MIT Technology Review: 90% of engineering firms will raise AI spending, but without an arms race

AI increasingly affects not only software but also things that work in the physical world: cars, home appliances and medical devices. A study by MIT Technology Review Insights and L&T Technology Services shows that engineers are ready to expand their use of AI, but they follow the logic of risk, verification and measurable results, not the logic of hype.

Why the cost of error is high

In product engineering, a model's mistake costs something different than in a digital service. A bad chatbot answer can be fixed with an update, but a defect in a design, an embedded system or a manufacturing decision can lead to a batch recall, an accident or trouble with a regulator.

So teams that bring AI into the design and production of physical products are not ready to trust general-purpose models as they are. They build processes with different levels of trust, mandatory verification and clear human accountability.

The study's authors stress that what matters here is not demo magic but first-time-right: the ability to get a correct result on the first attempt, or as close to it as possible. For companies that make cars, electronics or medical devices, this metric matters more than loud claims about transformation.

That is why AI adoption happens in layers: first in areas where the effect can be verified, then deeper into the product lifecycle. This is how companies reduce the risk of costly mistakes and build trust in new tools.

"Where an AI result affects a physical system and cannot be rolled

back, reliability and measurability take priority."

Where the budgets are going

The survey covered 300 product engineering, development and technology leaders in the US. All of them represent companies with revenue above $500 million across 16 industries, and the interviews were supplemented with the views of senior executives and industry experts.

Nine in ten respondents plan to increase AI investment in the next one to two years, but they do not plan a sharp leap. 45% expect spending to grow by up to 25%, about a third by 26–50%, and only 15% are ready to raise the budget by 51–100% at once.

Priorities are pragmatic too: the leaders are not smart assistants for their own sake, but tools that are easier to verify, defend before a regulator and tie to ROI. Analytics, simulation and validation come first, areas where engineers have clear feedback and historical data. Such tasks are easier to audit, approve and defend to the business.

The study identifies several areas around which most demand is currently built:

• predictive analytics for early detection of defects and weak points
• AI simulation and validation before production launch
• multi-level verification of models and results
• specialized, auditable tools instead of general-purpose systems

What is changing in teams

One finding concerns people. 73% of leaders expect AI to take over routine engineering work. This does not mean specialists become less important; the center of gravity itself shifts. Inside companies, value moves from manual execution of repeatable operations to architectural decisions, systems thinking and strategic assessment of trade-offs. The more operational work moves into tools, the more important the people who understand a model's limits and are responsible for the final choice.

At the same time, external partners and specialized vendors play a growing role. If part of the execution moves to third-party ecosystems, ownership of core logic, data and decision rules becomes a question of control.

The authors note another shift: companies measure success by more than time to market. Product quality and sustainability rank higher among goals, since customers, investors and regulators can see them. Cost reduction and team satisfaction fall lower in the priorities.

What this means

For the market, this is a signal: AI is entering physical-world engineering not through flashy demos but through verifiable, narrow use cases with clear ROI. The winners will not be those who promise a revolution the loudest, but those who embed AI fastest into simulation, quality control and engineering decisions without losing trust, safety and accountability.

This cautious model is likely to become the main template for adoption in the coming years.