Artificial Intelligence Archives - Gérard Métrailler

October 27, 2026October 17, 2026

Usage Pricing Breaks at Renewal

Charging for usage, or for AI tokens, turns every renewal into a new sale. The transition everyone cites as a success story shows where that breaks in the org chart, and how the comp plan has to change to catch it.

TL;DR: Informatica, the standard example of usage pricing that worked, reported renewals below forecast in February 2025. Its CEO told investors that when a customer’s executive sponsor changed, the account team knew and the renewal team didn’t, and that changing the teams’ incentives was the fix. Usage pricing makes each renewal a fresh purchase, and org charts and comp plans were built to sell once. AI token meters make it worse: usage swings with the customer’s agents, so commission on raw usage rewards noise. Compute the share of variable compensation tied to usage; that is the pricing model you actually run. Make one owner accountable for each defended renewal, while 2027 comp plans are being written.

October 20, 2026October 4, 2026

Shadow AI Is a Vote

Your approved-tools list records what the company authorized, not what it runs. A local agent installed in ten minutes is not on it.

TL;DR: Every AI tool your team uses without asking is a vote on the official path: too slow, too rigid, or missing. Read the votes before you ban the tools: giving people a sanctioned tool cut personal-account use where blocking did not, and nobody measures how long it takes to say yes. Data leaves through free accounts, tools expensed on a card, and local agents. A local agent installs in ten minutes with full access to the laptop. Praise AI in speeches and punish its early adopters, and people learn to hide. Run an amnesty inventory, leadership first, and publish time-to-yes by requester level. Fix the cause with training and a sanctioned alternative, offered fast.

October 13, 2026September 27, 2026

If You Had to Change What You Charge For, Could You?

Every AI contract you sign assumes a pricing unit the market hasn’t settled on. The vendors coping best stopped betting on one.

TL;DR: Salesforce has priced one AI product four ways since May 2025, and the market splits between charging for effort (tokens), output, and outcomes (all-you-can-eat wasn’t sustainable after all). The vendors coping best sell a currency instead: credits whose exchange rate they control, so the next change happens in a rate table. Fund usage data and the currency layer; defer billing, contracts, and comp tied to one unit. Your customer’s budget, built on your unit, is what you can never migrate. Output or outcome pricing leaves rising inference costs on your P&L, and not in a good way. Freeze the exchange rate for each contract term so customers can budget. Bill this year’s usage through two candidate units, on paper, and record what breaks.

October 6, 2026September 26, 2026

An Ode to (Choosing to Be) Busy

Working hours halved, and the complaint never moved. Busyness was doing another job.

TL;DR: Everyone is still slammed, though we work far fewer hours than our great-great-grandparents, and AI now turns out a first draft in minutes. Busyness is a choice, and a rewarded one: a full calendar makes its owner look scarce and in demand, no matter what got finished. The hours AI saves go straight back into other work, plus new work nobody owns. The exchange rate is cultural: Americans reward the overloaded week, Italians mark it down. My own model of a fully loaded leader’s week leaves two days open. Check next week against your prioritized list and what an agent could do; empty one recurring block.

September 22, 2026October 4, 2026

Everyone Bought the Computers. Six Sectors Got the Gains. Will AI Be Different?

A square poster of the frames illustration on a freestanding street billboard in early-morning Milan, with a pavement espresso bar three metres in front and every patron turned away from it toward the counter.

The computer revolution was in the 1990s. The AI transformation is playing out now, and hardly anyone has kept evidence showing which side they are on.

TL;DR: Your board will ask what the AI spending bought, and most companies can no longer answer, because nobody measured the work the deployments were about to change. The lag in returns is real, and it sorts the buyers from the beneficiaries. Six sectors out of sixty took 99% of the 1990s gain, and the heaviest spenders missed it. You cannot locate yourself in a distribution without a “before”, and almost nobody took one. Engineering and support kept timestamps, so those baselines can still be recovered.

September 15, 2026September 2, 2026

The Oldest Skill in Your AI Budget

An early-morning photograph looking down a long empty shared workspace toward tall windows, with a watercolor print of two pinned sheets displayed large on a board in the foreground.

Briefing a person and briefing an agent are the same act. Most people were never taught to do either, and the second one shows up in minutes.

TL;DR: Mid-market companies now put more investment into AI than into anything else, and most organizations credit it with under 5% of EBIT. What sits between the spend and the result is an ordinary management skill: stating the outcome, supplying the context, defining what done looks like. Only one in three workers has received any employer-provided AI training in the past six months. The existing training covers prompting and stops where the work gets useful. Nobody in your company has read anybody else’s prompt, so nothing compounds. Ask three people for the prompt each used on one task.

September 8, 2026August 22, 2026

Authority Delegates. Accountability Doesn’t.

A loose watercolor print lying on a hotel writing desk at night, showing a blank paper roll with a single blue thumbprint at its head, with the New York skyline soft beyond the window at blue hour.

Hand work to a person, and part of the risk goes with them. Hand it to an agent, and none of it does. The fix everyone reaches for targets the wrong variable.

TL;DR: A human delegate absorbs part of your risk because they have assets, a career, and legal standing of their own on the line. An agent has none of that, so authorizing one concentrates your exposure at the moment it feels like you are sharing it. Liability attaches regardless of how closely anyone was watching the work. People correct AI errors less when correcting costs effort, and money does not help. Skipping every approval prompt is itself the decision, taken once, by one person. Write down what it can reach and what it could destroy before you authorize the next one.

September 1, 2026August 18, 2026

Directionally Correct, Specifically Utterly Wrong

A pinned watercolor sheet on a cork noticeboard in an evening library reading room, showing a chain of fading links with a weight hanging from the faintest one, with Paris rooftops soft beyond the window.

The errors in a generated research report are not spread evenly. They cluster where the decision is, and the rule you already use to catch them is aimed at the wrong variable.

TL;DR: A generated research report is right only if every link in it holds, so accuracy falls away as the steps pile up. In one benchmark’s public filings, the same models score 83% on answering what a document says and 35% on answering what follows from it. Verifying the recent and the obscure misses it; decisions rest on public material. The mode that most resembles diligence fabricates the most citations. Mark every claim retrieved or composed, and re-source only the composed ones before money moves.

August 25, 2026August 8, 2026

An Ode to Electronic Communication (Now That Words Are Free)

An early-morning kitchen table in a city apartment above Tokyo, where a tablet shows a watercolor illustration of three small shapes, one enormous block, and the same three shapes again.

Writing got free, reading did not, which makes every message you send a withdrawal from somebody else’s day.

TL;DR: AI made writing free and left reading as expensive as it always was, so every inflated message moves work from the person sending it to the person receiving it. The meter is bolted to the side that got cheaper, and nothing counts what lands on the other. The padding a model adds is billed twice, once in tokens and once in attention. Five daily AI users voting together now spot machine prose 299 times out of 300. Change what you ask the tool for: cut the draft in half every time. Send nothing that came back on the first prompt.

August 18, 2026August 4, 2026

The Diagnostic You Already Ran

An open sketchbook lies on a worn workshop bench in low morning light, its page showing a hand-painted coffee grinder drawn taken apart, most pieces laid out free and three still tethered to the body.

How much of your work an agent can take depends on how completely you can specify it, and you are paying for the gap on every single run.

TL;DR: Everyone has the same models, and nobody has the same space around them, which is where the real variable sits. The framework for measuring that gap ran in a management journal back in 2005, written for an entirely different purpose. Two questions decide it: can you codify the work, can you measure the result. What ruled work out then, untransferable team knowledge, is exactly what you can now supply. A colleague is onboarded once; an agent without memory is onboarded every time. Specify once, where the agent reads by default; unclear thinking now runs a meter.