Artificial intelligence may be considered the new electricity, as it will fundamentally change all industrial sectors and our entire society at large. Also, AI is a general-purpose technology, exactly like electricity, as it is meant to be used to create other technologies. You do not use electricity itself (at least not intentionally); you use it to create other technologies. The same also applies to AI.

Continuing this comparison, electricity was often used in its early phases to implement the same processes and solutions that were previously being used with steam engines. People just replaced steam engines in the factories with new, more efficient electrical engines. Then, starting with Henry Ford and Frederick Taylor, people realized that you could completely transform the factory and its processes thanks to electricity.
 

The same path described above for electricity is often happening with AI today. People and organizations start using AI to automate, to make the same activities more efficient (automating call centers, back offices, quality inspections, etc.). Then, they start using AI for its real potential—for creating the future.

With this said, as we described previously here on LinkedIn and on Forbes.com, creating and maintaining an AI-driven organization is a complicated, intercorrelated and long-term activity. We described there what it means to become an AI-driven company from a strategic, cultural and organizational perspective. Here, we will continue elaborating on this fascinating topic, addressing how processes must be reviewed to become an AI-driven organization.

Many enterprises will not make the transition to become AI-driven because they still operate in a context where data, teams and processes are highly siloed. This often includes having data produced from existing processes only to "support" specific organizational goals. Even businesses that are currently "data-driven" (having successfully unified data warehouses) have a long way to go to redesign their processes before treating data as their most important asset rather than a byproduct of their processes.

It is extremely important to be realistic about the benefits as well as the limitations of AI while defining your new AI-driven processes. Most companies should consider a gradual path, maybe starting with robotic process automation and other "simple" AI solutions, before totally transforming processes with AI. In doing this, processes should always include feedback loops to verify the quality of the results achieved with AI, for leveraging this feedback back into the AI model, in the continuous "virtuous cycle of AI."

To begin the journey, the first question an organization should tackle is how to change its processes to improve the quality and quantity of available data rather than how to get better data from existing processes. This demands a total change in perspective, as you should have processes serving data and not the other way around. DataOps is a valuable emerging approach that can also be used to take a process-oriented, automated and collaborative approach to review processes and operations with an AI-driven perspective.

With AI, you may have smarter business processes and may have automated repetitive tasks, but more important than anything else, you can extract insights and intelligence from any task performed in your organization. Also, with AI, you may significantly simplify your processes and—sometimes even more important—personalize them, adapting the processes to the specifics of every single customer, employee, partner or stakeholder at large. Also, for this reason, it is key to build feedback loops from services and product usage, as personalization is based on different preferences and ways of operating and communicating.

Coming now back to the comparison I made at the beginning of this article between electricity and AI, I would like to share an example taken from one of the many AI projects I've executed in the past. My team was working for a large manufacturer, and the customer was thinking of supporting quality testing (previously done by people observing products produced) with telecameras and AI. Very similar to the early usage of electricity, the intent was to leave processes unaltered and just assist human quality inspection with AI (i.e., replacing steam engines with electrical ones).

Then, like with Ford and Taylor many decades ago, we realized that with AI, it was possible to insert additional quality inspections (just using simple, low-cost cameras) in many intermediate steps, thereby detecting defects at a much earlier stage of the process, correcting them when possible and avoiding wasting energy and prime materials (and costs) on those components that had no possibility of being repaired. Clearly, where it was impossible to insert many additional people for quality control in each step, this was instead quite simple to do with AI. We could deploy AI in a much more distributed way, totally transforming processes.

Although the example above is taken from manufacturing, you could find similar examples in practically any industry and function, including those related to CMO, CFO and HR activities.

Finally, in designing AI-driven processes, organizations must consider where they will store data that the processes produce and pay particular attention to data cleansing—the process of updating or removing data that is inaccurate. This can also be done using automated data cleansing tools leveraging AI itself. If the data are inaccurate, the output and any related business decisions will also be inaccurate.

AI can be used to rethink how we collect information, analyze data and use the resulting insights to improve our processes and our decision-making. In doing this, the best way to make predictions about the future is to create the future. This can begin by creating AI-driven processes.

In the next article of this series, we will dive deep into the technological infrastructure as well as the skills and capabilities needed to sustain your AI-driven product and services business. The best is yet to come.