OPEX’s Predictive Solutions Advisor, Colin Shearer, has been a pioneer and thought leader in AI and advanced analytics for over 25 years. In this article Colin discusses why it’s important to combine human expertise with artificial intelligence to achieve business outcomes.

Colin Shearer, Principal Advisor - Predictive Solutions

Obsessed with AI

Today there are new challenges in the industrial analytics space in the triangle between advanced analytics software, the profession of data scientists and deep domain knowledge.

The biggest trend I am seeing is that people are obsessed with…. sorry… seeing the potential of AI. In around 2010 we saw the Big Data wave which has now blended into the AI wave. Both have had a positive effect in that most senior executives now appreciate there’s potential value in these technologies. But their organisations often have no clue how to proceed, and so they embark on what end up being science projects.

Data excitement is a pitfall here – the idea that simply throwing data at smart technologies will give transformational results. It doesn’t, and that’s leaving a lot of would-be adopters disappointed.

Blind faith in AI will eventually fade, as the value of incorporating human expertise becomes more and more evident.

Focus on Business Outcomes

The combination of data, advanced analytics tools, data scientists and domain experts is powerful, but it’s crucial to have the business outcome in sight. It’s business first and tools later. In industry it may be prevention of unplanned shutdowns, solving quality problems, efficiency gains and much more.

Projects have to be driven by business problems and goals because these determine which analytics approach to use and which data to apply it to.

And you have to plan from the start around how your business will benefit. It doesn’t matter how technically brilliant your analytical work is; until you do something effective with the results and enhance your current operations it is meaningless.

Getting AI right has the potential to transform virtually every aspect of business, with improved decision-making driving better outcomes across the board. This is also the case for industrial analytics, where it is possible to embed analytics in the process.

Real Time is too Late

In manufacturing and industry real time is too late. A lot of cost and effort has gone into deploying condition monitoring, but the end result is usually a blizzard of alarms that surface when the equipment is already seriously damaged, often too late to prevent unscheduled downtime.

It is so much more powerful and valuable to get an early warning hours, days or weeks ahead, in time to adjust and avoid unplanned production failure or quality issues. And today, the streams of data painting a highly complex picture of system and process behaviour give you the possibility to predict and prevent through the knowledge of correlations and causality in data.

To me it is quite obvious that success and value in industrial analytics comes from bringing together a relevant business problem, the scientists or engineers knowing the process and product, and the data scientists with an advanced toolbox. While you may now utilise big data, sensor data and a whole array of new analytics capabilities, you cannot get the outcome you need unless you include the layer of human expertise.

Big Data is a source for improvement for companies, but its value is only released when it’s combined with human expertise.

Machines and GDPR

In contrast to the challenges we currently face in other application areas for advanced analytics: machines do not care about GDPR! So adding sensors, adding more and more streams of data to be exploited with analytics is the way to go. The key point is, though, that you can’t reap the benefits unless you analyse in the context of your domain knowledge.

Today’s analytical technologies can deliver new levels of knowledge, and are key to more automation, greater predictive power and significant value contributions. But you need to align these powerful tools with scientists and engineers who understand the complexity of the systems and processes being analysed.

The Industrial Revolution of Analytics

Just like the way manufacturing was transformed, the craftsmen in analytics will also face an industrialisation. Stand-alone one-man analytics doesn’t deliver results efficiently and it doesn’t scale. Data science needs to go through its own equivalent of the Industrial Revolution, with more focus on automation and deployment.

The base line for me is, that it’s not just a matter of applying smart technology; it’s essential to incorporate human domain expertise. In industry deep knowledge about systems and process engineering is at the heart of the analytical approach, and this knowledge is key to interpreting the output of the analyses.

Breakthrough recommendations derived from analytics in a complex process are most effectively delivered “expert to expert”.

Adapted with permission from Camo Analytics who originally published a broader editorial as part of their Industrial Analytics Thought Leadership series.