Voice Biometric Adoption challenges in the Call Centre Industry

Matt SmallmanHome Page, Make the case

There is no doubt that Speaker Recognition using Voice Biometrics in the contact centre allows organisations to improve the efficiency and security of their calls, while at the same time making it quicker and more convenient for callers to get their needs met. The underlying technology is proven, implementation best practice established and caller acceptance is high. Some of the most trusted brands have implemented it and every week there is news of another deployment. Industry awareness is also high as we are yet to meet a senior leader of customer contact operations who doesn’t wish they could deploy this technology. So why is it that the majority of these deployments are in substantial businesses and mostly in Financial Services but adoption elsewhere is low?

Reality Check

In talking to dozens of potential and actual end-users around the world, it is clear that it is hard, expensive, and time-consuming to deploy this technology today.

Hard

  • Is it Active or Passive? – To start with, organisations must generally choose between implementing an automated service using text-dependent voice biometric technologies, or a more natural agent-based service using text independent technologies. On the one hand, integrating with automated services can deliver significant call avoidance benefits but generally has lower adoption rates due to the requirements to repeat the same phrase. On the other, agent-based systems have higher adoption rates and are far more effortless for the customer but only deliver marginal handle time benefits because it’s too late for self-service by the time you are speaking to an agent. While it might be desirable to do both, and they can be complementary, this is nearly always impractical as a result of the cost and complexity of delivering what are effectively two different solutions.
  • Internal Integration – For implementation, every solution then needs to integrate with the organisation’s telephony platforms, agent desktops, and customer databases. On the business side, multiple existing business processes need to be updated, hundreds or thousands of agents trained and numerous internal stakeholders engaged in addressing issues such as privacy, security, and fraud risk. Even in small, relatively agile organisations, the degree of internal co-ordination required to achieve this is challenging and rarely the only activity underway.

Expensive

  • High Internal Costs – This complexity is not without its financial cost with capital expenditure on hardware, network changes, internal development, business change, and professional services for integration and configuration often totaling 3 to 4 times the Voice Biometric vendor’s headline licensing costs.
  • Challenging Relative Benefit Case – These costs must balance against the customer, efficiency, and security benefits. In high value at risk industries with frequent callers and long term relationships like retail banking, these can be substantial. They have, in many cases, been forced to evolve complex and lengthy knowledge-based authentication processes to mitigate their fraud risks, with the inevitable side effect of high genuine customer failure rates, dissatisfaction and extended talk times. In other industries however, the benefits (see our Making the Case article) are orders of magnitude smaller and challenging to make a compelling case.

Time Consuming

  • Choosing the right Voice Biometric product – Because of the expense, effort, and risk of being locked in, organisations naturally want to make sure they choose the right vendor, leading to protracted procurement processes. The lack of any independent benchmarking or vendor comparison requires organisations to engage with a plethora of vendors with significantly varying track records and competence. In many cases, these vendors would rather discuss academic concepts of relative accuracy and esoteric features, rather than the practical realities of designing and implementing a scalable service for adoption and business success.
  • Time taken to get optimal results – Once implemented, tuning and calibrating the core engine to achieve the levels quoted in the marketing literature usually requires large volumes of customer audio that must come from the production environment. This volume takes time to gather and only once analysed with changes implemented can the service can start to deliver meaningful benefits. Finally, because consent is nearly always required before registering or enrolling callers, there can be a significant delay between implementation and benefit realisation depending on the frequency of callers.

While we have seen determined organisations get to full-scale deployment with thousands of agents within 12 months, this is very much the exception rather than the rule, with many implementations taking several times longer.

The harsh reality is that with today’s products and delivery models, in the vast majority of cases even on a five-year time horizon, the return on investment is poor. Significant upfront costs outweigh the present values of those benefits achieved far later in the period and Voice Biometric adoption therefore remains low.

As a rule, only organisations with the highest value at risk, or significant scale to spread the investment over, are prepared to invest. Organisation’s without these features may still take the plunge for broader customer experience, innovation, and brand reputation reasons but these will inevitably be the exception.

Opportunities increasing Voice Biometric adoption

Hope, for the industry and technology, is however on the horizon, as a result of the conflux of several trends and innovations:

Voice Biometrics As A Service

While some vendors have always delivered their products this way, it has not been the preferred implementation approach for most end-user organisations. As they increasingly adopt this delivery model for other business-critical applications however, this attitude is rapidly changing. Perhaps, more importantly than the delivery model is the commercial approach that allows organisations to pay as they go. As organisations only really derive value at the point of verification, it’s only fair that they only pay for these so expect per verification charging to emerge as the dominant pricing construct allowing far more explicit alignment with value delivered. With the right design, this approach will also allow for more straightforward migration between vendors, leading to greater price competition and incentivising vendors to continuously improve their service to retain business.

Cloud Contact Centres

There is no doubt that the majority of contact centres will transition to cloud-native telephony platforms in the medium term and we expect smaller organisations to do this even quicker. The very largest organisations may transition more slowly but even they are likely to move to traditional platforms delivered in a more cloud-like way. Whichever way you go this transition provides an increasing opportunity to standardise and often simplify integration points and the increased similarity of audio will allow vendors to offer better out of the box biometric models.

Conversational IVR

The increasing use of natural language self-service applications and callers more conversational engagement learned from devices such as Amazon Alexa is increasing the length of audio provided to automated services. At the same time vendor’s speech scientists are improving the underlying algorithms, to the extent that these still short utterances can achieve acceptable verification results using text-independent enrolments. This holy grail will allow organisations to derive more significant self-service benefits than traditional approaches, while also improving agent interactions.

Conclusion

We’ll explore each of these themes in subsequent posts (so don’t forget to sign up to be notified when they are published), but it is clear to us that as these themes play out Voice Biometric adoption will increase, as the cost and complexity of implementation will fall, in some cases dramatically. Because this technology so significantly impacts the quality of interactions organisations have with their customers, we are excited about the impact it will have on customer service in the far broader range of industries for whom this is now affordable.

We are also looking forward to adapting and making more broadly available our hard-won best practice from the most successful existing implementations and plan to increase the transparency of the value and applicability of different vendor offerings, so that organisations can deliver the greatest possible impact in the shortest possible time.