5 Levels of GenAI Engagement Maturity
Level 1: Q&A #
This phase uses chat primarily as a way to ask how to do things. It might include proofreading or copy pasting in or out, but the primary interface is a dedicated chat window, and the output is text that a human reads and does something with.
Level 2: Tooling Augmentation #
This phase uses gen AI augmentation inside a tool, sometimes as a peer worker, sometimes as Q&A. The GenAI can recommend changes in real time, as well as be asked to make larger changes on behalf of the user. There are three components to this shift: the embeddedness in the tool, the passive context gathering, and the capability for proactive participation.
Level 3: Supervised Agents #
In this phase, the agent is the primary worker. It is something of the inverse of level 2--the agent is given a problem to solve, and it proposes a solution, including which tools to use, which the human either requests changes to or accepts.
Level 4: Unsupervised Agents #
In this phase, the input phase of an interaction is offloaded from a user interaction to a common pool; an agent or group of agents will work against a task or assigned tasks to completion on the behalf of the user. The user will then review their work--usually at the end user level.
Level 5: Agents as Entities #
In this level, agents have their own identities, accounts, credentials, and resources. They run autonomously from other users (sometimes as a bot or app), and follow the same workflows as other system users from end to end.