Most engineering decision-makers expect artificial intelligence to become embedded in core design and engineering software, but trust in its output remains much lower, according to IoT Analytics.
The findings are based on a survey of 120 manufacturers and point to a divide between expected adoption and practical reliance.
IoT Analytics found that 87% of engineering decision-makers expect AI to be built into core design and engineering tools, while 86% said the main benefit would be reducing manual effort in repetitive engineering tasks.
The strongest expected value was in simulation-related work. According to the survey, 78% of design and engineering decision-makers expect critical or high value over the next two to three years in simulation preprocessing and simulation results interpretation.
Other areas also ranked highly. The survey found that 69% of respondents see critical or high value in large language model-generated documentation and requirements generation, while 67% said the same for predictive design optimisation and code generation.
Trust gap
Despite those expectations, confidence in AI-generated output remains limited. Only 24% of respondents reported high or full trust in AI for generating documentation or specifications, the highest trust level recorded across the activities covered by the survey.
At the most confident end of the scale, full reliance was rare. "Fully trust and rely on" did not exceed 3% for any activity examined, IoT Analytics said.
Trust was lower still in simulation work, even though respondents identified it as the area with the greatest potential value. Trust in AI-generated simulation stood at 17%.
The contrast suggests manufacturers are more comfortable using AI to support preparatory or administrative work than to make or validate technical judgements. In practice, the data points to caution about handing over decisions that affect product behaviour, design accuracy or engineering verification.
Changing products
IoT Analytics linked that caution to wider changes in engineering itself. As products incorporate more software and receive continuous updates after shipment, design processes are shifting towards behaviour, validation and ongoing performance rather than physical form alone.
Harsha Anand, Senior Analyst at IoT Analytics, said changes in products are reshaping the discipline and increasing the central role of simulation.
"The fundamental reason design and engineering is being reinvented is that the products themselves have changed. Products are now software-defined, continuously updated after they ship, and increasingly intelligent. When a product is defined by how it behaves rather than how it is drawn, the engineering question shifts from 'can we build it?' to 'how will it behave?' and that puts simulation at the center of the design and engineering discipline.
"Our research shows simulation is both the fastest-growing software category alongside cloud, and the single activity where manufacturers expect AI to add the most value (78%). But the research also surfaces a paradox: engineers are far more willing to let AI prepare a simulation than to trust its output, with trust in AI-generated simulation at just 17%. The manufacturers and vendors that close that gap between simulation ambition and simulation trust will lead the next cycle of engineering software," Anand said.
The figures indicate that the commercial opportunity for software suppliers may depend less on adding AI features than on persuading engineers that outputs can be checked and defended. In regulated or safety-sensitive design environments, that issue is likely to carry particular weight, because engineers remain accountable for final decisions even when software assists with analysis.
Vendor challenge
Knud Lasse Lueth, Chief Executive Officer at IoT Analytics, said adoption is no longer the central question.
"Our recent design and engineering survey shows that the question has shifted: it is no longer whether AI will be adopted in engineering, but whether engineers will trust it enough to act on its output. I believe the vendors that close this trust gap, for example by making AI outputs auditable and verifiable, will see faster adoption of their AI capabilities," Lueth said.
The results suggest vendors in design, simulation and engineering software face a two-part task: proving that AI can save time in routine work, and showing that its output is transparent enough for engineers to use in decisions. With only a small minority prepared to fully rely on AI in any tested activity, acceptance of the technology remains conditional rather than complete.