Enterprise architect argues AI fails to address upstream bottlenecks in process optimisation
A new analysis suggests that organisations misidentifying bottlenecks by focusing on coding speed rather than vague inputs are unlikely to see efficiency gains from automation.
Frederick Vanbrabant, an enterprise architect with experience establishing five enterprise architecture offices, has published a blog post arguing that artificial intelligence will not automatically accelerate business processes. In an article titled "I don't think AI will make your processes go faster," published on 15 May 2026, Vanbrabant challenges the prevailing assumption that automation is a panacea for organisational inefficiency.
Vanbrabant contends that companies often misidentify bottlenecks, focusing excessively on coding speed or the volume of resources rather than upstream ambiguity. Drawing on principles from management classics *The Toyota Way* and *The Goal*, he argues that software development delays are primarily caused by vague requirements rather than a lack of technical capacity. He suggests that process optimisation should instead focus on providing high-quality, predictable inputs to bottlenecks.
The enterprise architect uses software development as a primary example, noting that developers cannot simply "type faster" to speed up projects. Instead, the majority of time is consumed by translating vague problem statements into precise technical solutions. Whether using waterfall methodologies with scope documents or agile iterations with domain experts, the core delay lies in defining what the solution should actually achieve.
Vanbrabant asserts that AI-generated code still requires significant "handholding" and detailed problem definition from domain and product experts. He argues that this effectively shifts the workload rather than eliminating it, requiring a much deeper involvement from stakeholders to write out every feature and bug fix in minute detail. Without this precision, AI tools may generate code quickly, but not necessarily the correct code.
The author suggests that true process optimisation requires ensuring that those performing the work have all the means to do so. Citing a key lesson from *The Goal*, he states that bottlenecks should receive predictable, high-quality inputs. He warns that adding more resources, such as lawyers or developers, to a department struggling with incomplete documentation or unclear directives will not speed up the process.
Vanbrabant notes that his observations are grounded in his professional background, having been part of the creation of five enterprise architecture offices. He highlights that frustrations in development teams often stem from technical debt and estimates being treated as deadlines, further complicating efforts to streamline operations through simple automation.
The piece serves as a counter-narrative to the "unrealistic expectations" surrounding AI in the current market. Vanbrabant concludes that before organisations look to automation, they must first address the fundamental ambiguity in their requirements, ensuring that the inputs to their processes are clear, complete, and well-defined.
