Divided opinion & common themes, on applying AI, from #FSFConference
Yesterday morning, I had the pleasure of joining (maybe almost 100) FS leaders to talk about AI.
What interested me at this event was both the different emphasis (almost to the level of divided opinion), as well as common themes. I would, simplistically, separate them into the ambitious technophiles & the commercial savvy customer insight leaders.
Now, the FSF Club operates on Chatham House rules, so I won’t be quoting any one person or organisation directly. But, sufficient to say, I firmly align myself with the insight leaders, in contrast to continued AI over-hyping. Here are my recollections, of key points that landed with me. I hope you find them useful.
AI: The endless march towards The Singularity
Our first speaker was a futurologist. Firmly in the camp of technophiles, he laid out a vision of an AI dominated future. As well as, generously, giving away copies of his book – he raised the need for education. Rightly diagnosing a lack of (even digital) literacy, across large organisations. He proposed both education on emerging technologies & planning across 3 ‘time horizons’.
Those timelines were short-term (1-2 years), medium (3-5 years) & long-term (up to 10 years hence). The idea being to have planned a mix of optimising current operations, innovating new products/services & preparing for a digitally disrupted new business model. There were a number of good ideas here, but too much focus on technology looking for a problem, rather than the other way around.
I was also a little concerned to see Data Science & Analytics demoted to being just a part of “AI“. Clearly that latter term has now become the most fashionable. Even more concerning, to my ear anyway, was the future vision of ever-increasing automation & artificial intelligence. To expand on this he predicted 7 stages of AI maturity over coming years:
- Rules based Expert Systems (already mature, been with us for over 30 years)
- Domain Specific Machine Learning & Deep Learning (current focus for most applying ML)
- Context Aware AI (chatbots and other applications learning to use data about context to select reply)
- Theory of Mind/Reasoning Machines (able to independently learn a new challenge, most advanced now)
- Self Aware machines (conscious they are reasoning, currently still a theme of Sci Fi movies)
- SuperIntelligence (levels beyond human intelligence once aware machine collaborate to learn)
- The Singularity (where eco visions of learning from planet collide with most integrated consciousness)
Apart from sounding increasingly like ideas for the next blockbuster Sci Fi movie, what is missing here is human need or customer focus. Despite sharing front-runners in AI innovation, like BlackRock’s algorithmic trading, many examples sounded like tech for tech’s sake. If we really do let ourselves be led by what technology could do, rather than what customers/society needs – then perhaps we deserve this potentially dystopian vision of the future.
AI: Having more of a conversation with your banking customers
As a change of tone, our next speaker was from a major UK bank. For a refreshing change, his focus was on creating digital assistants to help their customers. Learning from other sectors, which is often a good idea, they took their inspiration from the most popular health tracking apps. With a focus on how they can help their customers stay “financially fit“.
It was interesting to see how a conversational interface could work instead of using buttons and normal FS app information displays. Think WhatsApp look & feel, rather than your classic mobile banking app. They have started simply, with a rules based system underlying the chatbot responses, but have refined quickly & well using live pilot with opt-in customers. They’ve also partnered with others & rightly considered how to design for failure (e.g. learn from Alexa et al & have less creepy responses whenever unable to understand/answer questions).
The continued customer insight focus, rather than machine learning for its own sake, shone through. Regular feedback from customers included: “you’ve brought the cashier back!” A feeling of a more human interaction with your bank appears to be a clear customer need/benefit & I’d expect to see further developments along these lines. With conversations also providing rich intelligence to mine on demand for services/products of the future.
AI: A creative vision for more industries
In a style that I welcome from speakers, our next presenter just told stories without slides. Working for a leading global brand agency, she encouraged us to seek inspiration outside the world of FS marketing. Such broader cross-sector reflections have certainly helped at other events I’ve helped lead.
Her first story was from the very different world of high fashion & catwalk shows. AI has been applied by a leading designer, to scan their entire back catalogue, learn their style & propose new designs. None of these were produced without question (no automating the designer yet), however they did help the creative process as prompts to select from & refine. One interesting creation, that did make the catwalk show, was a dress with lights woven into fabric that displayed a colour representation of twitter sentiment. Bots scanned twitter feed in real-time, to send instructions on colours to display on dress, visually showing reaction of crowd as it is seen for first time. Perhaps a gimmick, but interesting story.
Her second story concerned the most famous chatbot in China. Microsoft’s Xiaoice is an example of AI which has learnt beyond the confines of context specific feedback. Rather, her designers originally set the objective of achieving longer conversations, through a metric of ‘conversational turns‘. Through both being shown thousands of conversations & through live conversations with her (now 40m) followers, Xiaoice has learnt the art of conversation. On average customer service chatbots achieve 1.5-2.5 conversational turns, Xiaoice now achieves an average of 23. In fact she has proven so popular, she now reads the weather on a popular Chinese morning news TV show.
Interesting examples, where again the design focus has needed to be human interaction, as well as learning by doing. Taking up the theme of designing for failure, the programmers of Xiaoice have found a neat way out of problems. Through giving the bot the persona of a 17-year-old girl, when Xiaoice encounters a problem, or cannot understand what to say next – she just has a strop. She pouts and refuses to answer. Very human! 🙂
AI: Building it into all your products
Next was a genuine technology giant, from the FAANG Group of tech giants, it was interesting to heart their take on use of AI in FS. Focussing on how they can help everyone find & understand the information they need online, they have 3 themes to their development in this area:
- Assistance (providing more help & improved interfaces for people)
- Immersive Technology (embedding into our ‘real world’ & providing worlds that feel real)
- Machine Learning everywhere (adding intelligence into all their products).
As well as the inevitable corporate videos & product demos, there was some interesting thinking here – that will no doubt impact all our lives.
Examples of Assistants were both voice activated & text conversational interfaces. Because of security concerns, many FS customers prefer text communication with their bank. Tech providers are likely to provide the intelligence for either & more interfaces. An example pioneering in this area is Progressive Insurance in the US. They already have a chatbot that can be called up by popular assistants, to answer product or service questions & provide smoother interface to buy if relevant. Good to see design focus is on providing useful expertise first. They also see Smart Home developing towards Thoughtful Home, which offers promise for not just more comfort but also improved risk management for insurers.
Immersive technologies are mainly a combination of either Augmented Reality (AR) or Virtual Reality (VR) apps. Examples included insurers using drones to survey flooding & then being able to explore that space using VR. The more prosaic, but useful AR apps, like Halifax HomeBuyer and others letting people just point their camera & get relevant information from online. Expect to see this form of search growing. A more personal & important application was use of VR to enable families to hold their own fire drills. A chance to ‘test run’ family members ability, to react appropriately & use correct exits etc, in the event of different fire scenarios. When you think about it, it’s odd to go through fire drills so often at work, but not run them at home.
Finally, Machine Learning everywhere, basically meant a bit more intelligence in all their applications. As their product set is so ubiquitous in most of our lives, this will inevitably educate & set expectations amongst your customers. Like others, they are making more of their developments in machine learning available to developers. There was also an application that many leaders will welcome. Ocado is using ML to scan the thousands of emails/messages they receive, in order to prioritise those that need urgent action (like rescheduling delivery). If AI can help overcome the curse of overweight InBoxes, it really will help people live happier lives.
AI: Better marketing through wiser usage
Our next presenter brought a lot of marketing experience & pragmatism to this work. As a partner in a marketing/brand/growth consultancy, he reminds us that money is wasted too. A salutary reminder was that over $400bn is wasted annually in digital projects that go nowhere and this is happening in AI innovation spend too. He demonstrated this through both good & bad examples of such investments. Positively, the step change in the effectiveness of Google Translate when on holiday (from translating words to translating meaning of sentences etc). Negatively, the collaboration between Microsoft & a leading US Cancer Hospital that ended up being cancelled (wasting $60m) even though the algorithms work.
So, again, this was a plea to refocus on customer need & commercial relevance. Don’t ask, “where should we invest in AI?” Rather ask, “how could we better meet this identified need?” A useful reminder that navigating organisational politics, corporate governance & processes for releasing funding – are challenges to be faced in large organisations. As I teach on my Softer Skills course, stakeholder management & influencing skills are essential parts of the toolkit for success, even as a Data Science or AI leader.
He also included some interesting research on the preeminence of ‘brand relevance‘. Listing the top brands globally on this basis, brought to life how they each have clear jobs they help people get done. Relevance, in the mind of today’s consumers. Commercially, it was interesting to see the results of correlation analysis against 10 year CAGR. Brand relevance had a higher correlation than NPS, to this key metric of financial growth. Perhaps it is time to focus less on being a frictionless experience & more on insight generation to ensure you are relevant to people’s lives.
AI: Will the real Data Scientist please stand up
Our final speaker is, arguably, the most experienced Data Science leader working in Financial Services today. Certainly within the UK, his one of our leading lights & fortunately a wittily engaging speaker too. Currently working at a leading global insurer, he comes from a background in academia, agencies & other sectors as well as FS. Amusingly, he started by sharing 15 things about his work, family & personal life. Demonstrating how, now we knew that much about him, it would be much easier to have a personal conversation in the break. He then challenged us as to how much we know about our customers & how little this knowledge is used in our communications & interactions with them. Most FS interaction is still far from being a conversation.
It was delightful, to hear this very technically capable gent, advocate the importance of staying focussed on customers & insights. Simplifying the world of potential Data Science applications, to enabling “conversations at scale“. For insurers it was also a wake up call to value how much data people share with them. As a large composite insurer, own principle they might know the value of car on drive, home security (or not), value of possessions in-house & when people are away on holiday. Imagine who you would confide in, to know all those facts? Insurers are trusted by their customers, so greater transparency, protection and utility from their data is an outcome from GDPR to be welcomed.
Turning to the need to avoid ‘being creepy‘, he also shared the infamous Uber “rides of glory” advertising. How could such a provider think it appropriate to spot a pattern of potential affairs & share such analysis online. Applications of AI need to be for the benefit of those whose data they use. Revealing more of their challenges internally, he reminded me of the reality for many of my clients. Far removed from the hype of future cyborgs, there is much useful work to be done today on comms. Coordinating (and normally reducing) the number & timing of different communications customers receive. Only sending those you learn work for customers & improving personalisation step by step. Another reminder of the importance to earn the right to not be “unsubscribed“. Here AI can be intelligently used to stop being stupid & demonstrate the organisation remembers what customers have told them before. Why not making your communications more relevant & needing less of them – a goal for your application of AI?
AI: Were you at #FSFConference?
I hope those reflections are useful & prompt your own reflections on how to respond to AI opportunity.
If you also made it along to this event, or others on the same topic, I’d love to hear your feedback. Are you with the disruptive technophiles or the commercial customer insight leaders?
How would you advise your CEO if they asked you where to invest in AI?