Getting started

AI Hub uses large language models to deliver document understanding and workflow automation, transforming how organizations extract actionable data from documents.

Two ways to build in AI Hub

AI Hub offers two editing experiences for building document understanding apps: the app editor for guided, no-code setup, and the flow editor for building fully customizable pipelines with step-level control using Python.

Both experiences use agent mode — AI Hub’s LLM-based engine — for document classification and extraction. The difference is how much of the pipeline is exposed to you.

App editor

Use the app editor to build apps that classify documents and extract structured data without writing code.

Automation projects serve as the foundation for document understanding. In a project, you define the document types (classes) your organization processes and specify which data points (fields) to extract from each class. Projects function as blueprints for your document processing requirements.

Automation apps package project definitions into deployable solutions. When you create an app from a project, AI Hub turns your blueprint into a functional application that processes documents consistently.

In addition to custom-built apps, AI Hub offers pre-built apps for common use cases such as ID verification or invoice processing. You can use pre-built apps as-is, or customize them and modify the underlying project.

You can run apps on-demand or configure them for automated processing through app deployments. Deploying an app makes it operational by connecting it to upstream document sources and downstream business systems, with support for scheduled processing, human review workflows, and data retention policies.

Flow editor

Use the flow editor to build advanced apps: end-to-end document processing pipelines with full control over every step. Advanced apps support use cases requiring complex validation, custom processing logic, or deep integration with external systems.

Flows serve as the foundation for document understanding. In a flow, you chain classification, extraction, validation, and refinement steps into a single pipeline.

Like the app editor, the flow editor uses agent mode for document classification and extraction. In addition, you can:

• Chain classification, extraction, validation, and refinement steps into a single pipeline.

• Native support for LLM-powered classification and extraction.

• Write Python in custom function steps to add logic and integrations, with access to an LLM client for AI-powered processing.

• Connect to upstream and downstream systems.

Flows are published as advanced apps, which behave like standard apps — you can run them on-demand or deploy them for automated processing.

For full documentation, see Flow editor.

Human review

Both apps and advanced apps support human review workflows, allowing reviewers to move between review tasks seamlessly regardless of which editor was used to build the app.

For details, see Human review.

Architecture

AI Hub organizes functionality through two main areas: the Hub and workspaces.

The Hub provides centralized access to document processing apps — pre-built by Instabase, customized by you, or shared in your organization — and supports reuse of assets across AI Hub.

Workspaces serve as environments for creating and running apps. Each user has a personal workspace, where you can test features or develop personal projects. Shared workspaces are used for team collaboration. Organizations typically designate specific workspaces for development and production to maintain clear separation between environments.

Most functionality is scoped to workspaces: automation projects, app deployments, human review queues, and data sources. Documents uploaded to projects remain in their respective workspaces.

When you run an app from the Hub, you select a target workspace where processing runs and results are stored. This separates reusable assets (apps) from execution environments (workspaces) and supports standard development lifecycle practices.

Subscription offerings

AI Hub offers two subscription tiers with increasing capabilities.

Enterprise lite

Designed for focused functional teams, high-growth startups, and mid-market organizations that need to rapidly deploy powerful document automation without complex orchestration requirements. Includes:

• Up to five shared workspaces for collaboration

• Organizational document processing, including validation rules and human review

• Group- and role-based access management

• Extended data source support

• External storage integration (Amazon S3, Azure Blob Storage, Google Cloud Storage)

Enterprise

Designed for large operations teams and enterprises with advanced document automation needs, including complex workflows, high-volume processing, and sophisticated human-in-the-loop requirements. Includes everything in Enterprise lite, plus:

• Unlimited shared workspace creation

• Comprehensive document processing solutions, including support for advanced apps

• Production workflow prioritization

• SLA-based reviews

• Enhanced security controls, including single sign-on authentication

• Broadest support for digitizing non-Latin languages

• Flow editor access for building advanced apps (single-tenant deployments only)

Administration

Account management lets you control account settings and, for organization admins, manage organization members and assign roles to control access and permissions.

Data connections let you connect document repositories to AI Hub as sources of file input or destinations for processed output.

Security controls protect sensitive document processing. Organizations can centrally manage secrets for integrations, and enterprise subscribers can implement single sign-on authentication.

API and SDK support

With the AI Hub API and SDK, you can integrate AI Hub into your own workflows and tooling. Both the API and SDK support running apps and deployments and retrieving results, including advanced apps created from flows.