AI in customer service: questions to ask before taking the plunge

AI has always played a key role in customer service. As a flow-driven business par excellence, customer service is akin to logistics: how do you manage these flows quickly, with quality and at the lowest possible cost?
In short: fast, well and cheaply.

‍

For decades, customer services have been striving to optimize these aspects, with a particular focus on costs. The advent of offshore call centers and workflow management tools, such asAircall for voice and Zendesk for email, is a perfect example.

‍

These tools have long been integrating simple AI: automatic categorization and routing, the first chatbots using NLP models and decision trees...

‍

With the advent of generative AI, notably ChatGPT, incumbent players and new entrants alike have jumped into the GenAI race. The promise is great: to revolutionize customer service like never before since the advent of the Internet and email.

‍

‍

But how do I choose the right GenAI tool for my customer service department?

‍

The first question is: am I ready?

‍

Actually, that's the wrong question, because nobody is really ready. Even if you were, the speed of developments in generative AI would quickly catch up with you.

‍
The real question isn't whether you're ready, but how to get started as quickly as possible with a solution that suits you.

‍

‍

‍Forthis, it is essential to know your customer service well: what are the simplest and least risky questions (in terms of reputation and finances)?
‍

In general, the easiest flows to handle, the "low-hanging fruit", are asynchronous, written flows (email, WhatsApp, Messenger...), which require "static" knowledge (such as FAQs). These cases will be handled very well by GenAI solutions, and often still represent significant volumes(over 10% of customer questions).

‍

Partner selection

‍

Once the perimeter has been identified, it's time to look for the right solution. Several parameters need to be taken into account:

• ‍Complexityand speed of installation: tools that install very quickly are ideal for testing quickly (however, beware of those that promise installations in a few days that turn into months, make customer reference checks to avoid smooth talkers). But, on the other hand, heavier tools that require months of project management give you more time to get agents and other stakeholders on board.

• Pricing: several models exist: fixed subscriptions, fixed subscriptions with sliding-scale costs, success-based invoicing. Everyone has their own objectives here, especially in terms of ROI. The only thing to bear in mind is to compare dishcloths with dishcloths, estimating a cost per ticket in order to have the same standard. 

• Performance (and transparency of results) : in your benchmark, it is essential to define upstream the performance indicators to be monitored (utilization rate, productivity - DMT, number of queries processed...). Even if your own indicators are the most relevant, choosing a partner who shares performance indicators with you is an excellent sign of transparency, which is necessary to forge a lasting partnership.

• Potential for evolution: Gen AI is a brand-new technology, and new developments are coming out at breakneck speed, with substantial improvements every month. The risk of choosing a rapidly obsolete solution is high. Understanding your partner's vision and tech roadmap is key to ensuring that it will be able to evolve with GenAI: is the solution dependent on a single model (OpenAI, Mistral, etc.)? What is its vision for the various channels (e-mail, voice, chat, etc.)?

‍

Supporting agents through change

‍

Once you've chosen your solution, you'll need to support your agents through the change. This is a crucial step: without agent adoption, even the best GenAI tools won't work.

To achieve this, don't neglect agent training. Explain your objectives and reassure them of the expected benefits. To simplify this phase, choose a tool that has already been integrated into your own tools. A solution that tracks the use of AI by each agent will also be very useful for individualized support.

‍

Transforming the trial

Finally, it's time to monitor your performance and validate your partner's potential to evolve: are your objectives validated and does your future potential seem realistic?

‍

‍

Here are the keys to getting started with generative AI and choosing the best path for your personal situation. 

But if you were to take just one piece of advice from this article, it would be this: don't miss out on the AI train that your competitors will jump on.

‍