Investigate

Starting Your AI Journey

Investigate

Starting Your AI Journey

It is vitally important to have a clear grasp of customer intent and process challenges before embarking on an AI journey. This is a point that cannot be stressed strongly enough.

Without understanding all the reasons for contact, automating a process may actually lead to frustration and complaints, putting customers off any future communications. Similarly, if the processes that are being followed and the issues associated with them are not clearly understood, implementing AI on top may further complicate the situation.

For example, a process may be so complex, typically requiring hands-on intervention from a specialist within your team, that attempting to automate it would run the risk of skipping hidden but nonetheless crucial steps, negatively affecting the odds of a successful resolution. On the other hand, a particular type of contact may be so infrequent, or even completely new, that a process for managing it has not yet been formalised, in which case this must be established and refined before any sort of automation is considered. The absolute worst-case scenario of a customer being led in circles through various menus, none of which relate to their actual issue, must be avoided at all costs.

So, what is standing in the way of this holistic understanding? Put simply, conversations. Conversations happen everywhere, both formal and informal, structured and unstructured. Most organisations hold a potential goldmine of data here, but do not have systems and processes in place to track and analyse it, and so miss out on a range of invaluable insights.

Contact centres may have a basic handle on voice conversations, for example, or chat interactions, and most contact centre software solutions provide a degree of insight into these channels. But the information gleaned may only be basic, based on IVR choices, drop-down menus, or “disposition codes” that advisors enter at their discretion.

As a result, any reporting generated will not necessarily reflect what customers are actually saying and how advisors are actually responding. Industry adoption figures around technology bears this out. According to ContactBabel³, only 25% of contact centres are using interaction analytics, where deeper analysis of spoken and written communication takes place.

Moreover, there are often areas of communication that occur without any tracking or monitoring whatsoever, such as email or internal conversations. Whilst 46% of contact centres use email management software (ContactBabel³), it is often Microsoft Office, with no formal structure in place to monitor and measure service levels. In our experience, this can lead to blind spots, with significant activity occurring that doesn’t serve the wider customer experience.

³ The UK Customer Experience Decision-Makers’ Guide 2023-24 (ContactBabel 2023)

This is where AI can help, providing clarity on the nature of demand and shining a light on poorly designed processes and sources of customer and employee frustration.

This is a targeted machine-learning exercise that is performed on large sets of unstructured conversational data using purpose-built algorithms. Data sources may include call recording transcripts, emails, customer reviews, or other data that is relevant to the subject matter. The process is normally iterative rather than linear, as it may take a few rounds to identify and gather all sources and cleanse the data. Time is taken to ensure data format is aligned and suitable for ingestion. During this step, the selection of data sources is key, to ensure that the results are directly related to the area under investigation. It is also important that the data set is large enough to feed the AI algorithm.

The resulting output provides insight that would not otherwise be available by traditional means. This includes (but is not limited to) topic analysis, topic frequency, sentiment analysis, and routing and effectiveness analysis.

Contact us