Why Technologists are Key to Building the Long Tail of Agents

Today every conversation in enterprise AI is about agents. But who is going to build these agents? There are 3 options:

(1) End User

(2) Developer

(3) Technologist - someone with a background in technology, who may be in IT or in a business unit.

Let’s examine each of these 3 personas in more detail.

The frontier labs, e.g., OpenAI and Anthropic, are pushing for a very simple model. An end user describes an outcome, maybe defines tools to use and skills, and the model figures out how to auto-magically achieve the outcome. The accuracy of achieving the outcome is derived from both the model and the end user’s ability to create the right prompt, tools and skills. Although a few users will be able to learn the necessary prompt engineering and select the tools and skills needed, can we train all end users to become model experts? This seems unlikely.

The agentic platforms, e.g., Amazon Bedrock, Azure AI Foundry, Google Enterprise Agent Platform, and Langchain, are for developers to build, run and observe agents. These platforms need APIs or MCP interfaces in order to create the agents. The vast majority of workflows in the enterprise do not have well-defined APIs, and even fewer have MCP servers. Even if APIs were available, this is an expensive approach both in terms of cost and time. These challenges will limit the number of agents that can be created by developers.

We believe that a no-code platform for technologists that operates at the GUI layer (no need for APIs, MCP, etc.) is key to building the long tail of agents. These are workflows that are either executed infrequently by a lot of users, or frequently by very few users. There are billions of these workflows in the enterprise that cannot be solved by either a developer platform or reliance on the end user.

GUIDE is intended for a technologist to rapidly automate workflows with skills they already have - capture screen shots of workflows and describe the task in language.