"Where there is no standard, there can be no kaizen."

Robots

People

Love

Reading time: 3 min.

A copilot saves engineers and planners 10% to 15% of their time. Reshaping a whole function around AI lifts its efficiency and effectiveness by 30% to 50%, BCG estimates in a new report on industrial companies, where the AI leaders grew revenue more than twice as fast as laggards.

Most are still handing out tools. Almost 70% of US industrial companies that adopt AI change nothing about their organization, data or processes, Census data in the report shows, and much of their know-how sits with senior managers rather than in documents.

Start with one function, such as production planning or aftermarket service, and redesign it with the planners who run it, keeping its rules in a layer you own rather than inside one model. The agent advises first, then acts with approval, and earns wider limits only while the numbers hold against a baseline Finance signs.

How human-centric AI delivers business results

Your weekly briefing on tools that create competitive leverage

Crux Labs learns how your experts judge an agent's work, checks all live traffic against that standard and opens a code fix when the agent slips. It emerged from stealth with more than $6 million recently, and its CEO says agents lack "the years of heuristics" people carry in their heads. Ideal for Chief AI Officers deciding how much room each agent has earned.

Rainbird turns policies and expert interviews into versioned rules that agents call over MCP, with the evidence behind every answer. You own the rules and can host them in your own cloud. Right for CIOs in banking and insurance who must show a regulator why each decision was made.

Ben Duffield helped create Palantir Foundry. With Lumaril, which he co-founded, he is working on knowledge that rarely gets written down: how veteran plant operators adjust hundreds of interlinked settings by feel. Lumaril's AI plugs into the controls a plant already has and learns from its best operators. The company raised a $66M Series A recently, and it gets paid on the throughput and yield it adds.

In 1985, Matsushita's engineers couldn't get a new bread machine to knead properly, even after comparing X-rays of its dough with a baker's. Software developer Ikuko Tanaka trained with the head baker at the Osaka International Hotel and noticed his distinctive way of stretching the dough, so engineers added ribs inside the machine to copy it.

Your plant has its own head baker. Write their rules down with them before an agent touches their work, then test those rules on last quarter's decisions.

Ikujiro Nonaka tells the whole story in HBR.

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

NUMMI, the GM-Toyota plant in California, ran in an old GM factory with mostly former GM workers and still came close to Toyota's own quality. James Womack, Daniel Jones and Daniel Roos explain why in The Machine That Changed the World: across car plants worldwide, automation explained only part of the productivity gap, and how the work was organized explained much of the rest. Worth an evening if most of your AI budget goes on seats.

Almost 70% of US industrial companies adopting AI follow the red dress.

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!