Artificial intelligence (AI) is in transition in technological solutions and how it is used. Companies are increasingly taking pilots out of test labs and deploying them at scale, with some reaping significant benefits.
Regardless of any uncertainty surrounding this science, ignoring its potential risks, companies doing business ‘old fashioned’ are going under.
But for other organizations, getting value from AI can be elusive. Your models may need to be tuned, and your training data sets may need to be more significant.
In addition, there are concerns about bias, ethics, and transparency. Launching an initiative into production before it’s ready or expanding a strategy beyond its initial phase before properly examining the results can cost a lot of money or send an organization down a business-damaging path.
How can a project of this magnitude transform or, on the contrary, sabotage a company? They have to get creative without clear return on investment (ROI) numbers. And IT leaders are measuring the value of AI.
Mature versus innovative technologies
Measuring the business value of any initiative or technology is only sometimes a linear calculation. AI is no exception, especially when considering degrees of maturity and potential.
Proven and predictive variables such as data mining, cost and training savings, investment, and the ability to facilitate new uses influence decisions regarding acceptable ROI. Still, it is essential to trust the technology, regardless of how new or established it is.
Some use cases are at a high level of maturity. We also have the tedious processes of all companies. So we automated a lot of things, like ticket processing, data search and extraction, and review of contracts and subcontracts.
For technologies with a medium level of maturity, JPL looks at whether they can enable new use cases and at what cost.
Here, AI will help new use cases that are not currently possible. In the case of NASA, it could analyze images from space and send a million text titles back to Earth to describe, for example, a dry lake in a particular direction.
Finally, the measure of success for the most cutting-edge experimental AI technologies is whether they allow new science to be done and new papers to be written and published.
Companies like Google and Microsoft have easy access to gigantic volumes of training data, but at JPL, the data sets take a lot of work to come by and require PhD-level experts to analyze and label them.
AI measurement and its spheres of influence
When there is no direct way to measure the business impact of an AI project, companies will mine data for related key performance indicators (KPIs).
These variables are generally related to business objectives and may include customer satisfaction or employee retention rates, among other variables.
An example is Atlantic Health System, where patients are at the centre of every decision, says Sunil Dadlani, its senior vice president and CIO. In many ways, the return on investment in AI is measured by looking at improvements in patient care.
These patient-centric metrics include the reduced length of stay in the facility, faster treatment time, or faster insurance eligibility checks.
Measure success incrementally
Automation leading to cost reduction is the most straightforward way to show the economic benefits of AI, says Sanjay Srivastava, head of digital strategy at Genpact. But it can also facilitate new revenue streams or even completely transform a business model.
For example, an aircraft engine manufacturer saw that it could improve failure prediction and logistics so it could begin offering engines as a service. “It is a new business model; it changes how a company operates because AI technology allows it.”
Alignment with the strategic vision
Then there is the reality that some projects may hurt the bottom line in the short term but still be meaningful and transformative in a long time. For example, a company that implements a customer service chatbot can eliminate everyday tasks.
But these can be harmful because some people are good at upselling and want to interact with people.
He says it goes back to what kind of company you want to be. At some point, you have to ask yourself if this is the kind of company where if a delivery goes wrong, for example, customers can call to ask where it is, and then you try to sell them your product.