New trend! Gartner points out that in 2 years, "small-scale specialized AI" will be used 3 times more than LLMs!
The need for specialized solutions, reliability, and cost-effectiveness are key factors driving the shift to smaller, more task-specific AI models.
Gartner predicts that over the next two years (2027), organizations will deploy small, task-specific AI models at a rate at least three times higher than traditional large language models, or LLMs.
Conventional large language models, despite their strong language capabilities, have a reduced response accuracy when it comes to tasks requiring specific business contexts.
The versatility of business process tasks and the need for greater accuracy are driving a shift to using specialized models tailored to specific functions or domain data. These smaller, task-specific AI models provide faster responses and require less processing power, reducing operational and maintenance costs, said Sumit Agarwal, vice president, Gartner analyst.
Organizations can customize LLMs for specific tasks using Retrieval-Augmented Generation (RAG) or Fine-Tuning techniques to build specialized models. In this process, an organization's data is a key differentiator. Overall data preparation, validation, versioning, and management are required to ensure that relevant data is appropriately structured to meet the needs of fine-tuning, which is the process of customizing an AI model with specialized data.
As organizations become more aware of the value of personal data and insights gained from specialized processes, they are more likely to start monetizing their models and offering access to these resources to a wider audience, including customers and even competitors, moving from a defensive approach to a more collaborative, open use of data and knowledge, Agarwal added.
By commercializing the model, organizations can create new revenue streams while fostering a more connected ecosystem.
Deployment of small, task-specific AI models
Organizations looking to deploy small, task-specific AI models should consider the following recommendations:
- Pilot Contextualized Models : Deploy small, context-specific AI models in areas where business context is important or where LLMs cannot meet quality or speed expectations.
- Adopt Composite Approaches : Identify use cases where a single model is insufficient and move to a hybrid approach involving multiple models and workflow steps.
- เStrengthen Data and Skills: Focus on data preparation to collect, curate, and organize the data needed for fine-tuning language models, while investing in developing the skills of people across technical and functional areas such as AI and data architects, data scientists, AI and data engineers, risk and compliance teams, procurement teams, and business-side experts to effectively drive this initiative.
Gartner / PC & Associates Consulting (PR)