Building AI teams for today, developing the skills we need for tomorrow
Why small teams suit AI prototyping, and why they must make room for people earlier in their careers.
As the AI Capability and Enablement team works with AI use cases, one lesson is becoming increasingly clear: the way we compose teams for AI prototyping matters.
AI prototyping is different from taking an established digital service through delivery.
In the early stages, the priority is to validate the problem, understand what users need and quickly test whether AI can provide an appropriate, useful and viable solution. That changes both the shape of the team and the way it needs to work.
Start small, but bring the right skills together
For an AI prototype, we have found that smaller, multidisciplinary teams can work particularly well.
A core team might include contributions from a business analyst, service designer, delivery manager, lead architect supporting a data scientist and software developer focused on building and iterating the prototype.
The goal isn't to remove disciplines from delivery. It's about bringing in the right capability at the right point.
At the beginning, the team needs enough expertise to validate the problem and develop a rough service design that can be tested with users. From there, a small team can rapidly prototype, gather regular user feedback and continue iterating towards something that demonstrates whether the idea has value and could be taken forward.
When the work moves closer to production, the team can broaden. A fuller digital delivery and user-centred design capability becomes increasingly important as the service matures.
Why smaller teams can move faster
One of our biggest lessons has been that smaller can be better during prototyping.
Bringing a full digital delivery team together too early can add complexity before we know whether the solution is right. Small teams have fewer hand-offs and can respond quickly as we learn.
But there is an important condition: small teams need people who are comfortable working beyond narrow role boundaries.
AI prototyping particularly benefits from people with broad, T-shaped skills who can contribute their specialist expertise while adapting to what the team needs. Traditional role boundaries can become less rigid because the focus is on learning and iteration rather than following an established delivery path.
We've also seen value in bringing engineering and user research very close together, sometimes called forward deployed engineering. Changes can be implemented quickly, with less opportunity for requirements to become lost or misinterpreted between disciplines.
For a focused thin slice of a service, this can be an extremely effective way to work.
But there is a challenge we need to address
There is a risk hidden within this model.
If our small AI teams depend only on people who already have broad experience, how do the people earlier in their careers ever gain that experience?
Junior colleagues may understandably have developed deeper experience within a particular professional role but had fewer opportunities to work across disciplines, rapidly prototype or use AI tooling as part of multidisciplinary delivery.
If we only select people who already have that breadth, we risk creating a cycle where experienced people gain even more AI experience while those earlier in their careers struggle to get onto the first rung.
That isn't sustainable for AI enablement.
AI enablement needs to be a development journey
We therefore need to think about team composition as more than simply finding the smallest team capable of delivering a prototype.
Creating opportunities for more junior colleagues to participate is key. That could mean bringing a junior developer, business analyst, user researcher or content designer into a prototype team alongside more experienced colleagues.
The opportunity is to learn how disciplines connect, gain hands-on experience with AI tooling, participate in rapid prototyping and gradually develop the broader capabilities that future AI teams will require.
That creates a different model of AI enablement: we build capability while we build prototypes.
This doesn't mean everyone needs to become an expert in everything. Professional disciplines still matter. Deep expertise still matters. But AI delivery increasingly rewards people who understand their own discipline while also being able to work effectively across others.
For the organisation, it broadens the pool of people who can confidently contribute to future AI work.
Building the next generation of AI capability
Ultimately, successful AI enablement isn't just about delivering individual use cases.
It's about building an organisation that has the skills and confidence to keep doing this work. Combine pace with capability building, having experienced and junior staff working together.
That's how we move from delivering AI prototypes to building sustainable AI capability.