Why Design Intent is More Critical Now Than Ever

What’s the big deal? Behind every design decision there’s a design intent and, sometimes, it gets lost in translation. Yet it matters because it captures the “why” behind every engineering choice. It explains not just what changed, but the reasoning, trade-offs and goals that drove the change. It provides the vital context that is missing from version histories or design records alone, helping others propose alternatives that match the original goal.

Product requirements may set market expectations, but design intent is what an engineer uses to explain why they made particular choices based on those needs. Clarity on design intent means that other engineers working on the product further down the line can either maintain these choices or consciously update them, helping protect product quality.

Capturing design intent also preserves institutional knowledge, enabling new or future team members to understand why certain decisions were made. This prevents reinventing the wheel or repeating mistakes. It is particularly key when taking into account the emerging skills gap in the engineering industry. According to the Bureau of Labor Statistics, the share of advanced manufacturing employees over 55 has more than doubled in the past three decades. This trend means that decades of design expertise are at risk.

By valuing design intent, engineering leaders protect product quality, avoid repeated mistakes, accelerate decision-making and enhance their company’s ability to deliver products that truly meet user needs. And as engineering teams increasingly embrace AI—agentic AI in particular—the importance of documenting and applying design intent in every design decision is accelerating.

The prevalence of AI, and its support in decision-making, makes the task of preserving the rationale behind each decision more critical to ensure outcomes align with engineering principles and organizational goals.

Design Intent in an AI World

Yet the rise in AI experimentation does not translate into mature adoption, and the real test is how deeply companies embed it into decision-making. According to McKinsey, 79% of companies say that they’re specifically using generative AI, which demonstrates that AI adoption is widespread at a surface level. McKinsey research also reveals that less than 10% of companies have fully scaled AI agents into any business function, and this number is only 6% in product development.

Agentic AI has the potential to transform product design when developed and deployed thoughtfully. It can remove bottlenecks, increase the scope of exploration and innovation, and empower engineers to focus on higher-value strategic work. But truly scaling agentic AI in the sector and embedding AI agents into design review workflows requires significant effort. Design intent is at the center of that effort.

READ MORE: Physical AI in Motion: How Machine Learning Drives Next-Gen Industrial Automation

AI agents depend heavily on historical design context. They need this information to accurately review existing designs and then recommend and generate meaningful design variants. When design intent is clearly documented and easy to access, these agents can draw on past insights, lessons learned and the rationale behind previous decisions to ensure their suggestions are technically sound, relevant and useful.

Conversely, if an AI agent does not have access to design intent, there is little to prevent it from drifting away from original goals, which dilutes the relevance and usefulness of its suggestions. Without the context from previous design decisions, the agent cannot learn from past mistakes or capitalize on best practices, ultimately putting product quality, safety and functionality at risk.

What is Holding Agentic AI Back?

Taking a step back from design intent for a second—the gap outlined in McKinsey’s research between the adoption levels of generative and agentic AI can be explained by the complicated processes and the preparation needed to guarantee that AI agents work.

To begin with, there is a high bar for deployment. For AI agents to work autonomously, they need well-defined guardrails and clear handoff points between engineers and agents.

Close attention is key to ensuring data access and control parameters are configured correctly. AI agents need access to the right data, but must be restricted from unnecessary or sensitive information. This demands precise data governance. Engineering leaders are also concerned about their organization’s data readiness and recognize that AI won’t help unless their data is well-organized and accessible.

READ MORE: Q&A: How Contextualized Data and AI Agents Enhance Manufacturing Automation

It’s also important to remember that successfully scaling agentic AI involves not just technology but significant changes to business processes, decision-making and organizational buy-in. Tools like Product Lifecycle Management (PLM) or Product Data Management (PDM) systems are heavily embedded into traditional design review processes and are often used as the primary “base” to start workflows.

But while these tools are essential for product design, they are not built to scale AI integration and have substantial limitations. For example, PLMs and PDMs are good at recording what changed and when, but they are not designed to capture the reasoning and trade-offs behind those changes in a way that’s usable for AI.

The concept of design intent becomes even more challenging when unstructured data enters the fray—things like CAD review comments, ad-hoc annotations and digital sticky notes where much of the reasoning often lives. To enable AI agents to effectively support design reviews, the core software platform must be able to capture and structure design intent from both formal reviews and everyday decisions. It also needs to support seamless integration with other software tools across the entire ecosystem.

The final obstacles to successfully adopting agentic AI at scale are reliability and trust. To overcome caution about handing over workflow autonomy to software, AI outputs must be proven trustworthy. While some of that confidence will build with time and repeated success, giving AI agents the richest data set to learn from can help expedite the process and earn user trust more quickly.

How Can Teams Successfully Integrate Agentic AI?

Adopting and integrating agentic AI can feel like a huge task with numerous risks, but it doesn’t have to be this way at all. Engineering leaders can start small by identifying a high-impact use case and choosing a specific business process or engineering workflow where AI could provide clear value—for example, automating drawing reviews or running simulations.

Once the agent is deployed across the chosen workflow, the team can monitor results, gather user feedback and refine the process so that it is as reliable and valuable as possible. Once this use case has succeeded, it will be easier to get wider buy-in; leaders can apply lessons learned to additional workflows or teams, thereby scaling AI adoption step by step.

READ MORE: Leo AI: How CAD-Aware AI is Changing Mechanical Design and Engineering Workflows

While there’s plenty of debate around how AI will transform engineering teams, rest assured AI agents are unlikely to take the place of skilled engineers. Large language models (LLMs) are only 90% accurate at tasks that take a human roughly four minutes, and their performance drops as tasks become longer and more complex. And there will always be essential decisions and trade-offs where human expertise is vital.

But with the right system-level prompts and data access, AI agents can support drawing reviews across different formats, from informal peer checks to formal reviews. While these reviews vary in rigor, many include repeatable checks that consume significant engineering time. AI can assist with this work, helping engineers focus on high-leverage decisions that require experience and judgment.

What does design intent have to do with this? A lot. Without comprehensive access to structured and unstructured design intent, engineering teams won’t be able to unlock the potential of AI.

Design intent bridges the gap between raw data and human reasoning, enabling AI to be truly useful and reliable in engineering environments. Design intent provides the context AI agents need to interpret data accurately, make relevant recommendations and support sound engineering decisions.