One of the key insights provided at Gartner Symposium concerns trends around the investments that CIOs expect to make over the next 12 months. High on the list this year is business intelligence, including so-called ‘smart data discovery’ and artificial intelligence technology.
Those attending the event would be forgiven for thinking that business intelligence technology stands at something of a crossroads. With usable, enterprise-ready artificial intelligence applications just starting to come to market, it was never going to be long before commentators began thinking about what might happen if businesses asked AI a different type of question – not ‘what is the meaning in this data?’ but ‘how can we increase the bottom line in our business?’ Listening to some of the sessions at this year’s Symposium, the casual observer could be forgiven for concluding that this is the future of business intelligence.
That, however, would be the wrong conclusion, at least in the medium term. Whilst artificial intelligence is extremely good at sifting through unstructured datasets to make sense of the information that they contain (and it will only get better), it cannot yet understand the extremely complex trade-offs that go into running and growing most businesses. In order for it to make a useful, meaningful contribution, it needs to be taught about how the business is run, and how it interacts with its stakeholders and the wider economy, as well as the substance and context of the task it has been given.
Artificial intelligence is very good at drawing a rational conclusion from a limited set of data with a rational basis. However, the leaders of a business must cope with information that has no rational basis – the behaviour of consumers, of markets and of regulators is not always rational or predictable. Similarly, the leaders of a business must work from a limited set of data – the data relevant to the questions they need to answer – but in selecting those datasets they must consider a very wide set of questions about the future performance of the business.
This process, of working out what business questions and data are relevant, and interpreting the answers effectively, requires experience, insight, instinct, and understanding of the organisation’s appetite for risk. Artificial intelligence cannot appreciate the nuances of this process, so, until we reach the era of the artificially intelligent CEO, artificial intelligence will only be able to contribute to an answer, not provide the answer itself. If we model the business mathematically, as a tool to support decision-making, then artificial intelligence could certainly contribute to some of the elements of that model (for example, by analysing customer buying behaviour). However, it could neither create the business model, nor interpret it. That is why, for the foreseeable future, effective business intelligence systems will always require an underlying model created to reflect a business and the priorities of those who govern and run it.