In the next two decades, the way we live and work will be transformed by data science, machine learning – and, at some point, true artificial intelligence (AI).
Data is to the modern world what oil was to the 19th century – an extraordinary resource which can make winners of those who learn how to exploit it effectively. For businesses, data science offers major advantages over other analytical techniques, for instance its predictive power, dynamic potential to learn and improve over time and the ability to integrate it into standard business flows compared for instance with a graph or slide. But as with any technique, the quality of the output will depend on the quality of the input, in this case the data, the rate of dispersion of the techniques and the level of comfort with data science in employees, customers and the wider public.
So far the applications of AI in business have tended to concentrate on a few obvious areas of gain such as detecting patterns in large data sets to identify long term trends or predict future demand. But, just as modern techniques can extract so much more from existing oil deposits than was the case even a few decades ago, modern data science techniques can reveal value even in smaller or incomplete data sets. And while deployment varies by sector and data maturity, the majority of companies haven’t yet started on this journey.
At our own business, ASI, we have done research-grade data science at the cutting edge of what’s possible, for instance working with the UK’s Home Office to develop a machine learning model that can identify Daesh propaganda with an incredible level of accuracy to enable video sharing websites to block it. In addition, we have conducted over 220 applied machine learning projects for organisations that are achieving radical improvements in efficiency from relatively simple applications of AI. This is AI aimed at solving very specific business problems such as optimising my marketing spend (and keep learning from real-time information as customer behaviour evolves); how can I predict likely asset failure so I can replace it in advance; what’s the optimal mix of products to carry on a flight from London to Prague versus Madrid to Copenhagen.
For instance, ASI built an adaptive scheduling system for a bus operator that predicted bus “bunching” – the old problem of no buses turning up, then three arriving at the same time – 38% better than the current system, enabling bus controllers to intervene. Cue happier passengers, less crowded busses, and big savings for the bus company. Over the next five years we can expect much of the value from AI to come from the gains of relatively simple AI permeating the economy until its use is standard practice. For some companies the optimum approach may be black-box AI – for instance cybersecurity solutions – but many companies will want a competitive advantage from the technology and will have to build their own to achieve that.
The challenge for CEOs, as well as CIOs, is to ensure that they are choosing wisely between building and buying solutions that will directly impact the strength of their product or service offer and their efficiency. Increasingly enterprise spending is moving away from infrastructure and towards tools and bespoke software that can deliver. How many big-data procurements lie cold and lonely today because they weren’t well integrated with business workflows?
As the market develops, we believe the model needs to shift from AI dependency to AI capability. This is why ASI’s transformation service is increasingly working with companies to build the capability they need to run AI projects internally from developing a strategy which integrates AI with the commercial strategy, the technology needed to support data science, the data flows & governance, and recruiting a data science team.
Any new in-house team needs the right environment in which to build and test its machine learning models with access to the right data sets on which to train them, a familiar interface or dashboard and which uses the, usually open source, tools with which they are familiar. That is why we have built our own best-in-class data science platform, Sherlock ML.
Some people believe data science can be done via “drag-and-drop” interfaces. We have made a strategic bet that intelligent data science is, and will increasingly be, fundamental to unlocking competitive advantage from data, and the key to all this is the right people.
The increasing demand forthe best data scientists and engineers is already creating a shortage of talent. ASI is playing a part in the solution to this through its Fellowship programme, Europe’s most prestigious programme for PhDs and postdocs moving from academia to data science roles in industry, which attracts applications from more than 10% of the UK’s STEM PhDs.
For many businesses, the application of machine learning and AI represents a huge strategic opportunity but the decision about which model to go for requires a rigorous assessment of where on the maturity scale the business is. For most business the real gains come when they move from curious, with no in-house data scientists, data silos individually held by different departments and only an aspiration for a data science strategy to expert, with a clearly defined strategy, a rounded in-house team of data scientists and jointly held data accessible by the data team.
Now is the time to ask yourself, where are you?
