Interpret
Filter and Analyse
InterpretFilter and Analyse
With machine learning providing a wealth of conversational insights, the next step is to apply a human lens to filter and analyse the results.
First and foremost, the results must be interpreted according to the organisation’s goals and strategy. Some issues will be more important than others. For example, if an organisation primarily focused on sales conversion discovers that their advisors are spending considerable time servicing queries from existing customers, that will be a high priority and may drive a process change to separate this activity and improve the service experience.
This can be done using customer journey mapping, which is an entire topic in its own right, but will typically require interviews and workshops with stakeholders involved in the end-to-end process, as well as live observation of interactions as they occur.
This will bring the first phase’s finding to life in ways that machines cannot and so expose further process bottlenecks. For example, advisors may be resorting to manually taking notes on paper as calls progress, or referring to printed manuals because they don’t have the right information to hand via the system in front of them.
These findings will also have to be viewed with brand values and customer personas in mind. It may be tempting to automate some of the problematic processes, but this cannot be at the expense of the organisation’s principles and values. A brand that is leading on the personal touch rather than positioning itself as the lowest cost provider should carefully consider where automation best fits.
Similarly, if the characteristics of the customer base are such that a considerable proportion are vulnerable or unfamiliar with digital technologies, the strategy must reflect this. Although the public’s familiarity with AI continues to grow, with a recent UK Government report revealing that 34% of individuals surveyed are now using chatbots on a regular basis and 25% doing so for work****, this should never be assumed to be the case.
Another important human consideration is that of emotion and complexity. How important is the journey to the customer? How sensitive is it to them? And how complicated is each interaction for both the customer and the advisor? Some interactions – such as notifications of bereavement – will always be emotionally charged, and so unsuitable for automation. Similarly, some interactions may be highly complicated, with a large degree of risk, and should therefore be fronted by a person. AI can be used to support them, improving their performance and efficiency, but they must remain in the driver’s seat.
The diagram below can be a useful aid to help determine the role of the human versus the role of the machine, and who should take the lead for a particular interaction:
Contact us
