Build AI Workflows: HATZ Platform for Industrial Automation

Starting a new workflow

The video begins by showing how to initiate a new workflow, noting its similarity to building apps but with added flexibility for moving elements.

Workflow interface and navigation

It highlights the navigable canvas for workflows, allowing users to move elements and use a mini-map for orientation as workflows grow in complexity.

Individual LLM selection per step

A key difference from apps is introduced: each step within a workflow can be configured with its own distinct Large Language Model (LLM), offering greater customization.

Tool integration for LLMs

The video explains that workflows support tool integration, allowing LLMs to interact with external data sources like the web, calendars, or CRMs, provided a compatible LLM is selected.

Connecting steps with outputs as inputs

It details how the output of one workflow step can be explicitly used as the input for a subsequent step, emphasizing that this connection must be manually configured.

Advanced prompt steps and complexity

The presenter mentions advanced prompt types like web scraper and document processing steps, noting that workflows can quickly become intricate and technical.

Run history and sharing workflows

The video concludes by explaining that workflow runs are recorded in a personal history and that workflows can be published and shared in the same manner as applications.