The Layperson's Guide #3: How to Get Started with Conversational AI

If you've read my preceding layperson guides and are of the opinion that Conversational AI is a solution your business could utilise, there are a number of steps to take on this journey. That said, I think this Mark Twain quote is worth bearing in mind from the outset:

The secret of getting ahead is getting started. The secret of getting started is?breaking your complex overwhelming tasks into small manageable tasks, and starting on the first one.

The first thing to say is that you need buy-in across your business, both at your executive level and the leaders who will actually be responsible for implementing and managing the conversational AI strategy. This isn't a technology that sees its best results being deployed in silos. That simply leads to a disjointed customer experience and may do more harm than good. Conversational AI is transformational and can impact every corner of the business, both internally and externally facing, so once everyone is pulling in the same direction you stand the best chance to make it a success.

It doesn't have to be complicated, it just needs to work

Just because conversational AI has the power to transform how a business interacts with its customers doesn't mean a strategy the length of War and Peace is required. Transformational journeys take years to reach full maturity and like any great journey, they all start with a first step. Often this will be with a high volume, low complexity customer intent; think of your FAQs and most common reasons customers contact you.

Let's take an online retailer as an example. One of the most common FAQs they'll receive from customers is an enquiry regarding their order. Such an intent does not need a costly human agent to reply each and every time a conversation starts with "where is my order?". Instead, a conversational AI can understand what's the real intent here, i.e., my order hasn't arrived within a timescale I expected it to and I require an ETA for its arrival. Using this intent, a chatbot can then conduct some simple Q&A with the customer to verify their identity and relay the order status using information from the retailer's CRM. In a nutshell, that conversation is a request for a simple piece of information which can be easily retrieved using a chatbot. It is fully contained within the bot journey without the need for any human intervention.

Let's say the same customer wants to reschedule their order delivery based on the current ETA. Now the conversation is a transactional one. For illustrative purposes, the first iteration of this chatbot is unable to handle such an intent yet. No problem. This is where the bot should be able route the conversation to a human agent with the correct skillset to handle the customer's intent and complete the conversation.

Ultimately, good conversational design is what's key here. At any point where the bot fails to recognise the customer's intent or sends the customer in a loop, it should automatically route to a human. The purpose of this technology is to enhance the CX by making it quicker and easier to arrive at a desired outcome. As soon as that's not happening, get it escalated.

So take a look at what some of your most frequently asked questions are from customers, think about how this can be automated and you'll be on your way.

Entry points & communication channels

Customers aren't going to have contact details for businesses they transact with saved to their phones. Chances are they'll search online for how to contact a business. Or maybe they'll have the app. Or perhaps they received a letter or bill with contact details through the post. Whatever it is, there are numerous ways a customer can search for how to initiate a conversation with a business.

A company website will be the default go-to for many customers. This is a great entry point to introduce your chatbot. The 'Contact Us' page, often a page plastered with the company's phone number in big, bold font or a multi-field web form inviting customer contact, is prime real estate to instead direct customers to start a conversation via web messaging, WhatsApp, Apple Messenger or whatever messaging channel is most appropriate for your business model and customer demographic.

A letter or package can have a QR code which initiates an in-app conversation or an email can have a hyperlink that launches a WhatsApp message. If voice is one of your biggest channels, taking your customer out of the IVR queue by playing a message that says "Press 1 to start a messaging conversation" is an incredibly powerful way to take volume away from the contact centre. These are all examples but the point is that looking at the journey through your customers' eyes will help you identify what entry points they face and introducing your messaging channels as an option there will start to shift your customers toward your conversational messaging strategy.

Name and personality

This technology is artificial intelligence, let's not forget that. You shouldn't be starting off with the intention it's going to fool your customers into believing they're conversing with a human. That will be a very jarring experience once they realise it's a bot and erode trust. Make it clear at the top of the conversation that this is a bot. Many brands will call their bots digital assistants, virtual stylist (if you're in fashion for example) and so on. Giving the bot a name will help bring it to life. The personality of your bot should then reflect your brand. The language it uses will depend on your industry and customer demographics so you'll need to decide its tone of voice, how serious or playful it is, the type of words and terminology it uses, whether it uses colloquialisms, emojis, GIFs etc.

This doesn't need to be fully refined on day one. Get started and it can be improved and tweaked over time.

Conclusion

This is barely scratching the surface but then again, that was the intention. There are other things to consider such as what metrics you use to measure the efficacy of your messaging channels, the in-house skills and resources required and so on. While it's important to think about these and have plans, it's not critical to have them all nailed down from the beginning. As the saying goes - you don't have to be good to get started but you have to get started to be good.

I've worked in SaaS most of my career and can confidently say a good software partner makes all the difference. I've seen failed software deployments over the years and often this was because the customer was left to their own devices trying to figure out how to use the technology, and ultimately failing. Initial investment aside, the opportunity cost and detriment to the end customer have a huge negative impact as well. There are many providers of conversational AI technology on the market - my advice would be to scrutinise their track record and customer success stories and avoid being lulled in by the cheapest bidder; it will ultimately be a false economy.