Most organizations have bold ambitions to deliver AI technologies internally, however they lack the skills, experience or capacity to deliver these projects successfully. Conversational AI is becoming more popular as a way of automating messaging and speech-enabled applications that offer human-like interactions between computers and humans. But it’s often hard to get these projects right.
Let’s look first at where conversational AI can make a difference. There are three main forms of communication in every organization where conversational AI can dramatically change how the organization works, and they are: the speech elements around customer dialogue, email, and instant messaging. To access data and information that is useful to an organization from these communication formats takes a huge amount of manual processing. With the right conversational AI in place you can deliver cost savings and improve customer experience significantly, improve NPS scores and increase customer retention and revenue.
Many companies have used voice recognition technology in the last ten years. However, until recently, customers were saying it was a poor experience as the technology wasn’t good enough, and it wasn’t really helping improve the customer experience. As voice recognition software has become more sophisticated, we are seeing improved customer experiences.
The ability to analyze emails is at the forefront of the new conversational AI tech. In the past, companies have had to shift through vast amounts of email information and input it into large, cumbersome case management platforms that cost millions and took three to five years to implement. However, the technology exists now to gain more insights from your email - what traffic is coming through your email, the types of conversations your employees are having with customers and partners, what type of requests are being made etc. The software then takes this unstructured information and structures it. Unsurprisingly, it’s the financial services market and investment banks who handle vast amounts of data and customers who are leading the way with this type of conversational AI. They are the early adopters and where we see some of the best outcomes.
In terms of instant messaging and query management, chatbots are leading the way. There are three main use case scenarios for the use of conversational AI:
- These are the best cloud storage solutions (opens in new tab) on the market right now
- non-personal or un-personalized answers – for example, when a basic question can be answered with a generic answer, e.g. where can I find company policies on annual leave?
- personalized answers – for example, how many days of annual leave do I have left?
- personal transactions – for example, where you are asking a chatbot to make a transaction on your behalf. This involves identification and verifications of the person, an employee or customer, e.g. I’m a customer and want to make a payment on my account. In this instance, the chatbot has the additional capability to link into back-end processes.
With technology improvement, it’s not surprising that conversational AI is now on most boardroom agendas – but how do you get it right?
1. Ensuing you have access to the right skillset
Most organizations have bold ambitions to deliver AI technologies internally, however they lack the skills, experience or capacity to deliver these projects successfully. Conversational AI is such a niche discipline, and companies trying to do it themselves simply need time to learn. It’s not only about the technology skills either. AI engineers need to have an understanding of the business, the customer and the challenges facing the department.
Getting the first project right is so important – it not only delivers business benefits but also gains buy-in from senior leadership and sets a precedent for future projects. Partnering with an organization that specializes in conversational AI can help bring the skills immediately to accelerate the delivery of the first project whilst building capability internally and setting the foundations for success.
For example, understanding intents from email is complex. Experienced data scientists will help you to simplify the whole process of preparing large data sets. Email is harder to deal with as there is greater complexity in it. It’s a good idea to work with specialist companies that have data scientists on board who can help you examine what data your emails contain that could be used in a conversational AI platform to improve business outcomes. One company stands out in terms of extracting information from emails, and that is Re:infer. They have built a sophisticated platform that can scan free and unstructured text and understand the intent of the email, classify it, then extract the key piece of information required. This is enabling organizations to radically improve their outcomes from the use of conversational AI.
2. Understanding the potential for conversational AI within your organization.
One of the most common challenges with adopting conversational AI is that the leadership team do not have a deep enough understanding of the technologies, their capabilities and limitations. The starting point should be education. You should try to understand how other companies (in your industry and across other industries) are deploying conversational AI. Look at larger organizations that are innovating and the use cases they have deployed. Learn about the capabilities of each technology, and once you have the foundational knowledge, you can start to share this across your leadership team and manage expectations around what conversational AI can and cannot do for your business.
Take for example, speech. More and more voice recognition platforms are being used across financial services organizations for example, Natwest is using it to support digital identification and verification of customers. For things like instant messaging use of technology like Microsoft Azure Q&A Maker or chatbot solutions allow you to automate common queries for business functions like HR or finance. Standard requests can be programmed easily into the chatbot, and responses given quickly.
- Check out our take on the best cloud hosting services (opens in new tab) at the moment
3. Selecting the right use cases
The most powerful use cases are those that have the potential to transform customer and employee experiences. However, organizations that fail to spend time understanding their customer preferences and customer pain points often find themselves deploying technology that has little value. If deployed incorrectly, conversational AI can force customers towards other manual service channels which effectively shifts cost from one part of the business to another.
Companies that successfully deploy conversational AI tend to start with assessing what the customer needs and then work back to the technology. It’s important to understand how to make a frictionless journey for customers, which includes understanding the types of digital service channels your customers prefer to use and the technology to match.
Segmenting your customers, developing personas and documenting the customer journey are all great techniques to understand detailed pain points and develop evidence-based recommendations on how to improve. In financial services, for example, conversational AI has been used to great effect within Digital Identification and Verification (DIV) of customers, complaints handling, and servicing queries.
4. Aligning to the business’ objectives
Like other strategic automation tools, adopting conversational AI requires a strategic approach, setting objectives that align with business goals, and measuring ROI. Without this, you won’t be able to assess the results against the business case. All too often, companies fall into the trap of trialling new technology without truly understanding how the technology can solve a business problem. Conversational AI should target improvements in the biggest areas for improvement: increasing NPS, improving retention rates, reducing complaints and reducing costs.
5. Resolving data issues before you start
Conversational AI systems require data to learn, and to work well they need rich, high-quality data to overlay the AI solution so that you get great results from it. And, the richer and more accessible the data, the easier it is to produce models that can accurately understand and interpret conversation. This can be an issue for organizations that aren’t up to speed on their data architecture.
Many companies face challenges obtaining the data they need. The first step is understanding what data is available already: do you have direct access to verbatim conversation logs through speech, chat or messaging channels. Are those data sets available through modern integration layers such as APIs? Once you understand what you have, you can easily start to understand what’s needed.
Conversational AI can deliver real business benefits if executed in the right way. And as we move to a world of low code, no code software, it is becoming more and more accessible to companies. The shift towards cloud-based platforms like Microsoft Azure, AWS, and Google Cloud has also made access to software as a service simpler and quicker. All this means that organizations have the potential to deploy conversational AI more easily and start reaping the benefits of improved customer service and revenue faster. Planning is key to success, though and working with a conversational AI specialist will ensure your project delivers the business outcomes you desire.
- These are the best cloud storage solutions for photos and images (opens in new tab)
Dan Johnson, Director of Automation, Future WorkForce (opens in new tab)