What: A workflow is a pattern of sociatechnical activities or processes that are used to complete repetitive work tasks. Sociatechnical, because workflows involve the interaction between humans and some sort of equipment, usually in the form of information technology.

Workflows tend to be complex, holistic and hierarchical. Hence, within radiology we can talk about the workflows associated with an entire carepathway, or drill down through a hierarchy of sub-workflows to the sequence of activities it takes to dictate one sentence of a report, or type your password to a keyboard.

Why: Understanding workflows is critical when deploying AI technologies because we must aim to seamlessly integrate these tools into the preexisting workflows. In order to achieve meaningful optimisation and a return on the investment, without increasing the cognitive or operational burden.

Simply embedding tools within software environments without understanding how the tool integrates into the respective current state workflow, is wishful thinking. Poorly designed tools can have a significant negative impact on performance. For example, a standalone application with its own workstation is unlikely to be used because it adds a new set of complex workflows sequences, which reduces efficiency.

Hence we need to stop taking workflows for granted, and learn to describe how we work in relation to operational impact, and the human factors of the tools we would like to adopt.

How: The conversion about deploying AI often starts with statements like “We are short of money – can we spend to save”, “We are short of staff – can AI do some of the simple stuff”, or possibly “that tool looks like it will increase profits”. Unfortunately this scattergun approach does not point to viable solutions that can be easily adopted at scale.

A more focused approach starts with the question: ‘Which of our workflows can be usefully optimized by AI’; and ‘which are worth investing in’. These questions set you on a process of discovery that will help you uncover which tools can really add value.

Ask yourself:

• Can we use data analytics to uncover operational bottlenecks and pain points.
• Are these problems at scale and/or involve high clinical risk.
• Which professionals are at the heart of the problem you need to address: Clinical, Administrative or Technical – it’s not just about Radiologists. Who in the organization understands how people are working?
• Will AI really improve performance? – Sometimes you can optimize workflows with more traditional methods.
• Once an AI tool is deployed how will it change the way people work? Will it cause any unexpected outcomes?
• Do I understand the way people work in general, or am I simply focusing on my own unique way of working?

Finally, as you begin to understand the interconnected workflows within your organisation, you will gain a clearer understanding of the AI use-cases that could deliver real benefit.