Artificial intelligence (AI) in engineering is delivering hugely positive results. Low- and no-code tools are democratizing AI capabilities that were once solely the domain of specialist data scientists. However, not everyone in engineering is reaping the benefits just yet. Getting productive with the new wave of accessible AI tools is easy. Engineers tell us their toughest AI challenges lie elsewhere. 

That’s why, in an earlier article, we shared practical tips for dealing with one of the most common bottlenecks: identifying viable AI use cases. Here, we’re examining five more hurdles you’re likely to encounter on your AI journey.  

Obstacle #1: Securing Buy-In from Business Leaders

AI in engineering isn’t new. However, as a recent Forrester report highlights, the emergence of generative AI (genAI) – specifically the public launch of ChatGPT 3.5 in late 2022 – reshaped how business leaders perceive AI. 

For many executives, AI has gone from a niche IT project to an essential corporate asset. AI strategies – and accompanying AI governance practices – have become must-haves.

This shift has significant implications for engineers looking to secure backing for new AI projects. Above all, project leaders need to demonstrate not only what their innovative AI solution will be used for, but also the impact it’ll have on things such as revenue, profitability, and costs. 

Starting out, it’s best to tackle realistic, quick win-type AI projects rather than ambitious, idealistic undertakings. Forrester’s report emphasizes the value of a “flywheel” business case, where modest initial gains snowball as an AI project’s impact gradually but steadily increases.   

Obstacle #2: Leaning into AI Governance

AI governance is often seen as a barrier to innovation. But rather than treating your organization’s AI governance framework as another layer of tedious bureaucracy, we encourage you to lean into it early. Doing so will help overcome one of the most crushing internal obstacles to AI success: lack of trust.  

Engineers aren’t the only people who need to trust AI models’ results – all stakeholders must have that trust. Winning support and building trust in AI throughout the organization is a golden thread that runs through any successful AI-powered engineering use case.  

Although regulations such as the European Union’s AI Act are starting to come into effect, individual organizations are proving quicker than lawmakers at implementing governance structures that address areas such as transparency, trust, accountability, and informed decision-making. On a more practical level, these frameworks typically include processes for risk assessment, model life cycle management, AI system auditing and monitoring, and compliance management. Engaging with these requirements early will go a long way in securing support for new AI use cases. 

Obstacle #3: Finding Trustworthy Data and Trusting its Results

Trust is massive issue when dealing with data at a more granular level. That’s often because many engineers feel overwhelmed by a deluge of both real-time and historical information. 

AI training data needs to be accurate, complete, free from bias, and reflective of its intended deployment. The key is to focus on data quality “hotspots” that will have the greatest impact on processes, outcomes, and decision-making. Engineers can minimize the risks poor-quality data poses by minimizing the target itself. Explore and protect only the information you need. Discard low-value data, which only makes it harder to identify the good stuff. Start small and evolve.

Many of the most impactful AI in engineering use cases leverage historical data from previous physical tests. In other words, it’s the data that engineers have been using and trusting for product development for years. Also, keep in mind that raw data is rarely complete, clean, and accurate. What’s more, data quality is an on-going process, not a one-time exercise. Project teams need to commit for the long term.  

Obstacle #4: Bringing a Project Team Together

Not everyone will welcome new AI initiatives. Much of the current debate around AI focuses on its potential to replace people. Things like agentic AI are only likely to further fuel such concerns. Even when AI doesn’t impact headcounts, it’s often associated with changes to existing roles and responsibilities. That’s likely to cause friction.  

AI trends such as democratization and “bring your own AI” don’t preclude the need for teambuilding. Don’t ghost your data scientists and tech teams; engineers need their support, particularly in areas such as tackling bias and complying with governance frameworks and regulations. AI in engineering often involves automating and synthesizing processes such as physical testing. Employees and customers impacted by such changes need to be on board. That’s why embracing governance is so important. 

Resistance to change isn’t unique to AI – it’s a timeless future of human nature. But one AI-specific problem is the fact that many people still can’t explain what AI is or what it does. As a starting point to building a happy AI team, project leaders should be able to communicate how a model functions, its strengths and weaknesses, likely behavior, and any potential biases and risks. Does all the data science and AI jargon confuse you? No worries – we’ve put together a quick guide to data and AI terms here. 

Obstacle #5: Finding the Skills and Resources Needed to Make your Project Work

Building a project team for AI in engineering may expose skills gaps and human resource shortfalls. According to Forrester’s Q2 AI Pulse Survey, developing skills is a top concern among AI decision-makers. A strong corporate AI strategy will address the need for ongoing professional development and recruitment. More immediately, engineers can also reach out for external help. 

Looking to get a comprehensive overview of your AI readiness? Book a complimentary data and AI assessment today.