5 minute read
The practicalities and best practices for deploying conversational AI
What financial organisations need to know to get the most out of their virtual agents
In its predictions for 2023, Forrester noted that, despite some dampening, data-driven automation would continue apace into the next year. Conversational AI is one such automation technology growing in profile for the banking and financial services market. Virtual agents are one of the most widely applicable and easily implemented data-driven automation technologies available to financial organisations.
Bespoke conversational AI platforms can transform finance businesses when regularly maintained and managed properly. However, this technology isn’t as commonplace across the sector as it should be. Many banks and financial institutions may have basic chatbot functionality to assist customer enquiries. Still, when it comes to fully-fledged automation, many are yet to leap into conversational AI as they are unsure where to start.
In order to progress and adopt a properly automated approach to customer service and the employee experience, businesses need to take four important steps to achieve the best results from a conversational AI solution.
Do your homework
The first step for any business is identifying a use case. No two conversational AI solutions are the same, or at least, they shouldn’t be. Businesses' use cases can be as simple as checking account balances or as high-level as automating multi-layered customer enquiries. Identifying what solution works best for you is key.
Responsible implementation is integral here. Conversational AI shouldn’t be all or nothing; slow and considered implementation works for everyone, customers, stakeholders, and the AI itself! The most successful conversational AI cases have been allowed to grow and take on greater responsibility at a manageable rate. Make sure you know what you want to achieve and set yourself attainable targets. Treating the AI as a digital colleague will help change perceptions of investing time into training the AI models. We know that to get the best out of humans, it isn’t just the initial training that matters; ongoing learning creates world-class output levels. The benefit of doing this with AI is that it's many times more scalable and consistent in customer outcomes.
Nurture your virtual agent
Once you’ve established a use case, it is time to implement your AI. Conversational AI uses Natural Language Processing (NLP) to interpret and understand voice and text commands by cross-referencing with a catalogue of relevant information. In a matter of weeks, your conversational AI can be up and running.
However, just because you have a virtual agent up and running doesn't mean your work is over. The most successful conversational AI-powered virtual agents are kept up to date, honed and tended to over time. Preparing a team to manage your bot is important during the initial scoping phase. Don’t worry – they don’t have to be quantum computer scientists; quite the opposite is required! The best AI trainers understand customer priorities and can craft the training data for the programmes accordingly.
Expect some challenges along the way
Of course, it is unlikely that everything will be smooth sailing, particularly if you plan to be innovative with your solution. For example, voice-adapted virtual agents are still in their relative infancy, but the human voice represents one of the ‘holy grails’ of automation. “Talking to a robot on the phone” is often the target of much customer ire, but poorly maintained voice assistants give the rest of them a bad name. Make sure you keep an eye on your bots and update them in line with innovations, as this will ensure you’re always at the cutting edge.
Moreover, NLP requires vast datasets and needs to straddle many different areas of a business to be effective. This requirement presents unique challenges for larger organisations with labyrinthine and archaic IT setups. Creating a hospitable data infrastructure for conversational AI is a huge task but a necessary one for all businesses. The benefits of a coherent data strategy are apparent. Still, the potential
for smooth implementation of conversational AI only emphasises these benefits and makes it imperative to have a clear data strategy.
Work out how you want to measure success
There are three pillars of success for conversational AI for customers.
1. Automation: Have your customer resolution rates improved? How widely adopted and used is your conversational AI?
2. Commercial value: Have you seen upsales because of your conversational AI? Is it adding value?
3. Customer satisfaction: Are customers receiving an improved service? Are you seeing less customer churn?
Conversational AI may not always run smoothly, but that is no reason to give up on the technology. Re-evaluating an underperforming virtual agent, reassessing the use case and retraining your AI solution can unlock unrealised potential. AI is a constantly evolving technology, and with patience and fine-tuning, it can work for organisations of every shape and size. The important thing is to understand your goals for the business and how your AI platform can help you to achieve these goals, as this will guide how you develop the solution over time.
Getting started with conversational AI, but should you Build or Buy?
The build or buy debate is one that the financial industry is incredibly familiar with. Suppose you’re in a fortunate enough position to afford to build your own conversational AI platform - you’ll need to wait several years before you can get going with it. You’ll need the knowledge and insight in-house to make it work. Whilst many large institutions will have vast development resources and feel the need to maximise their utilisation, the expertise in this new area will take time to achieve and delay further successful implementations.
Hence, we are seeing a rapidly rising level of interest from many organisations choosing to buy into solutions having taken a first stab at building themselves. Not only is it faster to buy readymade solutions to meet consumer demand for new services immediately. Third parties can take a more objective view of your operations, bring in their experience, and help to hone your solution with their knowledge and expertise in AI technology. Ultimately, whether you build your virtual agent yourself or bring in a third party, your involvement in every step is the secret to success. For example, establishing things like tone-of-voice phrasing needs to happen in-house, but outside experts can provide invaluable insight.
Time to get started with conversational AI
Conversational AI is here to stay. This is not a flash-in-thepan technology; automation is the future of customer service in the finance industry, as well as having many other use cases to help financial businesses operate more efficiently. Implementing any new technology to your business will be a little scary – there will always be unknowns. However, by following the steps above, and taking a pragmatic approach, banking and finance organisations can lean on conversational AI, virtual agents, and automation to drive transformational results for their business.
Sanjeev Kumar, VP of EMEA, Boost.ai