TL;DR
- Develop new expertise, even when it challenges existing products.
- AI shortens development paths and accelerates progress along them.
- Use AI to win customers and develop new services.
- People build the environment, set direction, and check results.
Experience holds the business together
A machine manufacturer does much more than assemble parts. It helps a customer choose a machine, designs it, buys the components, builds it, and keeps it running over time. Knowing how those activities fit together is part of what the customer pays for. A cheaper component may make the machine harder to service. A design that performs well under test may struggle with the material a customer actually processes. Experience connects these decisions.
For a company that has been doing this for a hundred years, much of that experience comes from repeated attempts to make things work better. Physical principles explain how a machine should behave. Years of building and servicing it reveal where that understanding is incomplete. A failed bearing, an awkward repair, or a customer’s unusual operating conditions can lead to a better design. As argued in The power of system design, improving a design can start with removing unnecessary parts. Competitors can study the finished product without seeing everything the manufacturer learned along the way. Reconstructing that knowledge takes time, which helps protect the business.
AI shortens development paths and accelerates progress
When used correctly, AI reduces the time needed to develop and test new designs. A competitor can use it to help write simulation software, connect engineering tools, and automate the comparison of design alternatives. The resulting improvement may come from better use of conventional engineering software as much as from the AI model itself. AI makes more ambitious development work accessible to a team that previously lacked the time or specialist capacity to attempt it. The competitor does not have to repeat the manufacturer’s decades-long learning process. It can explore different designs in simulation, discard unsuccessful approaches earlier, and find shorter routes to comparable or better machines. AI also helps automate the work along those routes, accelerating the pace at which the team can investigate and improve a design. This reduces the advantage built through decades of experience. Commitment to existing designs can even become a disadvantage. Engineers may spend their time refining a familiar machine while a competitor explores a different way to solve the customer’s problem.
Today’s lead in quality needs continued development to remain a lead. The risk extends beyond losing individual orders. A competitor could take over specifying and designing the machine, then contract out production. An established manufacturer might keep building machines while losing the customer relationship and influence over the design. Keeping its factory busy would then mean retaining a smaller part of the business.
How can established companies use AI to outperform competitors?
Established manufacturers can use the same capabilities to develop better offers sooner and win more customers. Experience may help with that work, but it should not determine which designs a company is willing to explore. Funding AI as baseline infrastructure gives people access to the tools. With that access in place, the task is to organize work so AI helps the company outperform competitors and grow. Across design, production, and service, this includes developing knowledge the company does not yet possess.
A manufacturer could serve customers whose operating conditions were previously too expensive to accommodate. Its service history could also help it develop a maintenance service for customers who currently buy only the machine. Those offers would still need customers willing to pay enough to make them worth providing. Entirely new products may require expertise that lies outside the company’s history. Pursuing those opportunities means being willing to learn unfamiliar approaches and reconsider what the company should sell.
Build an environment that can do the work
Build a company-specific environment in which AI can carry out work and develop new expertise through research and experiments. This surrounding system is the harness: the information, tools, instructions, and checks that let AI models carry out useful work. For the manufacturer, it would connect a design to the customer requirement behind it, the tests it passed, and the service problems it later caused. It would give AI access to engineering tools and make clear which changes it may attempt and which require review. The same environment should support investigating approaches that begin with the customer’s problem without assuming the next product must resemble the current one. Pushing AI integration covers the mechanics of connecting tools and instructions. The purpose is to develop and apply expertise fast enough to create better products and new services ahead of competitors.
Consider a manufacturer investigating how a customer could achieve higher output. An AI system could compare earlier designs, investigate service failures, propose changes, and run permitted simulations. It could check whether the proposed parts are available and whether technicians can reach them during maintenance. Engineers would review the evidence, resolve uncertain assumptions, and decide which alternatives deserve physical tests. That investigation should also consider different ways to achieve the output, including approaches that require a new machine. AI can help research those approaches and design experiments that develop expertise the company does not yet have. A successful alternative may take sales away from the existing machine. The company may choose to cannibalize its own product because the replacement gives customers a better solution.
Keeping design within the company does not require a person to perform every calculation or draw every part. It requires the organization to retain the ability to specify a design, judge it, and improve it using what happens in production and service. Some analysis and coordination previously performed by employees could be carried out by AI supplied on demand. The company would still need to direct that work and retain the knowledge gained from it. That capability gives it a basis for defending existing business and taking on work it previously could not offer profitably. Building this environment is continuing operational work. An experienced engineer may need to explain why a promising design was abandoned. A service team may need a way to feed recurring failures back into development. Someone has to maintain the simulation tools, check their assumptions, and evaluate whether a replacement AI model performs the work reliably. Organizations have to give people time and responsibility for these tasks, alongside delivering today’s orders.
The company can buy access to models without training its own. Its accumulated knowledge and the means to apply it still need to remain usable when a supplier changes. That includes retaining design decisions, test evidence, and the connection to customer outcomes. Otherwise, the company risks becoming dependent on a service it cannot adequately evaluate or replace. The environment should make rented AI useful while preserving the organization’s ability to steer the work. As more work can be delegated, more human attention can go toward choosing worthwhile problems and improving the conditions in which AI tackles them. Better machines and shorter development cycles should help the company win orders, reach new customers, and provide services worth paying for. Developing new expertise may take it into products its past would never have suggested. Existing designs should compete with new alternatives on what they can offer customers, regardless of how much has already been invested in them. Past spending that cannot be recovered is a sunk cost, and it should not justify choosing an existing design over a better alternative. Success can include creating products that replace the company’s current range.