DeepSeek V4-Pro just matched Opus 4.7 on coding at a seventh of the cost. The four-bucket framework for rewriting your 2026 AI budget — by Q3 ideally.
OpenAI and Anthropic disagree on whether AI should act on its own initiative for societal good. Why this gap matters when choosing an AI vendor today.
Shadow AI agents running on staff laptops with live credentials to company systems. KiloClaw named the crisis — and most governance is lagging behind.
Anthropic just cut OpenClaw from Claude subscriptions on 4 April. The platform tax has arrived — and it changes how to build on any AI provider today.
93% of AI spend lands on infrastructure, 7% on people. Why the unphotogenic work of retraining people around AI is where the real value actually sits.
AI didn't break marketing. It exposed which content was built on substance and which was noise. A practical audit for business leaders, not just CMOs.
Nearly 80,000 tech jobs cut in Q1 2026, half blamed on AI. Even OpenAI's CEO calls it AI-washing. A simple test separates theatre from transformation.
Anthropic shipped Managed Agents in public beta this week. Not a capability breakthrough, but a friction breakthrough - and why friction matters more.
Walmart's ChatGPT checkout converted three times worse than its own website. The Sparky pivot reveals a fundamental truth about who controls commerce.
Zencoder's team went from 36 engineers to 30 and nearly doubled throughput. A data-backed look at what AI-first software development really means now.
Bezos is raising $100 billion to buy manufacturers and rebuild them with AI. Why industrial AI, not chatbots, is the smart money bet in 2026.
AI agents have rewritten the compute equation permanently. Ben Thompson argues this isn't a bubble — and your business strategy needs to catch up now.
By 2028, Gartner predicts 90% of B2B buying flows through AI agents. Your brand now needs a schema as much as a soul. Most marketers aren't ready yet.
Each nine of AI reliability costs as much work as the one before it. Most businesses are at 90%. The march to production-grade agents is brutal maths.
Most businesses think their agentic AI is at level three. Most are at one. A five-level framework for honest self-assessment and practical next steps.
Forget workflows — AI agents run decision loops. Map where autonomous cycles belong inside your business, and you will outpace everyone else doing it.
85% of enterprises want agentic AI but 76% lack the operational foundation. Why process visibility matters more than model capability.
Most people use AI like a Ferrari in first gear. The literacy gap isn't about access — it's about understanding what these tools can do.
Y2K had a deadline. AI doesn’t. The real challenge isn’t fixing a bug — it’s managing work that now moves faster than your controls.
Most AI pilots return nothing. The five execution failures behind the $154 billion mistake — and the practical playbook to fix them in 90 days.