Having spent most of my career working with procurement and development in technology organizations, I have often pondered why some, particularly larger and more complex, tech organizations often seem so inefficient.

According to a January 2022 study, the average developer spends less than an hour a day actually writing code.

According to this 2019 study, the average developer spends less than 50% of their time actually programming (which, besides coding, can include debugging, testing, and other related activities).

The time a developer spends coding or the efficiency of a product or technology organization is individual and depends on many factors such as technical debt, team, organizational structure, and culture.

However, it is clear that there is great potential for complex product and technology organizations to become more efficient. This is not only a value for the company as a whole but also means that employees can spend more time doing what they enjoy, developing products and services that create value for customers and society.

AI has already particularly helped inexperienced developers improve the pace and quality of their work, and with new AI-based tools like Devin and Github Copilot Workspace, AI will become a larger and more important part of a product organization’s everyday life.

So what is required for AI to really make a breakthrough in larger technology and product organizations? Below are my thoughts on what needs to be in place.

Better AI Models

Not just language models, but image and multimodal models also need to be significantly better than they are at present. Today’s models can help individual employees with various specific tasks, but to really create transformative improvements in a business, they need to be able to handle much larger and more complex data and information volumes than today’s models can.

As of the writing of this article, Gemini 1.5 has a context window of about 40MB. However, my experience is that models show signs of “confusion” at just a fraction of this. Even with techniques like RAG (Retrieval Augmented Generation), a 40MB memory is not enough to understand the complexity of a multinational company.

New User Interfaces

Today’s AI tools primarily use “prompting.” A user writes a question (in text) and receives a response back (mostly in text).

The key to unlocking the potential of AI in organizations will be the interaction between AI and humans, where users and teams work together with AI to facilitate and streamline work. This requires entirely new types of user interfaces where AI and users give feedback to each other to create trustworthy deliveries.

The next generation of AI-centered tools will not only answer questions or produce documents and other content. These tools will support the actual coordination of work, thereby reducing the need for coordination among employees. This places entirely new demands on user interfaces and communication between AI and users.

Accelerated Work Methods

The automation of entire work methods and processes will not happen overnight. Instead, it will happen gradually, just like with self-driving cars. From point-by-point automation here and there to whole parts of work methods that can occur with only human supervision and management.

As this gradual automation progresses, the time from idea to finished product will decrease, and product sprints will go from months to weeks to days (and perhaps hours and minutes in some cases). This implies entirely new hyper-agile work methods, where the winner is the one who understands their customer the fastest rather than the one who produces the most content.

Smaller Teams

As “white collar” work methods are automated, the need for man-hours decreases. This means that teams will become smaller, which in turn means less need for coordination (product meetings, requirement meetings, coordinations, etc.).

This reduced need for coordination becomes the true driver of efficiency as a result of AI. A consequent effect of this is that senior and experienced employees, instead of leading and coordinating more junior employees, can perform concrete and value-creating work, which will further increase the efficiency of teams and companies.

New Leadership

As with all other transformative changes, the biggest challenge will be leading the development of work methods, organization, and new tools that come with AI - building teams and cultures that not only tolerate change but also embrace it.

Having worked concretely with the development of AI tools and also having spent the majority of my career helping organizations through change and transformation, I am convinced that AI will radically change how companies work with technology and product development - but it will take time.

It will not be the technology that is the limitation, but our human capacity for change, both individually and together.