A survey of 350 engineering leaders in organisations with 1,000 employees or more has revealed a widening gap in AI adoption, according to the ‘State of Engineering Report’ from SimScale
Of those surveyed, 9% have a mature, scaling AI program in place, 80% are working on limited deployment pilots, 8% are planning to start this year, and 3% have no plans to start at all.
The survey results show that many companies that planned to start AI projects last year (44% in 2025) have now taken the next step and started pilots (80% in 2026). The speed of progression from a pilot program to large-scale deployment tells a different story, with 7% of organisations having mature, scalable AI programs in 2025, rising by only 29% in 2026 to 9%.
The speed at which engineering organisations are scaling AI into workflows is also a key differentiator. According to the survey, the average deployment timeframe from pilot to mature AI is eight months, but 55% reported it was taking between seven and 12 months, and 9% over a year. The fastest-moving organisations achieved a timeline of three to six months.
David Heiny, co-founder and CEO at SimScale, said, “Moving from AI pilots to scaled deployment is still one of the most critical challenges for engineering organisations right now. The fastest moving teams are moving from pilot to scaled projects as quickly as three months, but many are lagging behind, creating another big gap.”
Despite delays in program progression, AI is playing a meaningful part in engineering workflows. The survey revealed that 36% of design and simulation projects conducted in the last 12 months used AI or agentic engineering methods, but it has not become universal across all projects.
The survey found three primary blockers to faster and more comprehensive scale-up:
- Data preparation and availability for AI (74%)
- Governance and compliance concerns (48%)
- Software interoperability challenges (42%)
Many engineering AI applications, including agentic assistance, workflow automation, and design exploration support, can begin delivering value with far less data preparation than teams initially believe. Governance is also becoming less of a concern, with 87% of respondents saying their organisations permit AI to make pass or fail decisions at design gates.
Heiny said, “The fact that over a third of design and simulation projects are using AI demonstrates a tipping point, but we need AI to transition from being a special project to becoming the organisational norm, and this means addressing the broader challenges and perceived blockers. When it comes to software interoperability challenges, while modern cloud-based infrastructure is critical, the primary bottleneck in engineering is the workflow itself, not a lack of computational power. SimScale has built the architecture that makes AI actually work inside engineering workflows.”
After running a successful AI pilot, Convion, a clean-technology company, has established a new standard for Physics AI-driven research and development. By making simulation insights immediately accessible through validated AI models, released as internal tools, the engineering team can explore changes interactively without direct reliance on computationally intensive solvers.
RLE International, a global engineering and consulting firm, also used SimScale to develop an end-to-end AI prediction workflow for automotive computational fluid dynamics (CFD) that can predict vehicle aerodynamics in seconds.
One area where there is almost full unanimity is in the return-on-investment (ROI) opportunity. 99% of respondents say they are confident their organisations will realise meaningful business value from AI or agentic engineering in the next 12 months.
This is already evident from the results being seen by first adopters. Where conventional simulation requests take up to 17 hours, whole AI workflows are being completed within six. Introducing AI into workflows can increase simulation iteration speed from as low as once a week to multiple times a day, and increase the number of design variants from as low as 1-5 to more than 200.
Heiny added, “The teams using Physics AI and agentic AI in engineering workflows are innovating faster as they’re able to explore thousands of ideas in seconds. This is not just about accelerating time-to-market, boosting revenues and reducing risks – although it does all that – it’s about a shift from human-paced iteration to machine-paced iteration, fundamentally changing how physical products are invented. If companies aren’t using these tools – they’ll be left behind.”
