Understand and Exploit GenAI With Gartner’s New Impact Radar

Gartner recommends that when developing GenAI-enabled products and services, you:

• Create a plan to deploy and test

• Focus first on the most prevalent use cases — those that are already delivering real value to users

• Draw an investment roadmap that prioritizes opportunities

• Strive to create a competitive edge

• Hold off on long-range future GenAI technology investments

This theme is at the core of GenAI offerings and highlights elementary components, such as large language models (LLMs), as well as innovative approaches to business models, such as models as a service (MaaS). The following technologies and trends fall into this category:

• Light LLMs can support use cases where massive (or heavy) LLMs are infeasible.

• Open-source LLMs are deep-learning foundation models distinguished by the terms of use, distribution granted to developers and the developers’ access to the source code and model architecture.

• Multistage LLM chains are libraries that connect different LLMs to complete multiple tasks.

• Model hubs are repositories that host pretrained and readily available machine learning (ML) models, including generative models.

• Diffusion AI models are generative models that use probabilistic variation to add noise to data and then reverse the process to generate new samples of data.

• AI models as a service (AIMaaS) provide AI model inference and fine-tuning and are offered as a consumable service by cloud providers.

This theme highlights the user’s critical role in reducing the risks and setting guidelines for vendors to responsibly manage GenAI. The following technologies and trends fall into this category:

• User-in-the-loop AI (UITL) is a workflow that requires users to be looped into any stage of the AI system development pipeline.

• Hallucination management refers to managing incidents when LLM-generated content is nonsensical or blatantly factually incorrect.

• Retrieval-augmented generation (RAG) is an architecture pattern that combines a search function with a generative capability to ground the output from generative completions.

• GenAI extensions are tools that augment the capabilities of GenAI models by giving the models the ability to retrieve real-time information, incorporate business data, perform new types of computations and safely take action on a user’s behalf.

• Prompt engineering tools provide inputs, in the form of text or images, to GenAI models to specify and confine the set of responses the model can produce.

• Provenance detectors identify whether text, audio or video content was produced using GenAI.

This theme covers some of the critical steps and decisions in building and advancing a GenAI model. The following technologies and trends fall into this category:

• Knowledge graphs (KGs) are machine-readable data structures that represent knowledge of the physical and digital worlds, including entities and their relationships, which adhere to a graph data model.

• Multimodal GenAI models allow multiple types of data inputs and outputs, such as images, videos, audio, text and numerical data, within a single generative model. 

• AI-generated synthetic data is a class of data that is often derived and extrapolated from a set of real data but is artificially generated rather than collected from real-world events. 

• Scalable vector databases provide vector (semantic) search capability and are used in conjunction with LLMs to apply the model’s ability to respond to natural language with information that is custom or specific to an enterprise or domain.

• GenAI engineering tools enable enterprises to operationalize models faster, balancing both governance and time to market.

This theme focuses on the expectation for emerging applications, some of which will enable new use cases and others which will enhance existing experiences, over the next three years. The following technologies and trends fall into this category:

• Simulation twins leverage the best of digital twins and what-if of AI technologies.

• GenAI-native applications consist of software designed with GenAI technology and capabilities at its core.

• Workflow tools and agents are functions that agents (AI programs/algorithms) can use to interact with the world. 

• Embedded GenAI applications are existing software applications that have been enhanced by embedding GenAI capabilities to improve on existing use cases or deliver new ones.

• AI molecular modeling uses simulation techniques to rapidly test a wide range of potential treatments by modeling how different compounds will bind and interact with target molecules. 

• Multiagent generative systems (MAGs) fuse computational software agents and LLMs to simulate an environment of complex multiagent system behaviors and interactions.

• AI code generation uses LLMs to generate code based on prompt instructions a user submits.

• GenAI-enabled virtual assistants (VAs) represent a new generation of VAs that leverage LLMs to deliver superior functionality.