The hype surrounding generative artificial intelligence (GenAI) and the pressure to use it sooner rather than later can lead IT leaders to apply it where it is not a good fit, increasing the risk of higher complexity and failure in their AI projects.
IT leaders should evaluate if GenAI is the right fit for their use case or whether to consider alternative AI techniques.
GenAI is currently at the Peak of Inflated Expectations on the Gartner Hype Cycle, which means that publicity around the technology and a few success stories are often outweighing the actual failures we are seeing in project development. It is not too far from the Trough of Disillusionment, where interest wanes as experiments and implementations fail to deliver. This is the phase before more instances of how the technology can benefit enterprises start to crystalize and mainstream adoption starts to take off.
IT leaders must pick and choose their spots to highlight GenAI. The question of whether to use GenAI models is a use-case-by-use-case decision. Using a prioritization tool is a good first step to determine if a use case is valuable and feasible, regardless of the AI technique being considered.
GenAI is not always the best option
GenAI is not a silver bullet, even if it may seem that way due to its popularity. In fact, it is often not the right fit for most AI use cases.
For instance, for prediction and forecasting large language models (LLMs) and GenAI are not the best fit. LLMs are not currently designed to do the kind of numerical predictive and statistical modeling required for things like demand prediction, sales forecasting, time-of-arrival estimation, weather forecasting and supply chain forecasting. Supervised machine learning might be a better fit in these instances than current generative AI models.
Planning requires exact calculations, which is not the current strength of generative models. This limits GenAI’s utility for valuable use cases like inventory optimization, field workforce scheduling, route optimization, financial portfolio optimization, pricing optimization in retail, and resource allocation.
Decision intelligence is a complex activity that requires an ability to model and choose between courses of action to achieve a target outcome. Current GenAI models are not built for decision making; their output is unreliable, they lack explainability and they are not able to model decisions in an explicit way to achieve outcomes. It is risky for organizations to rely on GenAI outputs to make critical decisions.
Finally, autonomous systems highlight how GenAI has an autonomy gap, often rendering it not useful. Current models are not currently robust enough to be autonomous, requiring close human supervision given the inaccuracies and hallucinations. This limits the usefulness of GenAI for use cases such as industrial robotics, delivery drones, smart factories, algorithmic trading and autonomous vehicles.
Many business problems will require a combination of different AI techniques, which are likely to be ignored if organizations maintain a short-sighted focus on GenAI.
Organizations can consider alternative AI techniques
For many organizations or business units, GenAI will be their first experience with AI and will start conflating GenAI with AI. However, GenAI is only a small part of the AI landscape.
A couple of the main alternative AI techniques organizations can consider are:
Nongenerative Machine Learning (ML): Also known as “predictive ML,” this is a set of techniques that make predictions with ML models that have been trained on historical data. This has been the dominant AI paradigm over the last decade, until GenAI became more prominent in 2023. It can be used for all kinds of high-value forecasting use cases, as well as problems like customer segmentation, anomaly detection, recommendation systems, customer churn prediction and predictive maintenance.
Rules/heuristics: Rule-based systems aim to capture expert knowledge in a structured manner — often in the form of rules — and use these for decision making. They are used in fraud detection, loan approval, risk assessment, medical diagnosis, quality control, knowledge discovery, anomaly detection and many other use cases.
These alternatives can be more efficient, effective,reliable and better understood than GenAI models in many cases. It is key to consider what is needed for the specific use case in terms of explainability, performance and reliability. GenAI models tend to be less reliable and explainable than other techniques.
Organizations that develop an ability to combine the right AI techniques are uniquely positioned to build AI systems that have better accuracy, transparency and performance, while also reducing costs and need for data.
Leinar Ramos is a Sr Director Analyst at Gartner focused on GenAI where he counsels AI Applications and other IT leaders on key management priorities regarding AI.
This is an opinion from an industry expert or analyst invited to contribute. They do not represent the opinions of Fierce Network.