Challenges in adoption
Yet despite its benefits , it is this journey to integrate AI into automation processes that presents distinct challenges , particularly around data quality , governance and trust .
Data quality is key as it is the fuel that feeds the process of automation . However , getting the amounts of data needed , in a clean and structured way , is a task in itself ( and one that automation would ironically help ).
Nick points out that “ automation on top of low-quality , ungoverned data is an untenable proposition ”, stressing that poor data management can stall AI projects .
“ High consequence , low confidence use cases are obviously not ideal proving grounds for use of automation in AI ,” he says . “ Conversely , high confidence and high frequency use cases may prove better situated to benefit from automated processes .”
In sectors like manufacturing and finance , where the stakes are high , companies must ensure the robustness of their data infrastructure before implementing AI in automation .
Ethical and regulatory compliance also complicate the adoption of AI in automation , particularly in sectors handling sensitive data .
One way to tackle such ethical aspects and thus build trust , as Urmila highlights , is to build trust in AI outcomes : “ Challenges in the field include gaining trust in the AI results , understanding areas where a human touch is still needed and evaluating the costs and evolving capabilities of large language models ( LLMs ) as they continue to develop ,” she says . “ Explaining and understanding AI results , as well as the ability to govern and monitor AI execution , can go a long way toward building trust .”
Trust in AI automation is necessary for its adoption and transparency is crucial in enabling organisations to rely on these systems . Moreover , data governance and cost considerations remain critical as businesses weigh the benefits of AI against the resource demands of adopting such technology and the regulatory punishments for not complying .
Luiz highlights , however , that this is challenging , with many enterprises operating in hybrid environments , causing compatibility issues due to a mix of premise-based , private and public cloud applications .
104 January 2025