Computers have brains. Obviously, we’re not talking about anything even remotely organic or anatomical, although nanotechnology may ultimately erode that dividing line over the years to come.
When we talk about computer brains we’re clearly referring to the core microprocessor that drives any given computer, be it a desktop machine, a mobile device or some smaller embedded unit. But in the age of machine learning (ML) and artificial intelligence (AI), we’re now also talking about computer brains that are accelerated and augmented by a decision engine.
Born out of a neurally combined intermixture of algorithmic logic and data at its lifeblood, a decision engine is only as insightful as the algorithm that drives it and only as smart as the data source it is able to ingest.
As organisations in every vertical now work to apply the intelligence power that comes from AI and ML, it seems clear that a decision engine requires a decision point before application… and this is a truth that plays out at various levels.
Starting Decision Points
In theory, ML can be applied anywhere, but the most practical and efficient way to get started is by focusing on decision points. This is the opinion of Rick Rider, in his role as VP for applied innovation at Infor.
What the Infor team says here is that if an organisation builds great ML solutions, it will (typically) allow that company to progress from degrees of augmented intelligence to full smart automation, all at its chosen pace.
“When selecting the right ML systems, it’s important to understand the details of the business problem and how that problem can be related to ML solutions. Choosing an algorithm first and then trying to apply it to a business case can be dangerous leading to biased results. Using ML in your enterprise should be a long-term, repetitive engagement and therefore you will need a flexible platform to grow with your organisation,” explained Rider.
Without this approach, he warns that any business will be stuck with a massive Total Cost of Ownership (TCO) burden per project. This is probably why (and this is another decision point) it’s a good idea to select a vendor whose technology matches the creative culture of the organisation purchasing the software solution.
Decide on Data Quality
As noted above, a decision engine is only as smart as its data source. It is for this reason that Rider and team say that data quality is crucial in applying algorithms to glean insights without explicit programming. Insisting on data quality is the decision point that either will or won’t lead a business down a route to maximising the potential of ML itself.
But training an ML system to ensure data quality can be challenging.
“One technique is to normalise the data during pre-processing to protect against misinformative data or outliers, but a bolder and more effective solution is to build models to correct or flag data at the time of the actual event or transaction. For example, if you have a customer that typically purchases certain types of products and mistakenly submits an order for something entirely different, ML within the system workflow can flag the exception in real-time and ultimately prevent both you and the customer from a negative situation,” explained Rider.
Having tools than can wrangle and clean data with no-code and low-code options can make sure that regardless of the state of your data it can be prepared for consumption by an ML model.
Human Decision Precedes Machine Decision
When it comes to deployment, getting ML systems to production and into end users’ hands is the biggest challenge. The Infor team draw from many years work in this sector and say that finding tools to perform ML is the comparatively easy part; but, even with this task covered, someone still has to develop the models, map business problems to ML solutions, integrate with the data sources and optimise and maintain the system over time.
“Having the right ecosystem to securely access, host, and administer those insights in a meaningful manner to users is where companies struggle. For this reason, organisations should focus their selection on vendors that can remove the barriers to entry and prioritise the consumption of insights as opposed to the science itself. The science behind delivering effective ML then becomes a commodity, much more quickly,” said Rider.
We can state with some certainty today that ML undoubtedly forms part of the future digital landscape, but the pace at which organisations embrace (and indeed, derive ROI from it) will vary significantly.
Infor’s Rider concludes that the companies that embrace the changes and realise that ML is just a creative medium will quickly see how they can differentiate themselves in terms of talent and other ways beyond just their products and services.
It seems clear that organisations will need a solution that matches their desired machine brain culture too and not one that simply checks the box or has the slickest marketing.
AI and ML are clever, but business needs to get smart over exactly how it builds smartness.