The bottleneck in business automation and how AI breaks it
Business automation has long been touted as a way most orgs can boost efficiency: eliminating repetitive tasks, connecting systems, and allowing people to concentrate on more valuable work.
But there’s a catch.
Most organisations run into the same obstacle: automation is only as effective as the skilled technical teams that design and maintain it. The business often doesn’t have a clear idea of what it needs from the start, and the tech team can only deliver based on what’s documented or the context they have. The outcome? A frustrating cycle of misaligned requirements, missed chances, and solutions that feel outdated as soon as they’re launched.
Fail fast? Not in this scenario.
The breakthrough
Enter AI.
For the first time, natural language serves as the gateway to automation. No lengthy requirement documents. No need for a translation layer between “what the business wants” and “what engineering can create”. You simply describe the process in plain English, and the system translates it into action.
When I tried out one of these new platforms, it completely turned my engineering instincts on it’s head.
As a software engineer, I was used to accessing data, in its rawest form, from basic queries to API driven data. The constructs were the same: a carefully crafted query that was explicit, detailed and repeatable, consistently delivering the same format and deterministic data structure every time.
But this automation tool operated differently - you created steps in a process, similar to Azure Logic Apps, for example. There is nothing unusual about having a user-friendly interface to accomplish powerful tasks. However, the steps were actually LLM prompts. All the connections and integrations happened behind the scenes, but now the query language was just sentences in English. 
It’s just data abstraction, right, but no-code at this level of criticality for the org?
The shock and the opportunity
My first reaction: unease.
- Could this be consistent?
- Could it be repeatable?
- Is it accurate?
- Is it secure?
- What if the model changed its behaviour overnight?
- What if it starts hallucinating? (they all do)
These are valid questions. But then I realised: that’s not the point.
Whilst this process felt wrong to the ways of working within software engineering there’s a bigger picture here. It’s the low hanging fruit that AI should be supporting, and that’s business empowerment.
By lowering the barrier to entry, AI lets teams’ experiment before they even know what the “right” requirements are. It creates space for curiosity and discovery, the kind of low-hanging fruit that traditional automation can’t reach.
The takeaway
AI automation won’t replace engineering excellence where stability and scale matter. But it will open the door for business users to test, learn, and move faster than they ever could through the traditional requirements pipeline.
It’s not about automating perfect systems.
It’s about automating possibilities.
And that’s the rewiring that changes everything.