A guide to training software development teams for the AI Era
The role of the software developer is changing again and will continue to change as AI evolves.
What's for sure is the skills of critical thinking and developing knowledge will not become obsolete, but what is happening is that how a developer uses these skills is flexing and changing to accommodate the biggest shift in software development in the last 50 years. In this guide, we look at how CTOs should be looking at supporting and training developers against the AI landscape. This view is built from hands-on experience, AI consulting and working to deliver AI-enabled software development across multiple sectors and seeing day in and day out the challenges and opportunities that AI is bringing to the developer world.
As AI accelerates implementation, the developers who create the most value are no longer the ones who write code fastest - they are the ones who understand the problem, the system, and the consequences of what they build.
This shift is taking the industry back to a proven model: The analyst programmer.
Not as a nostalgic concept - but as the future shape of high-performing engineering teams.
What’s in this guide
- What changed in software teams - How AI is reshaping traditional engineering roles
- Why the existing junior developer model is under pressure - Why faster delivery risks a weaker understanding
- The development bottleneck has moved - Why deciding what to build now matters more than building it
- What systems thinking means in practice - The questions developers must ask before writing code
- The analyst programmer returns - Why breadth of understanding is becoming the defining skill
- 6 Bespoke AI training needs to be developed now - How teams must adapt to develop effective engineers
- 5 Coreskills for 2026 - The capabilities that will define high-performing teams
- How Propel Tech can help - Supporting teams to adopt AI without losing engineering discipline
What changed in software teams
Over the past two decades, software engineering has evolved toward increasing specialisation: frontend, backend, platform, DevOps, and SRE. Bespoke and off-the-shelf served different markets and different needs.
That model made sense. As systems became relied upon to do more, they became more complex, and deep bespoke expertise was required to connect, build and operate systems reliably. This is still the case today - expert understanding of a business need sits at the heart of all great bespoke software and leads to a desired outcome for a business, usually efficiency or growth.
But…. AI is now compressing part of that complexity….
Implementation - once the most time-consuming part of delivery - is becoming significantly faster. As a result, the value of pure technical specialisation is being partially rebalanced by the value of understanding.
Developers who can add deeper analytical thinking and express this in simple business terms that make sense, and who really understand:
- Why is something being built?
- Who is it for?
- How does it fit into the wider system?
- What are we looking to achieve?
Are becoming more valuable than those focused only on execution, as these are the developers who can scale and put AI to work in the right way and critically understand the nuances of the implementation work AI is doing - without this, the layer of understanding that allows for best in practice work to be delivered becomes weak or altogether absent.
This is not a marginal change. It reshapes how teams should be structured, trained, and evaluated.
Why the existing junior developer model is under pressure
Traditionally, developers learned through gradual exposure, systematic and planned, starting small and gaining confidence as they learn. They built small features, fixed bugs, and absorbed system knowledge over time - through code reviews, incidents, and experience. Understanding emerged naturally because the delivery pace allowed it.
AI changes that dynamic.
A developer can now generate substantial functionality in hours or days. But the understanding that used to develop over the years does not accelerate at the same rate; our brains still need time to learn, to understand, to process, to consider and review - and this is critical to becoming a strong developer.
But…. in a world where AI executes quicker than humans can learn, this creates a gap:
- Output increases
- Understanding does not
Without intervention, teams risk producing more software, without a deep understanding, and this will ultimately fall down as the undercase ius not fully understood, and the agility that understanding delivers is the ability to change focus based on deep understanding. The risk of quick wins from AI is ultimately weaker architectural thinking, poorer decisions, and higher long-term cost - we are already seeing this in
In the AI era, learning can no longer be left to chance, or to opportunity to project iterations. It must be designed.
The development bottleneck has moved
This shift is often misunderstood as “AI writing code.” In reality, the constraint has moved upstream. If a working feature can be generated quickly, the real questions become:
- Should this exist at all?
- Is this the right approach?
- How does it behave under failure?
- What does it do to the wider system?
The bottleneck is no longer implementation.
It is judgment.
Teams that use AI purely to generate code tend to produce more defects, more rework, and more operational complexity. Teams that use AI to accelerate validated design decisions produce better outcomes, faster. The difference is systems thinking.
What does software systems thinking mean in practice
Systems thinking is not abstract, is rooted in the basics of software development which is user understanding. It is a discipline of asking better questions before writing code.
“Who is this for and what are they actually trying to achieve?”
Not what the ticket says, but what the user needs in context.
“Where does this sit in the system?”
What depends on it? What does it depend on? What breaks if it fails?
“What does 'done’ really mean?”
Is it observable, supportable, and understandable in production?
“What are the failure modes?”
What happens under load, with bad data, or when dependencies change?
The analyst programmer returns
At Propel Tech , we have always focused on understanding business and making tech possible as problem solvers. In the event of AI, the analyst programmer, the problem solver has become even more important, the developer who combines business understanding, system design, and implementation. Breadth of understanding, curiosity and problem-solving becomes the primary differentiator again.
This does not mean removing specialisation.
It means the most effective developers are those who can:
- Understand business intent
- Translate requirements into system design
- Evaluate trade-offs
- Own outcomes, not just outputs
- Deliver with a curious mind and a desire to solve human problems with technology
AI handles the mechanical translation into code. The human owns the thinking.
At Propel Tech, this has always been the expectation. Our teams are trained to understand the problem before implementing the solution, to challenge requirements, consider system impact, and design for real-world use. What has changed is that this approach is no longer optional; it is becoming the industry standard.
6 Bespoke AI training needs to be developed now:
Teach context before implementation
Developers should understand the business problem before writing code — not just the ticket, but the intent behind it.
Make architecture visible from day one
Expose developers early to system design decisions, trade-offs, and reasoning.
Train review as a core skill
Reviewing AI-generated code is as important as writing it. Developers must assess correctness, structure, and suitability — not just syntax.
Build operational awareness early
Developers should understand how systems behave in production: logs, monitoring, incidents, and user impact.
Create deliberate practice
AI reduces natural repetition. Teams must reintroduce it intentionally — through exercises, debugging, and design challenges.
Normalise “what could go wrong” thinking
Failure-mode thinking should be part of everyday development, not an afterthought.
5 Core skills to build into your team for 2026
The most valuable skills are not new. But they are now unavoidable.
- Systems thinking
- Clear communication
- Critical evaluation
- Understanding user intent
- Designing for failure and scale
AI has not changed what matters. It has removed the ability to hide from it.
Teams that invest in these skills will not just move faster - they will make better decisions at speed, which is where real advantage sits.
The Propel Tech AI engineering model™
Adopting AI successfully is not about tools alone. It requires structure, discipline, and a clear operating model.
We use a structured approach to integrate AI into software development while preserving architecture, governance, and long-term maintainability.
01 - Structure
Establish clean architecture, codebase health, and governance foundations before introducing AI acceleration.
02 - Accelerate
Apply AI-assisted development across build, refactoring, testing, and migration workflows.
03 - Validate
Ensure senior developer oversight, architectural control, and production readiness.
04 - Govern
Implement secure data boundaries, access controls, and full auditability.
05 - Optimise
Use AI to enhance documentation, extract knowledge, and continuously improve systems.
This ensures AI increases speed without compromising engineering quality.
How Propel Tech can help
Propel Tech helps organisations adopt AI in a way that strengthens - not weakens - their engineering capability.
We work with teams to:
- Introduce AI safely into development workflows
- Build the skills required for AI-augmented engineering
- Maintain architectural integrity and governance
- Develop developers who think in systems, not just code
If your team is navigating this shift, we help you move beyond experimentation to structured, production-ready adoption.
Contact us today to find out more about how we can help you shift to Bespoke AI software solutions. Or take our AI readiness Audit to see where your business stands on its ability to implement AI.
Your partners in possibilities
As experts across existing and new technologies, we don’t simply solve software problems, we find solutions that help manage change so that your business thrives and grows.
We’re eager to hear about your project goals and turn them into reality. Get a free consultation to make tech possible.
Brief us