How to harness AI to transform development and engineering

Some organisations expect to double growth and productivity and even triple profitability by 2030. But what is often ignored is that the nature of development and engineering itself is also changing rapidly. Merely plugging in AI tools to existing ways of working is unlikely to deliver the promised impact.

In a new study, Arthur D Little, working with partners NAE (Netherlands Academy of Engineering), IVA (Royal Swedish Academy of Engineering Sciences), and KIVI (Koninklijk Instutuut van Ingenieurs), uses evidence from over 900 AI and technology case studies, multiple hands-on projects, as well as survey feedback from 95 respondents across six industries to show how companies can best tackle this critical challenge. And the key finding is that people, not just systems and technologies, are key.

While there are already many specific examples of AI and other complementary technologies delivering major benefits in tasks like design, planning, and problem solving, levels of adoption are still surprisingly low. Generally, only around 15% of companies report mastery in applying AI to development and engineering, with just over half having merely initiated pilot projects.

The path forward toward AI maturity requires addressing the most impactful challenges, including developing capabilities, encouraging appropriate mindsets, and enhancing trust in AI's reliability, explainability, and security. These challenges are predominantly people-based. Solving them requires going back to basics on how developers and engineers create value: both through knowledge expansion and, increasingly, through knowledge integration.

AI can be especially powerful in enhancing engineers' capabilities in the knowledge integration dimension, for example, through knowledge management, project/portfolio management, scouting, intelligence, and collaboration tools. At the same time, via the knowledge expansion dimension, AI tools can also be extremely valuable in critical engineering disciplines, such as simulation, modelling, analytics, and problem solving.

Moving beyond AI pilots to full integration requires an approach that is people-centric. It is based on the concept of a balanced portfolio, something very familiar to engineers and innovators.

First starters, important baseline moves on AI integration need to be prioritised based on systematic consideration of user needs. Typically, these involve improving productivity, efficiency, collaboration, and learning, especially in the knowledge integration dimension.

User needs need to be analysed by considering personas, jobs to be done, pain points, and gain points. There will only be wide adoption of AI applications if they provide gain and/or reduce pain for user personas, just like other digital applications. The study identified 17 discrete user personas, 124 jobs to be done, and more than 200 gain and pain points, helping to identify over 900 corresponding AI use cases from 3,500 solution providers. From these, 50 were identified that were applicable for every development and engineering user persona. Heat maps can help individual organisations prioritise use case relevance by user persona to suit their specific needs.

But there are also other AI use cases that are important for key must-win battles; in other words, gaining competitive advantage as part of the technology strategy. These are especially relevant for the more specialist knowledge expansion dimension, which are less reliant on widespread user adoption for their impact.

Analysis of competitive strength versus competitive impact can help prioritise these AI use cases. This may also include less mature AI technologies that have high potential, such as multimodal models, graph neural networks, reinforcement learning, and others.

Identifying a balanced portfolio of AI use cases is the starting point, but merely plugging in new AI tools to the existing organisation is unlikely to deliver the anticipated benefits. Development and engineering functions must transform themselves toward an AI-enabled future model.

The report illustrates a four-step transformation approach to empower people through adoption of AI (democratisation), renewing ways of working (ambidexterity), leveraging internal and external data (data collaboration), and making essential changes to the technology infrastructure (enablement).

For development and engineering organizations, doing nothing is not an option given the huge value at stake. And while doing something, like conducting AI pilot trials, is not terribly difficult, the real benefit will come from AI implementation that is fully scalable, widely adopted, and properly aligned with technology strategy. A people-centric approach that fundamentally re-examines ways of working is the best way to ensure that the potential rewards of AI integration can be realised.