AI : Coming of age?

08 December 2017

Now I’m a Computer Science graduate, and have been aware of Artificial Intelligence (AI) for decades, but it transpires that for a raft of reasons AI is really starting to become mainstream so I thought it would be interesting to think about how to apply it to a “normal” company. By “normal”, I’m excluding hi-tech or high data dependent companies, more your average midmarket UK company just going about its business.


The definition of AI according to Wikipedia is “…intelligence displayed by machines in contrast with the natural intelligence (NI) displayed by humans and other animals”. Clearly the intelligence demonstrated in humans is fairly diverse in scale, and AI is similar especially where scoping definitions come in to play. One mans AI is another mans “just computers doing their thing”, so despite everyone getting excited about AI at the moment no-one really has a clear definition of what people are talking about. Reminds me of the cloud-hype about 7 years ago.

That said, there are a few consistent pre-requisites to AI being useful. Volumes of analysable data is one of them. For a machine to become intelligent, it needs analyse a set of data points, learn some conclusions, then apply those conclusions to future data points with a percentage certainty score. With only a small amount of data available, this process becomes very flaky, but clearly for decades society has been creating and digitising data at a ridiculous rate, and this is giving “the machines” enough material to chomp through to apply to future scenarios.

So, all very interesting – what’s it’s practical application? This is very much a question that only the person asking can answer. The ability for a machine to learn from a series of data points and make an “intelligent” guess as to what happens next can be applied to pretty much any scenario. Clearly much of the current discussions surrounds common human based tasks in the consumer world, such as driving a car, which has over a century of human learnings to learn from. “If its wet, slow down”; “if you detect someone in front of you, slow down or stop”; “if you are running out of fuel, GPS to the next fuel station” – these are very easy to understand and good applications for consumers.

But a “normal” midmarket company, such as an lawyer or a manufacturing plant, is a bit more difficult to find applications for. Especially if a commercial case is required to drive investment. AI technology is now very accessible for relatively trivial amounts of money, but the skills to make it work continue to be a premium in the market, and a set of resources to implement / train a machine to perform a complex business process more efficiently than an experienced worker can be a tough project to justify. Any project justified on improved productivity is clearly a leap of faith too, especially as we are at the early doors of this AI industry.

I think to make any “normal” business worthy of an AI implementation, question one is “how much accessible historical data do I have that is of use?”. The legal profession is a clear example of where this could get very interesting. Notwithstanding that much of the historical legal data is inaccessible (on paper, in a leather bound file, within a walnut panelled cupboard, locked with only one key etc), once this is digitised there could be a world of applications that would make for interesting use cases.

Hypothetically a law firm could ingest their entire contract library. They could set some rules around reading the words in the contracts, such that the AI machines could “learn” when a contract is well worded or not, has specific clauses, and probably associate a score to it around risk. Potentially this could be compared to case law such that you could proactively approach your legal firm clients with an automated statement on “your contract contains this clause, which might cause you a problem – for a mere £1000ph I could fix that for you”. Clearly I’m using a cheap lawyer for the example.

This hopefully provides a good backdrop to how AI can be deployed though. AI machines are only as good as the information you give them, and if your information cannot be easily given to the machine its going to be tough to train it to work. Information architecture therefore becomes a critical discipline, to gain the most from your data you need to ensure its structured and able to be processed by a machine longer term.

Most midmarket clients I work with do not have a mature approach to information management and therefore are going to struggle with leveraging AI technologies to drive their businesses forward in the short term. The more progressive ones are already looking at their data stores to see if they can use them, and these are the guys who will secure competitive advantage from AI. For the rest, I don’t see AI as being a major thing, probably for at least another 5-10 years. Unless the front runners leverage the technology so effectively that they consume their laggard competition.

AI is being hyped up by technology marketing departments who have run out of steam around GDPR. It genuinely has the potential to be very influential in a number of aspects of business over the coming years and beyond, but probably only for businesses who are willing to embrace it fully and make it part of their business strategy. The rest won’t be immediately effected, so a cautious enquiry rather than an “all in” strategy would probably suffice. Ignore it at your peril though.

Cue "token robot stock image" for the LinkedIn article. How cliché.

We use cookies and other technologies to allow us to remember you, improve our service and display relevant ads to you. To accept cookies, continue browsing, or view our Cookie Policy to find out more, including how you may withdraw your consent.