4 steps to developing successful AI strategies

As with any new transformational technology, business leaders are often rushing toward any new “shiny object” that promises to streamline their business. You could say that this is what happened to Artificial Intelligence (AI) in 2020, as the results of A study in which 43% of companies around the world were accelerating their AI initiatives in response to the pandemic.

Unfortunately, many of these companies rushed to integrate without stopping and asking who, how and why. As companies seek to enjoy the benefits that AI can provide, it is important not to try to implement them at all costs without sense.

AI may seem magical, but it is not. Bad algorithms lead to bad results. While investment and experimentation are extremely important, the biggest and most common strategic mistake companies make in using AI is failing to define a clear use case and desired outcomes with metrics in the first place. quantifiable.

Many companies have rushed to integrate AI without stopping to ask who, how and why.

To solve this problem, companies turned to the principles of design thinking. An approach that begins with an analysis of who will consume AI, how they will consume it, and why they need it. This begins with critical thinking about the challenges facing the business, framing those challenges in ways that AI can solve, and then identifying and refining the use cases that are critical to business goals.

With a data- and human-centric strategy, we can design an AI that successfully connects all strategic data and is appropriate to the companies’ business objectives. This strategy must respond to these four steps:

1. Set a purpose. Many companies don’t really have a clear idea of ​​what they hope to gain from AI beyond a vague notion of efficiency. That’s why it’s important to refine your goals by spending time studying how AI can be applied within your current business strategy. Is the safety of the workers sought? Keep customers happy? You have to start with an established purpose based on business objectives.

2. ID. Once the main objective for implementing AI has been determined, the use cases and the types of AI solutions that users need and that will eventually be integrated into the infrastructure must be defined. Artificial Intelligence is advancing rapidly in numerous fields, from computer vision, which determines what is in an image, to natural language processing for chatbots and virtual assistants.

3. Evaluation. The evaluation stage involves figuring out what data is needed to make the identified use cases efficient. Different types of teams focus on different priorities and different sets of numbers, which means that most industry data is siled to some degree. Implementing successful use cases through AI requires ensuring that AI is being fed accurate data that is drawn from across the organization.

4. Plan. The last step of the design thinking approach focuses on establishing concrete actions using statements of intent as a guide for technical implementation. The goal is to help clients operationalize AI across the business by connecting each solution to the defined AI strategy.

Fundamentally, an implementation strategy must earn user trust: how will customers react to your organization using data in this way? How can consumers and the public know that the AI ​​implementation is measurable and reliable?

Designing an AI strategy is also based on who is involved in the process. It is important for companies to include diverse voices and the right stakeholders at every stage of the process.

When we go to our clients’ offices, strategy-setting sessions are attended by senior executives who set the intent, define the types of information, develop business hypotheses, identify use cases, and infuse the business ethos into the strategy. The technical sessions invite data scientists, designers, and developers to come together to translate the proposals established in the strategy session into a more detailed plan, defining the use cases, evaluating the data, and planning the execution. Throughout each exercise, visual narratives, images, and graphics are used to help ensure that, even though they come from different fields, everyone involved has the opportunity to speak the same language.

What are the most common opinions? Habitually, when working with clients, their “aha!” arrives during the evaluation phase. Too often, companies believe they already have all the data they need to run any AI model, and that’s rarely the case.

For example, a client in the financial services industry wanted to develop an AI solution that would help speed economic recovery for small businesses affected by the pandemic. But in evaluating the data needed to create value for selected users, the team realized for the first time that their data was disorganized, siled, or unusable. Before starting to implement a trustworthy model, the data collection, infrastructure, and platform issues that hinder the development of trustworthy AI need to be addressed.

There is no doubt that AI is transforming business today. From healthcare organizations using natural language processing to help manage COVID-19-related inquiries to financial services companies using AI to parse tedious compliance documents, early adopters of AI continue to develop new use cases by the dozen. But what all of these successful implementations have in common is a clear intent and plans that connect the benefits of AI to a company’s top priorities.

By Francisco Montero, Data & AI Tech Sales Manager at IBM