Baffled by bots? One of Australia’s top experts explains how the tech helps marketers
Mark Halstead is the MD of iCumulus, the digital direct response agency that now builds and trains chatbots for clients. Here, he gives his jargon-free guide to how artificial intelligence works, and how it can be used to generate leads, save businesses money and ensure customer-service queries are that little bit less painful.
Today’s chatbots differ from older models because they’re intelligent – they’ve got a brain, which means they can interpret written text and speech. The technology has been quietly improving over the last few years, and it’s gone mainstream because it’s now so much more accurate. Think about how good the “did you mean” feature is on Google today. That’s all so important because if the bot’s not right, people won’t use it.
Specific chatbots need to be trained, which takes about three months. They start with what we call a “university qualified brain” – we use one provided by Google – and afterwards, we add niche vocabulary to move it into a field of expertise. The bots will never be perfect. We can launch the bot with 80 per cent accuracy but to get to 95 per cent, which is what consumers will put up with, it will take a further four weeks of training. For the mismatches, we hand over to a human, who can review the dilemma, resolve it and train the bot. These issues will always be resolved in the future.
People also forget about the importance of user experience and design. It’s crucial. It has to be quick and better for both operator and user, as otherwise, these inquiries turn into an angry customer because they can’t get through.
AI can, therefore, help businesses become so much more efficient. If 60 per cent of queries are basic – maybe how to reset a password – you can save money by not having your costly human agents deal with menial tasks. The staff remaining can use their talents to answer the consumer with a more human touch. AI won’t entirely replace humans, but they will make everything more efficient. The advancements we’ve made in the last five years will be double again in the next. I can see most customer service conversations starting with AI intervention very soon.
Today, AI is mostly about problems, but it could also be around scheduling: trying to arrange a test drive for a car, for instance. The bot looks at certain dealers’ calendars and matches that to a particular caller’s time. It then can make changes to one side’s calendar to make things match. Or how about a concierge service? If I’m in a shopping mall, I can walk in and ask a bot, “Do you sell sneakers?” and it will say, “Yes, here’s five stores. Choose one.” Then I’ll pick one, and it will take me there via directions. That’s not far away because a lot of the tech, such as online maps, is already here.
The specific nature of each one is so important. For example, one of our key clients is in insurance, so we built a bot to focus on sales and one that took into consideration that the business had a lot of outbound call centres. Why build a bot? Because not many people are picking up their phones anymore. Sales teams can develop prospects, but if potential customers aren’t answering, they can’t give them quotes and therefore sell. So we built an email bot to start communicating that message. At the backend, we’re teaching the tech to respond to the reply. It qualifies the lead before it’s sent back to the human.
All the big tech giants are fighting to be the market leader. As an agency, we want to be able to recommend the best platform based on the requirements – each has strengths and weaknesses – but right now we’re leaning towards Google because of the years of learning through internet searches and user mistakes. Today, people are speaking into Google Assistant via their phones, too.
Google’s Duplex tech booking a hair appointment over the phone got people talking, but it’s always dangerous going on a public stage like that. It would probably be 80% real. The particular audience who watched it were elated, yet some of the feedback in the trade press was that 50 per cent were also scared out of their minds!
People are slowly getting used to welcoming AI into their lives. Myself, I now talk to my phone and say, “Hey, send a text to X, Y, and Z,” and then also ask it to write for me. I’m not touching my phone. But you’ve got to adapt to it. You’ve got to learn a little bit about how to use it. It’s fascinating. I think there’s a big, healthy future ahead for AI.
We have created Nitrobots, a platform that develops AI-driven bots for marketing. Our team has been developing a range of virtual agents for sales and customer service, that contact, engage, and follow-up leads using natural conversation until the lead converts or opts out. The sales assistant captures the prospects’ sentiment, while the platform extracts key information such as best times to call. Then the AI assistant hands off the lead to the appropriate human sales agent to continue the conversation to close the sale.