The rise of the machines is starting to look more like reality than science fiction. But as always the introduction of machine based intelligence into our lives will be somewhat more mundane than the images portrayed in Terminator. In a business setting we are already starting to see machine intelligence play a significant role in many operational activities. Sales, marketing, credit assessment and fraud detection are just some of the areas where machine learning is being applied, and eventually no stone will be left unturned. To get a grasp of where all this is headed we need to look at the big picture.
For half a century most businesses have used information technology to streamline their transaction based activity and business processes. This has all been done in the name of efficiency, and if transactions and processes completely defined a business the job would be all but complete. However there is an important – in fact the most important component of business activity missing here – the decision. Business is an inherently uncertain domain. If it wasn’t we could get our spreadsheet geeks to create detailed schedules and then retire to the beach, golf course, or whatever your opium happens to be. But uncertainty ruins our potential nirvana, causing us to deal with unpredictable circumstances and outcomes every day. Ultimately we deal with uncertainty by making decisions, and these decisions are always made with incomplete information. What if it were possible to get computers to handle many of our operational decisions? Examples abound. Which customers should be given credit, which machine is likely to fail soon, determining whether a transaction is fraudulent, which prospect is most likely to close this week – and so on. Most of these things can already be done by computers far more effectively than humans – sad but true. So the current fascination with all things ‘analytical’ is no accident. We are inexorably moving to a scenario where success is determined by the efficacy of our decision making processes, and the degree to which we can automate decision making.
So let’s talk about some of the technology. In an industry that creates a new buzzword every 15 minutes, it becomes difficult to define exactly what half these buzzwords mean – but that’s marketing for you. Here are some of the essential components needed in our brave new world of decision automation:
- Lots of data (I was recently told that the term ‘big data’ is so 2014, and so I won’t use it)
- People who know how to pick out the most meaningful parts of the data – sometimes called data scientists.
- Algorithms which pick out patterns that show repeated behavior, so that future behavior can be predicted. This is variously called data mining or predictive analytics and even machine learning – take your pick.
- Mechanisms allowing these patterns to be used in everyday situations.
- Optimizing business operations on an ongoing basis as business conditions change.
The first thing to say is that some of this is complex and needs highly skilled (and highly paid) people to deal with it. But business managers have another route – buy a solution, or more common these days, use a cloud based service – and there are lots of them. For example InsideSales provides a cloud based service that will outperform your sales reps almost by an order of magnitude when deciding which prospects are worthy of a sales rep’s attention.
There is another dimension in all of this that I have not mentioned yet – most decisions are made in real-time, and so our data flows and machine intelligence needs to operate in real-time. A batch run every week just will not cut it. And so there is a great deal of interest in streaming analytics – applying analytical techniques to streaming data. Again, organizations can choose to pursue a DIY route or buy a solution.
What will be the result of automating our decisions? Well your CFO will be delighted to know that more work can be done by fewer people. People will increasingly be used to handle exceptions, but since most decisions will be handled automatically there will be obvious efficiency gains. More importantly however, these decisions will tend to be more accurate, and so business efficacy is also increased – somewhat more difficult to put in a spreadsheet, but real all the same.
All of this is happening now, and the rise of the algorithm is probably already happening in your business (if it isn’t it might be time to jump ship). However we have been talking of operational decisions – the high volume, relatively simple decisions that are made day after day in all businesses. How about tactical and strategic decisions? Well at the current time we use different technologies to aid our decision making processes – mainly in the form of business intelligence. Eventually this will become partly automated too, and five years from now BI will look somewhat different from how it looks now. It will be more automated and intelligent – not just a platform for slicing and dicing data, as it is now.
But we haven’t finished yet. New types of intelligence will be built into our business applications that are not even a glint in your favorite technology supplier’s (an oxymoron?) eyes. Game theory provides a framework for making good strategic decisions under uncertainty. A few commentators are beginning to mention it, because it is a next logical step. It won’t be called game theory, some marketing whizz will give it a much more sexy name – but it will be game theory all the same.
All of this will take decades to complete, in pretty much the same way it took decades to bring our transactional activity and business processes under control. These are just the fairly clumsy early days – ten years from now we will truly be living in a brave new world. Whether it will be a better world is up for debate.