This article was originally posted in Forbes.
If 2023 was a banner year for artificial intelligence (AI), 2024 was the year that saw companies put their lofty expectations to the test—often with mixed results. Naturally, companies have focused on generative AI over the past two years, buoyed by the hype generated by the launch of ChatGPT in 2022. In addition, many have invested in general-purpose large language models (LLMs) designed to serve a wide range of use cases.
As organizations refine their early uses of AI, many are finding that these generic tools often provide shallow, low-value insights and are at greater risk of hallucinations. In contrast, the companies that are finding success with the technology are those that apply it to specific, localized business challenges, considering the full spectrum of AI-enabled solutions to identify those that best align with their needs.
Creating Value Through Specialization
Companies that try a “one-size-fits-all” approach to AI may report some disappointment, but that doesn’t mean the technology isn’t creating value. Over the past year, we’ve seen countless examples of how AI is helping organizations save time, analyze data, and better serve their customers.
Digital Marketing
AI excels at extracting insights from large volumes of data. Some companies are using AI to aggregate and analyze data from sources such as social media posts, reviews, and blogs to understand consumer sentiment toward their products or services.
AI can also be a powerful tool for creativity. Stefanini recently collaborated with a major cosmetics company on a campaign that emphasized the emotional power of scent memory. Through a combination of active listening, social insights, and real-time monitoring, the company identified an opportunity to relaunch a nostalgic fragrance.
To celebrate the launch, the client asked users to share memories associated with specific scents on social media. We used AI to respond to this user-generated content in real time, creating AI-generated images that brought these stories to life in a tangible way. As a result, the company sold 17 bottles per minute during the sales cycle, exceeding the campaign goal by 42%.
Application Development
From documenting user requirements to generating test scenarios, AI can streamline the documentation process in application development. Many software companies are successfully using AI to create accurate, user-friendly documentation faster than previously possible.
Additionally, AI is helping teams design and deploy advanced solutions that improve user connectivity. For example, we worked with a global automaker to create an app that allows drivers to fully control their vehicles from their mobile devices.
The app includes 37 features such as car locator, remote locking, tracking, and collision warning. This solution has not only provided customers with more convenience and control but has also boosted the company’s sales and customer retention. In 2023, the company delivered 50,000 vehicles with the app, selling 90% of its fleet by October.
Customer Service
Many companies have found success using AI tools to answer questions about products and services, as well as to support customers or internal teams. We recently worked with a multinational company to launch a real-time translation tool, facilitating communication between teams that speak more than 20 languages. The company, which previously struggled to collaborate due to language and geographic barriers, reported an increase in productivity, employee satisfaction, and profitability.
While these three examples come from different industries, they all share a common theme: the targeted application of AI to solve specific business challenges. Rather than relying on generic AI applications to solve a broad range of tasks and domains, these companies focused on specific use cases and implemented AI to increase efficiency in a specific area.
Emerging Trends
As the use of AI matures, the tools available on the market are becoming more advanced. One model that is rapidly gaining traction is agentic AI. In this model, multiple AI “agents” perform complex tasks with minimal human intervention.
As NVIDIA points out, agentic AI systems differ from more narrow applications (such as image generation or question-answering language models) in their ability to use sophisticated reasoning to perform multi-step functions autonomously. Because these systems can analyze challenges, develop detailed strategies, and execute tasks independently, they have the potential to significantly increase productivity.
NVIDIA showcases several practical ways in which agentic AI systems are already improving outcomes across a range of industries, including providing an advanced level of support and advice in customer service interactions and synthesizing clinical information to help clinicians make more informed decisions in healthcare settings.
The growing popularity of agentic AI reinforces the shift toward more focused AI applications rather than generic tools. Rather than relying on a single general-purpose AI, agentic AI uses specific “agents” for each individual task. This, in turn, drives the need for solutions that offer agent orchestration.
Orchestrators coordinate multiple AI agents, allowing them to work together efficiently. Like a project manager, orchestrators can assign and prioritize tasks, as well as integrate data between agents, making the experience more seamless for the end user.
With this new technology comes the need for new governance frameworks that enable individuals and companies to harness the enormous potential of agentic AI while limiting potential risks. Researchers at OpenAI outlined one such framework in a paper published in December 2023. Their analysis emphasized the importance of “human-in-the-loop” protocols, which limit the scope of actions that AI agents can perform independently. They also highlighted the need for greater transparency and traceability, through “chain-of-thought” explanations, automatic monitoring systems, and the use of unique identifiers to trace actions back to the responsible human party.
A Maturing Technology
Along with the emergence of new capabilities, we are likely to see increased investment in evolving existing tools by 2025. Within the maturity cycle, textual interactions with LLMs have reached a relatively advanced stage. While there have been promising developments in image and video processing, this technology has not yet reached the same level of sophistication. In the coming year, AI tools for image and video are expected to mature, unlocking new possibilities for visual and multimedia applications.
As more organizations recognize the tangible applications of AI, they will increasingly be able to leverage this technology in sustainable and valuable ways. This will likely drive a new cycle of growth in the industry, leading to the development of more targeted and impactful AI solutions.
