To Afiniti and beyond? What direction next for dynamic routing?

Nov 15, 2024

Last week, the customer service AI company Afiniti filed for Chapter 15 Bankruptcy, leaving the future of the firm in doubt. In a world where Generative AI is grabbing most of the headlines - and investments - it’s an apt time to assess the future potential of the specific use-case that Afiniti pioneered: dynamic routing.

What is dynamic routing?

Routing is one of the most important capabilities of any customer service technology setup. For years, the combination of IVR (Interactive Voice Response) and ACD (Automatic Call Distribution) were the key tools for getting a customer to the next available CSR (Customer Service Representative) based on their reason for calling, and managing service levels across the whole operation. More recently, cloud-based CCaaS (Contact Centre as a Service) platforms enable far more sophisticated routing of contacts between different channels.

Dynamic routing seeks to take this concept further, by matching a customer with the most suitable individual CSR for their need, using a range of data sources, including:

Personality profiles - e.g. match a less digitally confident customer with a CSR who excels at patience and understanding
Recent online activity - e.g. match a customer who has just been looking at their bill online with a billing specialist
Next Best Action triggers - e.g. if a customer is identified as a high propensity prospect for cross selling of additional products, match them with a CSR with great sales performance

By taking this more nuanced approach to routing - so the logic goes - customers will be better matched with CSRs who can ensure a more successful outcome. In other words, better results from the same pool of resource, without the need to retrain anyone. It seems like a no-brainer, but many companies have found these approaches to be challenging to implement.

Successful dynamic routing - understanding the limitations

An instant uplift in performance is an attractive proposition to customer service leaders. Afiniti themselves capitalised on this by offering an outcomes-based pricing model. Their corporate customers pay a fee based on measured increases in KPI performance, continuously verified using A/B testing against a control group that didn’t receive the dynamic routing treatment.

From working with customer service leaders who have deployed many different flavours of dynamic routing, I have come to learn that there are some important considerations and constraints. Adhering to these can increase the chances of success of a dynamic routing initiative.

1. Simplicity as a golden rule

However the setup, the golden rule is to always consider the trade-off between any benefit that dynamic routing brings, and the complexity it adds to the operation. For example, dynamic routing can make workforce planning more complex, or limit managers’ ability to quickly adjust routing rules to manage spikes in demand. Any implementation of dynamic routing needs to consider the full set of workforce management and routing requirements of an organisation.

2. Carefully select KPIs

Dynamic routing works best when optimizing routing choices to maximise 1, or at most, 2 KPIs. Go beyond that, and the trade-offs between KPI impact become so complicated that benefits are diluted.

This works well in sales scenarios, e.g. to maximise conversion rate, but it can be a tough choice in service scenarios. Here, leaders need to optimise for a balance of many KPIs, including Customer Satisfaction, First Contact Resolution, Average Speed to Answer, Resolution Time, Transfer Rate and Quality. I have found that First Contact Resolution tends to be the most useful KPI to optimise through routing choices.

3. Trade off routing choice vs. service level

Dynamic routing is most effective in ticket-based customer service operations - where a customer gets help by sending an email or raises a ticket through an online portal. In these setups, a customer can expect a response in a few minutes, a few hours, or even a few days. Because the queue times are generally longer, this provides more flexibility to route a request to any CSR, even if they are not working their shift at the time that the request is initially raised.

In synchronous voice or chat customer service, dynamic routing is limited by the impact it has on answer times. Customers expect a response within seconds, but if the best match CSR is busy serving another customer, we have to make a decision about whether to make the second customer wait. In an operation that runs at 85% occupancy, on average only 15% of the CSR population are available for a dynamic routing choice. At busy times of day, when calls are queuing, the available CSRs may go down to zero, so a better choice may be to route the customer to the next available CSR, regardless of any dynamic matching.

4. Plan for the people impact

Dynamic routing causes individual CSRs to have different experiences to their colleagues. Some may end up handling more contacts throughout the day, some may end up with only the more complex requests, some may get better cross-sell opportunities than others.

Transparency and careful change management is essential, because if CSRs lose trust in the system, then employee engagement, performance and attrition rates can suffer.

So what next?

In general, with AI assistance tools becoming more prevalent in contact centres, we see a trend towards agents becoming more generalist, driving an overall simplification in routing systems. Here, the role of dynamic routing is in something that is not used all the time, but deployed only in specific circumstances where the business value is well-proven. Some examples I have seen work include:

Routing frequent callers to CSRs who are especially adept at training customers on how to use online self-service
Routing customers who are flagged on a system as vulnerable to more experienced CSRs
Routing customers with a high cross-sell opportunity to CSRs who have better conversion rates

One way to get this right is to pilot different approaches comprehensively by manually applying routing rules, before embarking on any project to develop automated dynamic routing.

Latest perspectives from BCG

GenAI will fail. Prepare for it.

For all its powerful potential, generative AI (GenAI) can generate incorrect outputs, produce harmful or offensive content, and expose organizations to new security vulnerabilities.
Before launching GenAI-powered services, organizations should conduct comprehensive testing and evaluation to identify and mitigate these risks. That evaluation should rely on human testers augmented by automated platforms.
But even with comprehensive testing and evaluation, the risk of system failure with GenAI will never be zero.