Most business process steps involve someone being asked to perform an action based on some given information and to either move the information along the process or enhance it in some way - sorry about stating the obvious! These steps may involve making a decision on who or where to forward the information. But these simple steps can often be troublesome and time-consuming, slowing things down and causing delays. Emails end up in the wrong team, information is transposed as it entered into spreadsheets, events in CRM and finance systems have to be reversed out. So how can AI tools help?
AI assistants are great for hand-off and routing tasks
We have worked with Large Language Models (LLMs) such as Microsoft CoPilot and OpenAI's ChatGPT to successfully pick up 'bridging' tasks where automation had been thought to be impossible. The success has had a lot to do with how LLMs have been trained and how good they are at interpreting unstructured natural language - which, funnily enough, is how most business users involved in a business process communicate with each other.
Key information can be extracted and structured without complicated logic
We have also found these models to be surprisingly good at understanding what might be thought of as specialist industry language. These models are quite capable, for example, of determining the difference between a social and a business email, or between an insurance quote request and a claim enquiry, or between an ice-cream order or an electrical component specification. Once categorised, the emails can be automatically forwarded to the right processing teams and events initiated - including pre-filling data. It's also easy to add new types and conditions without having to write complicated programs and scripts.
By combining these capabilities with skilled prompting, and our deep knowledge of tech-led business process improvement, we have implemented successful email triaging solutions, have automatically lifted entities into databases with high levels of accuracy, and have triggered tasks in multiple systems.
If you want to find out more about what can by achieved contact us now.