Companies are opening their arms to the world of artificial intelligence and letting chatbots do the talking. These programmes engage in dialogue with us on websites, social networks and apps, taking the place – at least in part – of human beings. We’ve all dealt with them, and sometimes we come to think of them like proper people. Companies have even started giving them names: Siri is the virtual butler on Apple devices, Google Assistant is the Android version, Alexa lives in Amazon’s Echo Smart speaker, Cortana acts on behalf of Windows and Xbox systems and Bixby is the voice of Samsung smartphones.

And there are more and more being created on an almost daily basis. They listen, analyse and understand questions put to them, providing answers (by speaking or showing content on the device screen) and personalising the user experience to fit the user’s habits.

A new way of interacting

So what exactly is a chatbot? As the name suggests, it’s a piece of software that interacts via dialogue and can pop up on a website, an app (like Facebook Messenger or Skype) or a device (a smartphone, smart speaker or videogames console). Some chatbots express themselves through writing, others respond to vocal commands and others still combine the two, using the device screen to show search results and links.

They provide answers to queries spanning customer service, weather forecasts, news round-ups, traffic information and music and they can even alter the setting on your smart central heating system – the list goes on and on.

Their responses – be they text-based or vocal – overlay or replace the graphic interfaces of classic screen-based software. And in all cases, the brain of a chatbot is based around artificial intelligence, particularly automatic-learning algorithms that allow the programmes to acquire new skills and abilities.

“Human” software

Thanks to machine learning, these programmes are able to evolve as time goes on, collating new data and getting to know the user better and better. Spike Jonze’s film Her might be a science-fiction movie, but it provides a fine illustration of the technology, hinting at the extreme potential consequences (the protagonist falls in love with the operating system, who becomes his virtual partner). Set in the future, the film isn’t too far away from the reality of the situation: in 2014, tech giant Microsoft launched a chatbot on Chinese social network Weibo that was specifically targeted at adolescents.

Xiaoice – the name of the chatbot – is a search engine first and foremost, but its characteristics are designed to portray a carefully constructed personality. It has the appearance of a 17-year-old girl, a modifiable, familiar voice (she’s “sung” dozens of songs and hosted radio shows) and is often the subject of declarations of love on Weibo, where she has over 5 million followers.

The strength of personality

Xiaoice might be an extreme case, but nonetheless it shows how a well-put-together piece of software can become a media phenomenon – a web celeb, if you like. It’s a fine example of the power of chatbots – and should serve as a great source of inspiration for those in the marketing business.

The stronger the personality, the stronger the message.

Remember Clippy, the animated paperclip that used to pop up in Word and offer assistance? He was rather annoying, thinking back. And compared with modern-day chatbots, Clippy looks prehistoric. But the idea behind him was destined to go far – his big eyes and goofy smile did give him personality, after all.

In old versions of Windows, Clippy was the nerdy mate you could go to for technical support. Nowadays, they have Cortana. Microsoft hasn’t given the programme a “physical” character, instead deciding to name her after the female protagonist of Halo, a very popular series of videogames.

Fresh from her debut is Elisa, the new Europe Assistance chatbot. Active 24 hours a day on the insurance company’s social media channels and website, she provides product information, assistance with online policy purchases and support for anyone who needs roadside assistance. Made in the image of a young lady with a blue bob and red trousers, Elisa is helping to drive the brand identity by using the same colours and adopting a similar shape to the logo. Clearly, chatbots are a useful tool when it comes to “humanising” a brand.

More individual marketing

There’s been 20 years of technological evolution, cognitive computing, machine learning and data analysis between Clippy and Elisa. But another important change has been the transformation of the world of marketing, which has gone from a “rigid” model, based on showing customers how a brand satisfies their needs, to an “elastic” model in which the brand itself adapts to the tastes and desires of people.

The industry has been turned upside down. And chatbots play an important part in the new paradigm because they allow companies to personalise their message to fit user preferences. Data analysis can be made more or less refined, leading to proper insights in some cases which could have a real effect on the brand’s decisions and strategies.

The Mondadori Store chain is one Italian organisation that has picked up on this. Using IBM technology, the company has created a unique virtual shop assistant named MyStore. The assistant doesn’t appear in a chat, as normal, but in digital kiosks located in stores (the initiative is currently underway in Rome and Milan). It provides information on availability, price and the location of books on shelves, as well as personalised advice. The user accesses the service via their Facebook account, allowing the programme to garner data on their “likes”, friends, date of birth and more. By combining all this information, MyStore is able to generate a list of recommended titles.

Chatbots can be more than just a voice – they can also act as ears and listen to users. By monitoring content and key words in conversations, the programmes are capable of identifying the most frequent questions, the topics of greatest interest and the main issues with a product or service. Using this as a foundation, brands can create new communication strategies.

Who’s afraid of Hal 9000?

The robot pilot of the spacecraft in 2001: A Space Odyssey is cinema’s archetypal “bad” robot character. Hal 9000 is the personification of the fear that artificial intelligence technologies could get out of control and indeed start to become evil, as in Stanley Kubrick’s masterpiece. In truth, we don’t even need to make references to the big screen to show that it’s best to walk before we can run when it comes to machine learning.

Look at Microsoft, for example. After hitting the nail on the head with Xiaoice, the Redmond-based company made a huge blunder with another piece of software, Tay, which was introduced on certain social networks in 2016 to answer user questions. After an initial period in which the software simply responded based on the data it had collected, the programme started to make statements of its own initiative, imitating the way other people spoke. On Twitter, the Tay profile copied politically incorrect phrases from users and make some questionable – perhaps racist – remarks about Hitler and 11 September. Microsoft was quick to apologise and silence a technology that clearly was still in need of refinement.

There’s no easy answer to the issue of ethical responsibility or how to control artificial intelligence. Many companies have decided to team up chatbots with human beings who are ready to intervene if needed. In some cases, the software can independently deal with Frequently Asked Questions (FAQs), before passing the baton on to conventional customer support when necessary. It’s an approach that works outside the corporate environment too. The charity Europa Donna Italy, for example, is about to introduce an automated chat feature to offer advice and support to women fighting breast cancer. In the event that the software is unable to properly deal with the request, a doctor or another competent person can step in. AIMAC, the Italian Cancer Association, has done something similar, using IBM’s artificial intelligence technology Watson to create a virtual assistant names Filos, who gives information to patients and family members.

Chatbot tips

Where to start? First of all, it’s vital that you don’t wing it. Any thinking on what types of technology to use should be inserted within a wider marketing strategy, in order to find the most suitable solution for the target audience, company image and sales objectives. And given that virtual assistants are tasked with interacting with the public, they’ll have the delicate responsibility of being an ambassador for your company and brand. Here are a few things you should keep in mind…

  • Start with data. The databases you link your chatbots to will become a kind of encyclopaedia for the technology, while the data used to create the artificial intelligence algorithm will define its area of competence and reasoning power. New skills and abilities can then be added to the software a little at a time with each interaction.
  • Create a character. It’s not compulsory to give a name or an appearance to the chatbot, but if you do, make sure you choose a character that fits with your image. The choice between a virtual assistant that’s funny or professional, male or female and so on will have an impact on brand perception. A short, easy-to-remember name is usually best.
  • Allocate tasks. In very few words, problem solving is at the heart of the nature of chatbots: things like answering FAQs, providing technical support, receiving complaints, dealing with warnings and other customer service activities. They can send messages with varying degrees of personalisation – news on events or new offers, promotions, contests, service notifications and so on. Therefore, it’s important to decide what kind of tasks you can give a chatbot to help with your CRM strategy.
  • Find a home for your chatbot. Where is your chatbot going to live? Think about where, how, when and why your clients get in contact (or may want to) and, based on that, decide whether you’re going to provide support on the internet, on your Facebook account, on Skype, on a special app or across a combination of different channels. The most important thing is that the chatbot is always available where it can be of use.
  • Turn words into actions. Once you’ve defined your strategy, you need to actually build your chatbot. You could use external services or a system integrator, or make use of DIY platforms to build your chatbot without having to write code (ManyChat is one example). This are also different types of ready-to-use software organised according to type (, for instance).
  • Don’t leave the chatbot to its own devices. Monitor the behaviour of the software and keep the databases it uses constantly updated, including CRM and product/service catalogues. Set the boundaries for the programme’s scope of action – the limits after which humans need to intervene. Measure user satisfaction and don’t ignore signs of discontent.