Shifts in Perspective

I’m back from holiday and there’s a lot to catch up on. (We had a lovely time and it was very sunny, thank you for asking.) The world of AI has moved on a lot even in my short absence. For example, I’m discovering new terminology to get to grips with, such as the Dark Factory which seems to be vibe coding with engineering rigour.

Going away and getting some distance is always a good way to get fresh perspectives on challenges. I think the challenge of making the transformation to AI-native needs three shifts in perspective.

1) The transformation to becoming AI-native is no less than the next industrial revolution. We are still at the start of this revolution and cannot yet predict the consequences. So our goal is not a static destination, not a “what good looks like”. Instead we should master the ability to keep on this transformation path by institutionalising a capability for continuous adaptation.

2) Using new AI tools in existing ways of working won’t yield expected gains. Becoming AI-native is to learn to work the way agents work. Part of this is really understanding how LLMs work and the other part is to institutionalise the guard rails to mitigate weaknesses, such as mastering context engineering and starting up a new AgentOps function.

3) Most profoundly, this is more than a redesign of workflow; we need to redesign work itself. Once we learn to confidently delegate to agents, we can offload not just mechanistically repetitive but also reasoning tasks to the agents. This enables us to focus on defining the next problem to solve and not worry about the actual solution.

You may well be asking: shifting perspective seems very strategic and intangible; what does all this mean for my tasks today? To make strategy actionable, I find a good next step is to agree on decision principles. These are heuristics to help us choose when there’s a decision or trade-off to be made. I suggest these three principles that map onto the three shifts in perspective:

1) To know where to make the next capability improvement, elevate the next constraint: find the binding bottleneck, fix the cause, then find the next one.

2) Everything as Code: if it's important it's code and can be generated on demand or prepared to save cost. Why? Because the goal is to codify the problem itself so agents can solve it at run-time.

3) Humans Above The Loop: shift from reviewing every line to governing the verification systems that assure trust at scale. This is not vibing nor keeping a tight grip but a risk-based balance between the two.

When you’re next not sure how to proceed, following these principles will help you remain on the AI-native transformation journey.

© Mind Rocket Services Ltd 2026. All rights reserved.

Avi Sinharay

CEng, MIET, MEng, MA (Cantab.)

Fractional CTO, Director/VP of Technology

Core Expertise

Technology Leadership, AI Native Dev, Operating Model Design, Engineering Culture

Domains

Health Tech, Media Tech