AI is accelerating individuals 
When writing about software development, there is currently no bigger elephant in the room than the impact of generative AI. AI is helping individuals analyse, code and learn faster than ever before. Yet when I look back on twenty years of software delivery, I saw few digitalisation projects that failed due to a lack of individual expertise. More often, they failed because organisations struggled to align around what should happen next.
This makes me wonder whether we are paying enough attention to collective learning, not just individual acceleration.
Twenty years of improving collective learning 
Over the past twenty years, we’ve seen a number of big changes affecting the IT sector: cloud computing, mobile development, NoSQL data storage… But the most impactful changes have not primarily been technological. They have been about helping people learn and deliver together more effectively.
Think about some of the major shifts:
- Agile shortened feedback loops between teams and stakeholders. Instead of spending literally years documenting and implementing assumptions, teams learn every few weeks whether they are solving the right problem.
- DevOps recognises that no organisational learning can happen when the people building the software are separated from the people running it. Bringing them together around shared ownership turns production incidents, metrics and user feedback into opportunities for collective improvement rather than blame and handovers.
- UX research and service design challenge the idea that user needs can be understood by debating assumptions in a meeting room. Research allows teams to (in)validate assumptions before investing in costly development projects.
- Organisational frameworks like Team Topologies and toolkits like Sociocracy 3.0 gain traction because they focus on structuring organisations around people’s interactions. To build organisations that balance adaptability and effectiveness, we need to explicitly factor in all aspects of human collaboration.
Different practices, different contexts, but a common theme: improving collective learning.
An early lesson in the public sector
In 2005, fresh out of school, I joined a software development team responsible for the declaration of social contributions in Belgium. I assumed software development would mostly be about designing data models and writing code. Instead, I found myself in a team of more than thirty people delivering a new release every quarter. The solution involved systems written in COBOL, others in Java, and a plethora of systems to integrate with and external parties to coordinate.
What I remember most from that period is the effort invested in creating a shared understanding of the work. Every release cycle started with a large kick-off meeting. Teams came together to understand what we were trying to achieve, what had changed and where collaboration would be needed. The guiding artefact during these meetings was a massive Word document that tracked decisions and centralised knowledge.
I am sure those meetings would make some modern agilists cringe. But they created alignment, and that alignment made the rest possible.
One lesson became obvious very quickly: the hardest part was not the software. It was helping people understand and coordinate with each other.
Why human friction matters 
Human collaboration is messy and often comes with what we would call gedoe in Dutch. Stakeholders disagree. Team members don’t speak up. Our best intentions get lost in translation. Decisions take longer than expected. These moments can feel inefficient, yet it’s there that assumptions are challenged, trade-offs become visible and collective ownership is created.
Today, generative AI allows that much of this human friction can be bypassed. Need user stories? Ask an AI. Don’t want to write test code? Ask an AI. Need insights for your user persona? Ask an AI. There is value in this. But also risk.
Becoming more self-sufficient as individuals reduces opportunities for collective learning. AI can generate perspectives. It can challenge assumptions. It can simulate stakeholders.
But it cannot create commitment. It cannot build trust between teams. It cannot take responsibility for difficult decisions.
The next challenge
In our Public Service Adventure podcast, Alixe and I regularly speak with people working on public services.
What strikes me time and again is that their greatest contribution is rarely their individual brilliance. It is their ability to help people, teams and organisations find a way forward together.
Looking back on the past twenty years, many of our most successful practices have improved how groups learn together. Generative AI is improving how we learn faster as individuals.
Perhaps the next challenge is making sure we do not lose the first while embracing the second.
