The Hidden Cost of Scattergun AI Adoption

I haven’t written a blog post on the website in a while, so what have I been up to?

Firstly, I’ve continued to read, reflect and expand my knowledge base. I’ve embraced tools such as Obsidian for note-taking and knowledge management, while also learning how to use research and qualitative analysis tools such as NVivo. One thing that is becoming increasingly clear is that undertaking doctoral research is not simply about reading papers, it is about learning how to structure, interrogate and synthesise ideas over long periods of time.

At the start of the month, I submitted a short paper in support of presenting at the ICIS 2026 conference in Portugal. Writing academically and communicating ideas with clarity and precision is still something I’m refining, but the research direction is becoming increasingly concrete and a clearer PhD arc is beginning to emerge.

Outside of the PhD bubble, I’ve also been reflecting heavily on the increasingly scattergun approach many organisations are taking towards AI adoption. As businesses create mandates for employees to “use AI”, the rise of disconnected, locally developed agents and workflows is rapidly becoming the norm. In some ways, it feels reminiscent of the proliferation of locally built Access databases that plagued organisational data landscapes for years.

The difference is that AI systems are not simply storing information, they are beginning to influence decisions, workflows and operational processes. It raises an important question: at what point does the race to adopt AI begin to generate diminishing returns through fragmentation, duplication and governance overhead?

I suspect there may eventually be a tipping point where organisations are forced to step back and unpick large volumes of decentralised AI tooling in favour of more structured, governed and operationally sustainable approaches.

Looking forward, it’s going to be a busy summer ahead. The next major milestone is completing the systematic literature review building on the foundations of the initial short paper. There is still a huge amount of work to do, but I genuinely believe the research has the potential to contribute something different to the conversation surrounding Responsible AI adoption in high-stakes environments.