7 things to avoid when implementing speech analytics for risk

Written on
byJames Gallagher

Remember when you got that new mobile, or update on your phone? You no longer had a thumb print but a facial recognition scanner to unlock it. Now, if you swipe one way or the other a new pop up can be used to perform an action much faster. At first you were a bit confused, but now you wonder how you used to live without that ‘do not disturb’ function, to have peace and quiet after work hours.

This neatly illustrates a time where you implemented a piece of software, somewhat unaware; to understand your needs, understand the features, and finally improve your user experience by applying those features to your needs. This methodology is the starting block your firm needs to adopt, when using speech analytics, to boost off toward better meeting customer needs and company goals.


Goals, Comprehension and Implementation

There’s little point in buying a top of the line fancy new smart phone just to make calls. So too is there little point in purchasing speech analytics software if there’s no intention of trying to understand and use it for its wealth of potential.

It’s paramount that you understand what you want from speech analytics. Apple and Android do that for you with your new phone, but it’s up to you to realise how to adopt technology to better meet the goals of your firm. To understand what your firm’s goals are there are three approaches. The first is clearly asking the customer what they feel they need and want. The next two approaches take this further by predicting what the customer may need or want even if they do not realise this. This can be done upfront by interviewing frontline workers. Asking them what they believe customers think and most want from the company. Also, by working with senior management and their strategic goals for the future of the company. Once goals are set, it‘s much easier for employees to figure out to what end they are working toward and how best to use resources at their disposal to achieve it.

For a firm to understand how best to implement and integrate systems to meet these goals, it’s necessary for an individual or team to spearhead complete understanding of the software with its functions and metrics. This group can then be a bridge between the specialism of the firm and the speech analytics tools, adopting them where it is most efficient.


7 Common mistakes:

  1. Lack  of data – To create an accurate, meaningful solution, NLP platforms need to be fed with a certain amount of data specific to the company and the insights it wants to uncover or processes it wants to streamline. The reason why AI is so great is that it learns and so the more you feed it, the smarter it will be. For example, if you have a large volume of calls where customers are distressed then speech analytics can show how certain responses by the agent improved or worsened the situation across a population. This will help in training call centre agents. To learn more on how NLP works check out our very own Barry Haddow’s blog. He highlights the importance of using a large amount of data to create the best contextual embedding for your audio calls.
  2. Unclear labelling – Insufficient data labelling is a common challenge. Many firms provide completely raw data but to create a solution that solves a specific problem, engineers need to understand what the data means in relation to the firm’s service or product. It’s mutually advantageous for firms to help the engineers understand what your data is and what you want to learn from it. Like a personal trainer the engineer can help you to achieve your fitness goals but you need to set out those goals and state what you can or cannot do so they can tailor a programme for you.
  3. Generalisations – We all know that writing a sentence into Google may not give you the right result. It’s about using keywords. Further, it’s about using the right keywords. For example, if a customer says he is “very disappointed” he may be speaking about the weather or the football result and not the firm’s service. However, if certain keywords are persistent such as “interface” the probability that a customer is referencing another issue is much less likely. Therefore highlight specific words or phrases (not just “disappointing” or “bad”) that are unlikely to be used in another context – or are central to your service/product.
  4. Not making the most of the solution – As we said before, don’t buy a new smartphone if all you want to do is use it for calls! Use the software functions specifically for every process that it can enhance. You will know if speech analytics will be useful to your firm’s issues if you put in the proper research. Ask many questions and you will find many answers in the form of solutions to your problems. How can we use this for frontline employee training? Will we be able to more easily identify vulnerable customers? Do frontline employees prefer using speech analytics technologies, if so will this reduce turnover?
  5. Lack of demographic understanding – Speech analytics and NLP is a fine tuning art. Basic BERT software has been developed but needs you to understand what it is being applied to and looking for. This means understanding your customer base and its different segments. Young versus old customers may use different terminology. Therefore, embedding words into a context will require thorough research, especially if terminology is not financially sophisticated as may be more prevalent in younger customers.
  6. Not using metadata – Speech analytics doesn’t just look at the actual words. Just as in real life we pick up lots of clues from body language, so too can we learn from data such as tone. Many customers can be very polite and it will take them a long time to open up about any problems they have. Understanding metadata such as the length of pauses can contribute to understanding the feelings of a customer – a longer pause could be a sign of deep rooted frustration.
  7. Incorrect risk weighting risk – When determining the risk magnitude of a call it is important to remember that different customer circumstances have different needs. Therefore, someone who has become, for example, a widow may not be as risky as someone who; has lost their job, has low financial resilience, and does not have a computer to access online banking services. In other words, risk weighting is not just identifying the most risky circumstance but viewing all circumstances on a cumulative level of risk.


Improvement, Innovation and People

When implementing speech analytics, it’s good to set a starting point and after a specified time to look back at what has been done and determine if your goals have been met. This self evaluation helps you to reconfigure your approach to better meet targets. It will also help you and your employees to realise the benefits it provides, continuing the motivation to keep using and improve the program.

Innovation is the next step. After reviewing how speech analytics has benefitted your firm, you can expand its functionalities, and continue to fine tune the software to better meet your new goals. Feedback from employees and the specialist team on how well it is working and how it can be expanded will be crucial.

Consideration of who to work with to build a solution is a crucial stage. You may think that all solutions are the same so that the cheapest off the shelf option will do, but not all solutions are the same! Choosing the right firm is the differentiator between a competing success or a stagnating failure. It should be noted that larger speech analytic firms tend to churn through adding new businesses to their portfolio. Whereas we understand the importance of taking care of each client throughout their journey to build the best bespoke solution with them. Further, just like any consultant you want an expert in your field who has experience with companies similar to yours so they can flag potential pitfalls before they are arises. Like a seasoned Sherpa, small experienced firms will direct you to what you need to provide and consider before the journey so as to find the most suitable path for you.

To round off the benefits of AI-driven speech analytics in a more tangible way, Gallup research shows that employees who are not engaged in their work cost the business 18% of their annual salary; but those who are engaged are 43% less likely to leave with profits 23% higher for the firm! Despite the rhetoric robots are not taking over the world of work and technology is a great way to enhance engagement of employees. Gallup also finds that employees feel they can do their work better with new technology. Speech analytics must be used in tandem with your employees – so continue to enhance performance and the software so as to make both work better for you.


To learn more about how we use speech analytics, visit Aveni Detect 

Follow us on LinkedIn , Instagram and Twitter

Other recent posts

Podcast logos-03

AI’ll Take Your Job

Join Joseph Twigg and Jamie Hunter, the dynamic duo of financial services and AI, as they unleash their wit and wisdom on the game-changing influence of recent AI development on the industry.  

Replay social 2-06

An Introduction to ChatGPT in Financial Advice

Our CEO Joseph Twigg was joined by Iria Del Rio, our lead NLP engineer to talk about the explosive rise of ChatGPT and other large language models, what got us here and what this...

Three confident business people having discussion while working in the office together

Beyond the basics: Debunking common QA assumptions in financial services

Quality assurance (QA) is a critical business function, ensuring that products and services are compliant with regulation and meet customer needs. However, there are some common assumptions about QA within the financial services industry...

Board member and FCA proposal

Aveni partners with Delta Capita to power their Consumer Duty offering

Aveni has partnered with Delta Capita, a leading global capital markets consulting, managed services and technology provider, to advance their Consumer Duty offering by providing their clients with access to our cutting edge AI...

limitation of ChatGPT in financial services

Limitations of adopting chatGPT and other large language models in the financial services industry

There has recently been a lot of hype surrounding ChatGPT, and other large language models (LLM’s) and rightly so. Their extensive capabilities make them impactful for a wide range of use cases across various...

It pays to plan properly for the future. Cropped shot of a senior couple getting advice from their financial consultant.

Age Partnership selects Aveni Detect platform to enhance customer outcomes and mitigate risks

Aveni.ai, has been selected by Age Partnership, one of the UK’s leading Equity Release Advisory firms, to revolutionise its Quality Assurance systems setting new standards for customer service and outcomes.     Following a successful...


Demonstrating Consumer Duty Compliance with Technology - Key Takeaways from Aveni’s recent webinar

In our latest Consumer Duty webinar series,”Demonstrating Consumer Duty Compliance with Technology,” Joseph Twigg, CEO of Aveni, sat down with John Liver, Strategic Adviser at Kore and NED at Barclays, and Alan Blanchard, Head...

Aveni SPW

Schroders Personal Wealth adopts AI-based Aveni Detect platform to transform compliance function

Aveni, has been selected by Schroders Personal Wealth (SPW) to transform its compliance function. Through the deployment of the Aveni Detect platform SPW will use the latest advances in Natural Language Processing (NLP) to...


Human in the loop 101: what is it and why is it so important?

Financial services firms have been turning to Natural Language Processing (NLP) solutions to extract valuable insights from vast amounts of unstructured data. But even the most advanced algorithms can’t match the intuition and creativity...