BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

AI Is Going Mainstream—Here Are Three Ways Companies Can Communicate Its Value

Forbes Technology Council

João Graça is the co-founder and CTO of Unbabel.

Have we reached the point where AI has become mainstream? As a recent study from Deloitte points out, 66% of organizations consider AI to be critical to their success. IDC also predicts global spending on AI systems will rise from $85.3 billion in 2021 to over $204 billion in 2025. With investments increasing dramatically and some of the largest companies like Meta and Google doubling down on AI research, the industry is reaching a new level of maturity.

Rather than having to hire a team of Ph.D.s, companies can now train open-source machine learning models on their own datasets with the tap of a button. Tools like Hugging Face and TensorFlow help with model discovery and customization for training. The practice of machine learning operations (MLOps) can now get these models into production easier than ever before. Once a technology like AI becomes simple enough to use, it rises to the level of mass adoption.

We've seen this story play out before with technologies like the cloud. Before Amazon, Google and Microsoft cloud platforms rose to prominence, the idea of hosting your data warehouse on a public shared server felt like an unbelievable risk. Now, some of the world's largest enterprises operate on the cloud, even those within highly regulated industries. The conversation has progressed from the idea of the cloud itself to its benefits—most prominently, its economies of scale over traditional data warehouses.

Now, it's AI companies' turn to shift the narrative from novelty to strategic value for stakeholders. Here's how.

1. Make explainable AI a priority.

Explainable AI is a set of processes or methods that make it easier for human users to understand how an AI system achieves its results. Ideally, by making the results of an algorithm explainable, it's easier to trust. Often, explainability is a key tenet of responsible AI and can help with some of AI's more serious issues such as bias, accuracy, fairness and transparency.

In addition, if AI companies continuously evaluate their deep learning models, they can more effectively help their customers and stakeholders fix potential errors that inevitably arise in training. In our company, a human-in-the-loop editing process ensures that machine errors don't fall through the cracks and that our algorithms can be retrained based on cultural and company-specific nuances in language.

Companies like Meta that are operating at hyperscale are taking a different approach to explainability by creating resources to address everyday people's potential questions about their systems. The AI System Card, for example, is an effort to explain how multiple AI models work together to provide certain results to users, such as the Instagram feed ranking. The idea is to give users additional context and transparency into why these algorithms are making decisions that impact what they see. While the System Card is imperfect, this type of content paired with model documentation is an important part of the future of AI explainability.

2. Create easy-to-use, visual interfaces.

A big part of making AI's value more mainstream is to make the user interface easier to navigate. The best tools blend into the background and are integrated into a user's existing workflow. In our company's case, our end users are often customer service agents. These people are incredibly busy and may be handling numerous customer chats and emails at the same time. By making our AI translation blend into their existing workflows, they don't have to waste time configuring a separate tool or adding steps to their already hectic workdays.

Visual user interfaces can also simplify what used to be incredibly complex processes within AI solutions. For example, correcting the output of a deep learning model and retraining it used to be a specialized process that involved specialized machine learning engineers. New systems are abstracting the complexity away from this process so that everyday users can efficiently complete these tasks, making their AI systems more valuable to their organizations.

3. Provide simple ways to measure success.

Making it intuitive to measure the ROI of an AI solution can allow stakeholders to communicate value back to the management team that allocates the budget. With emerging technologies like AI, it's easy to get lost in ephemeral promises and avoid answering questions about how a system is impacting the bottom line.

That's why it's important for AI companies to make this calculation clear to users. Depending on the industry, companies should align the results of a machine learning system with actual users' desired outcomes. Using the customer service example above, our AI system can measure the algorithm's performance against key performance indicators that matter to the customer service or support audience (e.g., translation accuracy, customer satisfaction scores, first response times, etc.). This information should be clearly and visually communicated in an easy-to-navigate dashboard so that teams can see how these success metrics are impacted by the technology over time. This type of approach can make the value of AI crystal-clear to users, their managers and even other departments that may want to leverage similar capabilities.

As we're seeing, AI is on the verge of mainstream adoption. If companies in this sector focus on making these systems more explainable (as well as simpler to use, change and measure), customers will more easily understand their value. Rather than focusing on AI for AI's sake, it's time to start speaking about how users can achieve major gains from AI. Whether that's by automating mundane tasks or providing strategic direction for the future of a program, this value shouldn't require a specialized degree to understand.


Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?


Follow me on LinkedInCheck out my website