Artificial Intelligence Without Business Vision is Waste

[Note: This article was originally published in Portuguese on CI&T on March 3rd, 2020.]

"Just because it's possible doesn't mean it needs to be done." This simple phrase serves as a good guiding principle for the current moment we are experiencing with Artificial Intelligence (AI). As I wrote in a recent article, the incredible possibilities of technology still generate much excitement and spark a race for investments in flashy tools and application development. However, this enthusiasm often results in few actions that truly create value for companies.

This happens because AI's ability to accurately respond to a variety of business questions sometimes functions like a siren's song, an enchantment, where the clarity of what should be done is lost, and the focus on the real objective fades. In a digital context, that objective is to prioritize meeting the needs of the consumer. In this lack of clarity, technology often ends up being used to satisfy sterile curiosities rather than generating relevant information for the development of effective solutions that bring impact to both businesses and customers.

A good measure for technology professionals to distinguish between mere curiosity and data that can generate value is to reflect on what they intend to do with the acquired information. For example, if the demand is to understand the business's consumer profile, with the right input of data and programming, an AI tool will undoubtedly be able to provide that answer, even breaking it down by demographic groups or any other desired clustering. But the key question that follows is: what are we going to do with this information? If there is no clarity in that answer, if it doesn't prompt action and a continued strategy based on it, the use of AI is a waste. Other tools, potentially simpler and more cost-effective, can address the business's curiosity, if applicable.

Technology + business = amplified results

Once we've overcome the barrier of whether or not to use technology, the next question for us, technology professionals, is that we cannot avoid understanding the business and consumer needs. Our role has shifted from merely taking orders from the business to facilitating the creation of value in a multidisciplinary and collaborative manner. This is because it's increasingly less about prescribing software to execute programmed steps and more about creating technologies capable of learning.

In other words, we must keep in mind that it's not about traditional software that provides static answers. The results will be dynamic and will change over time. Additionally, similar situations may yield different responses. All these aspects need to be directly translated into AI. And who can help us make this translation, understand what is relevant, and teach the software what it needs to learn? Business professionals.

This profile provides insight into the business problems that the solution will help solve, what questions make sense for the strategy, and what information should be considered to truly bring about impactful results. Therefore, before development begins, alignment with the business team is crucial to understand what type of learning should be programmed and through which paths.

Teaching the software to find the correct answer

I often liken the moment of teaching AI software to training a new professional who will perform the same task. The logic is the same. When training a new person, you don't provide the correct answer but teach how to evaluate which one is correct. This is especially valuable in the current context of high volatility and rapid changes.

To illustrate, let's imagine an e-commerce looking to improve its customer service. The first step is to understand with the business how real customer agents are trained to address customer concerns, what the step-by-step process should be. In a case of issues with the loyalty points program, for example, the agent needs to first confirm the customer's identity. Then, if the purchase was made, request a photo or copy of the receipt. After these steps, the missing points can be credited to the customer's account.

Knowing these and other patterns that are part of a standard customer service process, we understand how to train the software to do the same. However, a pre-defined set of questions will not always generate an expected set of answers. Thus, when the team observes real customer service, with visibility to some of the countless possible variables, they gain important insights to anticipate solution paths for new questions and emerging problems.

Promoting connection between teams

As mentioned earlier, to have this knowledge, it is necessary to instill a mindset of collaboration between technology and business professionals. This involves holding joint discovery sessions, design sprints to solve problems, and maintaining regular conversations to stay attuned to reality and changes in direction.

To establish an environment of collaboration and experience sharing for value creation, technology leadership must take certain actions:

1 - Establish a business partnership - It's crucial to identify individuals in the business area with an innovative mindset and an interest in solving critical business problems. These individuals will be valuable allies. They bring a broader and updated view of business problems, the methods they would use to solve them, and the value of these solutions to the company.

In return, the technology team should provide learning, context, and a foundation on the possibilities of AI technology. This way, business professionals will be better prepared to discuss new solution paths, identify AI application opportunities, and maximize value.

2 - Avoid alienating explanations - It is common for technology leaders, caught by the enthusiasm of explaining the numerous possibilities of Artificial Intelligence, to get lost in technical terminology that is inaccessible to those outside the field. This naturally creates a communication barrier. The listener becomes overwhelmed.

So, refrain from starting a conversation about (evaluation metric) ROC Curve, False Negative, False Positive, precision... Save these terms for a later time when there is greater understanding and maturity on the subject. This will happen during the collaborative work process. However, it's important to note that to simplify something complex, one must have a deep understanding of the subject. If it's challenging to find the right words or analogies for proper comprehension and teaching, more studying is needed. Dive deeper into the subject.

3 - Foster reflections on value versus cost of the new solution - To maintain good alignment between business and technology teams, it's important to jointly reflect on the value of a correct answer versus the cost of a wrong answer.

In other words, regardless of the teams' intention to use AI technology, it's essential to always assess whether the implementation costs are lower than the gains that can arise from the results generated by the solution. I would like to emphasize the idea that opened this text: just because it's possible doesn't mean it makes sense.

Therefore, to develop AI solutions capable of delivering value to customers and achieving impactful results for the organization, combine knowledge. It is through collective intelligence that companies gain the ability to identify and seize opportunities, using available technologies to truly innovate and delight.