Who is your “assistant” working for?

Will conversational machine interfaces ever work on behalf of consumers? AI-bots are currently a big thing for enterprise looking to manage their interactions with consumers, but is the environment right for these “assistants” to turn the tables and go to work for us.

Who is your “assistant” working for?

A big chunk of my life is spent leading a telecommunications company. It develops software that handles phone calls and other kinds of real-time communication implemented by businesses.

In the contact centre world “AI Bots” seem to be a big thing at the moment. The idea is that a business can reduce both it’s costs and customer friction by servicing customer interaction at least in part as a conversational human to computer exchange.

The attractions for the business are fairly compelling, at scale they can reduce their costs even below the current $single-digit per hour offshore benchmark for taking inbound telephone calls. Folks at the younger end of their customer demographic hate interacting by talking to them anyway. Robotic customer service agents also have the added advantage that a software based agent is unlikely to go off-script and threaten to bomb their customers if they don’t like their tone.

Consistent, risk free customer interaction at near zero marginal transaction cost is how the technology is hyped. Of course the practice is a lot harder than the theory and quite a few folks, even in the industry, don’t understand the difference between machine deep learning, and semantic language analysis with scripted responses (hint: the first could still actually threaten to bomb your customers if you use the wrong training data and/or expect too much).

Notwithstanding that, there are clear and compelling benefits to outsourcing conversations, or at the very least the start of them, to software in order to concentrate human resources on the place they add more value. There are several blog posts if not an entire book in the best way to do this but this summarises as be transparent about the nature of the “conversation” and have a clear escalation path available for when you hit the limitations.

This is all pretty mundane stuff, but the thing that is really interesting to me is whether the idea of a robotic customer interface sitting inside individual company contact centres has any long term future at all.

Enter the assistant as a consumer agent

I’ve recently been playing with the Google Assistant API. If you aren’t familiar with this it gives app developers the ability to extend the “OK Google” functionality built into it’s Android smartphones, tablets, and Home smart speaker systems to add the ability to interact with other external services. Using it, I can extend it’s functionality to turn on and off devices in my home, even join chatrooms in my favourite messenger app and interact with them by voice rather than onscreen.

More importantly enterprises can now extend the functionality to add “OK Google connect me to my insurance company” type phrases to interact with their consumers. I confidently predict that a large segment of users will love that functionality and positively embrace the fact that they are talking to an automated agent. Especially if it turns out to be a very smart one.

It is massively in Google, Amazon, Microsoft and anyone else who wants to compete in this markets interests to maintain the quality of their product by preferentially connecting to services that provide high quality transactions which are valued by their customers. They know a lot about their users and will therefore probably apply their own big data sources and consumer knowledge to mediate these conversations to the greatest extent that they can (although competition and data protection regulators may throw a spanner in the works).

Of course they could also choose to monetise the transactions which they are in a position to do due to the way their assistants are also gatekeepers. We are used to the way that plays out in things like search and there is no reason to suspect that it won’t turn out the same way for assistants. The provider of the best results for the consumer will achieve the greatest success with no excessive incentive for “greedy” short term monetisation maximising strategies.

So where is this going? maybe:

Hey Google my Broadband is really slow today?

Which could one day get the response

I can tell that your Broadband Provider is TPC, they have checked your line and can see a problem which is restricting your speed. They need to book an engineer visit to replace your router and I can see that your calendar is empty on Wednesday the 26th in the afternoon. Would you like me to offer it to them or shall I book a different date.

So far so good, the “customer agent” aspect of the assistant has delivered a win for the consumer and a win for the supplier.

Customer side agents may not be exclusively a win for providers though. It may also choose to add:

Alternatively I can see that the average peak time speed of TPC customers in your area is 8Mb/s. The best provider in your area is Gigacomms who provide 1Gbit/s fibre to the premises for £xx a month. Would you like to order a Gigacomms install and cancel TPC instead?

This may seem far fetched, commercial and legal reality could well get in the way. Even so, it isn’t clear to me that machine interactions are going to be exclusively about suppliers using robots to “manage” their consumer interactions to keep them at arms length. Maybe some of them will be the other way around, which could leave them being careful what they wish for.