Careful understanding and planning are needed for businesses implementing AI
We all know that Google, Facebook and Amazon, are investing heavily in AI, with results widely reported.
It’s no surprise that businesses built around collecting and analysing data are leading the way in AI. But what about more traditional businesses? Businesses built in the pre-digital age, based on physical products and infrastructure. Business that underpin transport and energy, or that develop new medicines or materials. How should these businesses take advantage of AI?
Such companies, who survived many years without investing in data, are now going through the process of digitalisation and the value of their data is increasing rapidly. Not only does their data hold value in its own right (maintenance schedules for example), it holds additional worth as data that can be used to train AI to deliver powerful, business transforming insights that add further momentum and value to an enterprises’ digital transformation.
But the AI opportunities presented to digitalising companies, are very different to those being harnessed by the likes of Facebook and Amazon, which we will call digital companies. The insights that can be generated, the type of data that must be utilised, and their adaptation to the digital world, are all very different.
AI for digitalising businesses
A key difference for physical, digitalising business relates to the type of data they collect and what they can do with it. Their data is largely about real-world processes critical to operation: jet engine performance data, transport schedules, chemical processes in R&D pilot plants.
Digitalisation is not just about AI – there are many tools to generate value from data – though AI is the jewel in the crown because of its extraordinary disruptive potential.
Examples of AI applications being developed by digitalising businesses include: automating routine maintenance, spotting signatures which predict component failure to know exactly when to take a machine (such as a plane or oil drill) out of service for maintenance and predicting successful active-molecule formulations to reduce costs associated with R&D in pharmaceutical, materials and chemical industries.
These applications are very different from the data collected by digital companies, which is largely on consumer behaviour. This matters when building a precision AI.
Digital businesses can often accommodate some imprecision in AI. In consumer marketing, it is reasonable to run an AI on vast data sets, unsupervised, without understanding the data well. It might find that people who tweet about shoes respond to shoe adverts. If targeting shoe tweeters leads to 10% more sales, it does not matter who within that group bought – just that shoe buyers are a statistically significant subset of shoe tweeters.
On the other hand, if you are using AI to predict when an aircraft engine component might fail, you need certainty. Knowing 1% of planes might fail isn’t very good – you need to know exactly which ones and when.
Using AI to make precise predictions requires a rigorous approach to data and AI training regimes. You cannot let an AI loose on all the vast unstructured data from your engine measurements. You need to turn it into rigorous training data, tagged and curated by experts who understand what the data is telling them (in this case what physical changes indicate engine failure).
It’s easy to forget in the digital age that data represents the real-world, an object, a physical property, an event. Understanding data needs people who understand what it represents – material strain, temperature readout, chemical reactions. It is these people who can apply this knowledge and rigorously apply it to design effective AI training regimes.
Getting people on board
A second major difference relates to people in the business. Digital transformation is just that, a transformation. Getting thousands of people to change their habits does not happen overnight. Transformation must be done at a pace that brings people and processes with it, or it will fail.
Digitalising businesses should avoid launching into all-encompassing AI initiatives without careful understanding and planning of what they need to achieve. To advance this practically and effectively they should start by focusing on quick wins, building AIs to address well understood, immediate opportunities with a clear impact upon operations.
Multiple small AI projects can be done in parallel ensuring the best ideas are progressed rapidly, with bad ideas allowed to fail early. This ‘fail fast’ mentality is the cornerstone of how the digital giants explore new ideas to deliver innovations and is one area pre-digital companies should look to emulate.
Real-world results build confidence and trust in AI and calms fears about new technology, giving AI programmes momentum, securing buy-in from budget holders, and excitement from the employees who interact with the results.
Lessons and pitfalls from the digital giants
An enterprise beginning its digital transformation starts its journey from a markedly different place to a digital company, and realises value in different ways.
There are lessons from the digital companies. Some can be directly applied, such as the ‘fail fast’ approach, the use of A/B testing, and perfecting user experience. Others can be adapted, such as the focus on people over technology and platform, though digitalising companies should demand much greater sector experience and understanding of what data represents.
But many aspects of the approach of pre-digital organisations will be very different, including fundamentals of what data must be collected and how it should be used to train AI. Equally the process of getting existing staff on board with using AI is critical to success, something new companies with young digital native employees have less trouble with.
AI is an opportunity for established companies to continue to lead in the digital world. But it also brings the threat of disruption from competitors, start-ups, or digital companies trying to muscle in on areas they think they can do better. If they are to resist this threat, physical enterprises must harness the disruptive potential of AI for themselves, but they must do it in a way that works for them.
This article is based on principles laid out in Tessella’s new whitepaper: Maximizing Value from AI: The digital transformers’ guide, which outlines 11 steps that pre-digital companies must take if they are to drive growth and stay competitive with AI.
Orginal source: D/RUPTION