La Fosse was delighted to host the first in a series of ‘CTOs Anonymous’ roundtable discussions, with Jonathan Midgley from the facilitating. The topic, Practical Applications of Machine Learning and Artificial Intelligence, brought senior digital leaders together to discuss the challenges presented by the rise of artificial intelligence (AI) and machine learning.

For tech enthusiasts who grew up dreaming of a world of practical applications for intelligent machines, these are exciting times. The use of AI in everyday life has rapidly become ubiquitous, governing everything from how long we spend waiting at the traffic lights to the price we pay for a pair of jeans online.  

The business opportunities presented by AI, and machine learning within it, are vast – and it’s down to Chief Technology Officers (CTOs) to apply AI-driven predictive technology for the benefit of their customers. We sat down with a selection of leading C-suite tech executives from businesses old and new, in sectors from fashion to transport, and asked them to discuss what challenges they’ve faced, and share insights around the solutions found (or the failures experienced) in their organisations when it comes to Machine Learning and AI. 

Here are their top 3 key priorities.  

Keeping it personal

“The problem with personalisation is that if you always show people exactly what they want, they never discover anything new. We’re trying to find a way around that.” 

Some of the most innovative work in machine learning currently goes into building the engines which serve personalised content. Depending on the sector, an intelligent platform can offer a customer hundreds or thousands of options, based on sophisticated assessments of what they want, where and when they want it, and how much they are willing to pay. 

Successfully presenting those options to consumers can drive sales, traffic, and customer loyalty: a buyer is more likely to return to a platform if they know it can pick out the products they want, based on existing browsing and spending data. But even the most accurate engines can serve up problems as well as predictions.  

Developing an algorithm to crunch data on the choices a consumer makes online and predict what they will need next is simple enough in theory, but tricky in practice. Focus that calculation too closely on previous choices, and you can inflate a ‘filter bubble’ around a customer, so that they are only ever offered what they know they already like. Suggestions can become obvious, repetitive, or boring. But placing too much emphasis on other predictive variables, like age, location or gender, risks losing accuracy and precision. 

Tech leaders are already tweaking their engines to address the personalisation problem. Leading news organisations know that their readers want pieces which challenge, as well as support, their views. Fashion retailers know that when trends change, so do the tastes of their consumers. So, increasingly, CTOs are looking to develop tools which will serve some options with a lower degree of certainty – products or services which buyers might want, as well as what they will want. Some CTOs are even building two or three different versions of an engine, to experiment with that balance, allowing for more speculative predictions of what a customer may want, to supplement the safe bets.   

To Spotify, or not to Spotify?

“We’re not as agile as we’d like to be. Spotify isn’t really on our radar. I’m focused in addressing our legacy challenges first.” 

As competition for talented developers and data scientists intensifies, and digital-focused organisations place more importance on agility and adaptability, CTOs are finding that structuring and staffing have become an increasingly important part of the role. Every CTO will be familiar with software development principles and practices, like Agile and its sub-set Scrum. But that does not mean that they agree on how they should be applied to the production of new AI products and capabilities. There are new debates among management: one of the largest being, to Spotify or not to Spotify? 

Spotify is famous for its adoption of an ultra-Agile working structure, in which developers are organised into autonomous, collaborative ‘squads’, with freedom to decide themselves what they want to create, and how they should go about doing it. Many CTOs have embraced the philosophy and have set about instilling a culture with maximum trust and minimal control from the top. For others, though, the model isn’t realistic.  

Keeping a team responsible for developing new machine learning tools nimble and responsive is an especially important challenge for tech leaders in older companies, where manual workforces and traditional operating models present real legacy challenges. Whilst start-up CTOs enjoy a blank canvas on which to plan their internal structures and operations, those in mature organisations are not always so lucky.  

Too much data?

“I’m not sure if sending your team away to dig through the data and look for general patterns is the right way to do it. I’d rather they were productising from the start.” 

A true CTO will tell you that there is no such thing as “too much data”. But with the amount of information companies can store rapidly expanding, it is easy to collect data simply for the sake of it – without actually considering which customer problem it is there to solve. 

The temptation for many CTOs is to take all their available data, and then create a product that can use it – but that is generally thought to be the wrong way around. As one C-suite tech leader acknowledged: “It’s easy to find yourself trying to develop an app to use the data, simply because the data’s there.”  

There is no doubt, though, that having large amounts of information to analyse is a benefit, not a hindrance – it just needs to be used correctly. For some, that means delving into the data and searching for patterns with predictive value. Others would prefer for their teams to be “productising” from the very beginning of the process.  

CTOs agree, though, that it is wise to keep asking two basic questions throughout the development process, regardless of philosophy: “What is the problem that I’m actually solving here? How am I making the customer experience better?” 

Are you leading a digital team and considering how ML and AI can help drive your business forward? Let us know what you would add to this list.