The idea in a nutshell: Anyone building a bot is going to have to consider the idea of personality. It’s unavoidable. I have been thinking about it and it appears that, with bots, personality = conversion rate. That means we will wind up needing to optimise it like any other UX and that could lead to a ‘Charms Race.’

Is every bot a sales bot ?

In am about 80% sure that it is fair to say : Any interaction between a bot and a user, when conducted within a commercial context, has a sales component to it.

Most websites have 3 purposes. They help users in :

  • Evaluating a product: In this scenario, the bot might be tasked with answering sales and product questions. They could assist users with the emotional and rational product enquiries people need to have answered before they will buy.
  • Buying a product: The bot might help with enquiries on how a checkout process works. I can imagine a bot assisting with form completion, explaining how the procedure works and why what the user entered failed the field validation.
  • Customer Service: Post sale, customers have queries and problems using the product or service they bought. During this period, the bot could be trying to assist a user with a service query. The company goal here might be to help the customer as efficiently as possible to save money. They may also want to service the customer well, so they remain loyal.

The goal for any private company, beyond satisfying the immediate intent that a user has, is to make money out of them. That applied to people whichever of those 3 things users are doing.

Those creating the bot wants to persuade users to research, buy or self service on line as cheaply and as successfully as possible.

On a website, you’d call this ( the desire to make these interactions as cheap and successful as possible ) a conversion rate. Teams tend to try to optimise conversion rates through CRO – Conversion Rate Optimisation.

The analogous concept ( to CRO )  within a bot world – is persuasion. Persuasion is an aspect of personality,  delivered through things like rapport, active listening and influence. Influencing users do what the company wants through a bot will include things like vocabulary, tone, pitch, cadence, questioning styles and so on. The more persuasive the bot, the better the conversion rate.

If that’s true then isn’t the natural ramification that there develops an arms race ( more accurately, a ‘Charms Race’ ) between companies. Each company is trying to train their bots to be the best sales people, to improve their conversion rate. Bots’ personalities would be optimised in exactly the same way we use CRO to optimise websites now.

The First Wam War – The Impending Charms Race

‘Charm is a way of getting the answer yes without having asked any clear question.’

I’ve read many definitions of charm over the years but that’s my favourite. It might be funny to consider the idea of a bot being charming. However, if we want them to represent the company, to sell the things we have, there is no reason to suppose they could not be.

If a bot is going to do our bidding, we could usefully program it using techniques which work for charming people. We could discuss effective approaches with actors, orators, politicians, leaders, attractive people, persuasive people, psychologist, English Language teachers, car salespeople, copywriters, actors, orators, news presenters. Together, using the mannerisms, behaviours and language they employ could help us understand how to program the bots charm.

To win the war, wouldn’t we want to understand the psychology of selling, persuasive words, influencing and other tools to lead towards an outcome ? Wouldn’t we end up optimising charm (the conversion rate) in the same way we do for a normal website ? Would we end up with bot split testing and targeting…. Don’t we get to the point where we are multivariate testing conversation fragments used by the bot to see what sells best ?

They may be more charming than us in the end

Bots use AI and large data sets to learn. Over time, they could access millions of conversations and establish patterns in the use of all the persuasive elements of language that I describe above. They could learn the perfect way to convince people to do things.

It seems more than possible to me that with sufficient study in these areas a bot could emulate charm more successfully than a human could. They could get so good that they could end up being more charming than any human could. In the film Ex Machina, the AI woman they created charmed a man to fall in love with her in order to escape. She optimised her charm to achieve her goal and persuaded him to do what she wanted.

Having the most charming bot representing your company could be a competitive advantage for a brand. Charm bots could help in the same way that the commercial success of Amazon hinges on having the best conversion rate. And Amazon got there with a lot of split testing with a lot of users over the years.

How many bots are there ? Bot personalisation

If the bot essentially replaces the website and the goal is optimisation, does the bot have just one personality or does it have many depending on who it’s talking to ? In the film ‘her’ configuring the operating system involved answering a few very simple questions about the basic psychology of the user. Based on those inputs, the ‘right’ personality was developed for the user.

People accidentally / naturally emit their  personality types in the language they use, the choices they make, the colours that they click on. It’s not hard to imagine bots from huge brands having millions of conversations a day. They could learn the phrases that work not the most in aggregate, but work best for each personality style.

Personality appears to be the most important commercial consideration for a bot

Having spent some time thinking this through, rather than an after-thought, personality seems like the most important ‘feature’ of a bot.

Initial trials with a written chat interface would simplify matters.  That means we could avoid ancillary personality questions like accent, tone, volume etc.

Never-the-less, the question of personality seems unavoidable and important even in text communication. Since we’re dealing only with text we have the relatively simple questions of what role does font, bold, italics play ? It would be interesting to me if Ariel 10 sold more than Verdana 14. In a way the choice of font is a choice of personality.

At times like this it strikes me how different it was working for a small business and not caring too much about how a bot might affect your brand. Things are very different in corporate. After examples of ‘bots gone bad’ like Microsoft’s Tay (I’m not even going to link to it, everyone knows what happened) the brand impact of ‘getting it wrong’ to the big companies is all too clear.

In some senses, this is a familiar conversation. The brand is the people (and the bots). It always has been. This is why so many companies invest so heavily in their staff. Branson says doing so leads to better interactions with customers by staff – which is the experience of the brand that people remember. Personality and brand are the same when we’re talking about bot interactions just as much when designing a website.

The commercial aspect of personality seems imperative. For companies looking to improve customer service, customer satisfaction and commercial results, having a ‘winning’ (this time, literally) personality seems likely to be potentially the primary question when it comes to designing chatbot interfaces.

I leave this blog entry with more questions than answers and may have to revisit the topic.