So bots are the hot tech of the moment. Initially I couldn’t understand what all the fuss was about. It didn’t help that I kept seeing examples which made me want to avoid bringing this into my life at all costs!
The penny finally dropped for me once I started ignoring the clumsy examples being demoed, and gave myself space to think through some of the fundamental implications of the technology.
Is it going to guarantee instant sales growth? Unlikely.
Could it start solving some genuinely meaningful problems? Absolutely.
My FirstBank account
To explain my thinking, I’ve put together a fictitious chat scenario where we can examine the nature of this type of engagement, and some of the problems it can (hypothetically) solve.
The first thing to notice is the interruption. With a chat bot you’re not receiving an message alert in another conversation which you can switch to in your own time. It’s jumping in on your current chat stream.
So the sentiment analysis on the current conversation is extremely important if you don’t want to risk looking ridiculous or annoying.
Of course, there’ll be validation issues behind this type of data integration. It’s an old old problem but perhaps we have now an opportunity to correct inconsistencies in a central location. Maybe! In theory though it’s an easy win for all parties.
Now things get a bit more interesting…
Please bear in mind that this is just conceptual but what I’m describing here is the ability for organisations to leverage existing data sets to generate an accurate opinion of someone.
The messaging application doesn’t have to reveal its raw data, but can allow algorithms to analyse the data to generate a meaningful score on someone. *See addendum for issues around this approach.
Additionally, the vendor can potentially avoid some of the hassles of storing and protecting customer data.
Let’s keep going…
So what’s going on here? Well, video can be much better for identification purposes, with some older studies demonstrating 95% success rates. As camera phones improve in quality and computational resource becomes more available, this will only get better.
More recent innovations have been able to use a combination of sound and video to create 3D models of the subject.
There are issues with facial recognition processes, like lighting and make up, but we should bear in mind this is just a contributing element to a wider confidence score of identity. It’s the combination of all the mechanisms (multi-factor) used in the enrolment process that move us towards higher levels of certainty.
Let’s explore further…
So now we an advocate for our new recruit who also happens to be an existing customer with a track record. This will again contribute to our confidence rating, possibly of both parties.
Now we’ll finish up…
Yes, it’s a contrived example (and my playwright career is not likely to take off soon), but the intention is to provide a clear demonstration of how past barriers to customer experience have been removed. When used appropriately, we can offer a genuinely slick experience onto complex processes.
And the reason that this technology might achieve rapid adoption is the fundamental win-win nature of the interactions. All parties in the above example gain from the transaction:
- The messaging application (Facebook) can monetise the keywords selected by providers which trigger the interruption. They also keep the user in their application for longer, which has always been primary concern for apps.
- The vendor (FirstBank) is able to detect and act upon a strong lead, which ultimately leads to acquiring a validated customer.
- Tom gets a new bank account with minimal effort.
- Sue gets a bonus reward for helping a friend.
I’ve deliberately chosen a banking transaction for my example because I think service based products provide a genuinely valuable use case for bots. These types of engagement can be time intensive and although the process might be well defined, it often requires some level of bespoke engagement.
For the banking industry specifically, I think chat bot technology speaks (pun intended) to their key concerns:
- Cyber security and mitigating fraud is in the top 3 challenges for banks
- Mobile is regarded as a critical customer channel, with nearly a quarter of banks seeing it as a strategic priority
- Enhancing mobile and online channels is by far the top priority for enhancing customer experience
It’s hard to imagine that these same concerns are not top of the agenda of most industries, but before rushing into bots on the back of all the hype, we need to assess any initiative from the following perspectives:
1. User engagement needs to be sensitive to both context and sentiment
a. The engagement has to be appropriate to the context of messaging. You’re potentially interrupting people on a private channel, and this has trust implications for both your brand, and the messaging service itself. Care should be taken not to consider this as just another channel to flog shoes and sunglasses.
2. Collaborative partnerships need to be negotiated
a. There’s a definite opportunity to leverage the detailed data held by the messaging service itself, but of course they won’t be giving this away. So new ways of sharing this value need to be negotiated, where all parties benefit without losing IP or trust.
3. New techniques are required to track user behaviour
a. The primary benefit of this technology is that we’re able to engage the user where they already are, i.e. in the messaging app. The downside of this is that it can be much harder to track activity than with traditional social media. The temptation might be to use hyperlinks to direct the user to a website so that behaviours can monitored, but this breaks the model and risks users becoming wary of any engagements with your brand.
So despite my initial scientism, I think with attention to these three concepts, bots can continually support your brand with personalised and trusting engagements.
And of course, the cloud native technologies mean it’s all delivered on a planet-scale.
Since writing the above post, I came across the following news report that Facebook has policies banning the use of its data for customer scoring purposes:
Firstly, I fully accept that this is a contentious issue and feel that my second tenet of collaborative partnerships highlights that we entering into new territory, and we need to find new ways of working that are in spirit of providing the great customer service.
With that in mind, I would say that the Facebook-Admiral case simply reflects the conflicting incentives within the insurance industry. Greater customer insight might be desirable for the insurers, but it certainly doesn’t help the customers if they can’t get insured for things for which they need protection. Essentially, this scenario is not a win-win, and Facebook has quite rightly put up some resistance.
In my fictitious example above, the bot is being used to establish identity. This is a core requirement for the product being offered, and the technology is simply removing the friction and bringing added certainty to an existing, regulated process.
I feel the underlying purpose of the bot is critical for all parties, otherwise is just won’t work.
It’s all very interesting and will continue to evolve. My goal here was simply to stretch people’s opinion of bots beyond them being just another channel to push consumer products onto a resentful audience.