Proto

ERP•AI apps

ERP•AI apps and workflows

Turn a business process into a working app, with records and views people can use beyond the original conversation.

For Process owners and business application teams. Based on Proto 0.2.122.

Start with the business process

A supplier review needs suppliers, proposals, decision owners, and a record of what was accepted. An invoice exception process needs invoices, matching evidence, unresolved differences, and a next action. Describe those relationships to Proto in the language your team uses.

Proto can build and modify ERP•AI apps through native operations. A new app can include linked tables, calculated fields, contextual sample records, an overview dashboard, a workflow board, and table workspaces. Ask for the additional pages, reports, or automations that your process needs, then inspect what was actually saved.

Move from an example to operating data

Sample records help you understand the app’s structure and test how pages behave. They are not evidence about your business. Keep them distinguishable from imported or entered operational data, and confirm the field definitions before loading a real export.

Create an invoice exception tracker with suppliers, invoices, purchase-order references, exception reasons, assigned reviewers, and resolution notes. Include a review queue and a summary by supplier. Use clearly identified sample data so we can review the structure before importing our records.

During review, follow one representative case through the app. Check how records link, whether totals mean what the team expects, and whether a reviewer can find the evidence needed to make a decision. A useful app needs coherent relationships as well as a polished dashboard.

Add automation to a defined event

ERP•AI workflows can react to records or run on a schedule. Proto can create and edit these workflows, configure their steps, and test the saved graph. An automation might flag an overdue review, update a status, or send a message through an authorized provider connection.

Say explicitly when you want an inactive draft. The normal workflow-building path can activate a trigger-driven workflow after its current version passes its test. Missing credentials or a failed test keep it inactive. A preview can show intended writes or sends without performing them, so distinguish a passing preview from evidence of a real event being delivered.

Before activation, review the event, filters, destination, and expected side effects. After activation, check a real run and its result. A manual-only workflow and an event-triggered workflow have different operating expectations.

Work within the account’s authority

Proto’s ERP•AI actions use the signed-in user’s permissions. Access to records, structure, pages, workflows, and other resources depends on that account. The agent does not create authority that the account lacks.

ERP•AI deletion and sensitive changes to roles, user access, and invitations are disabled by default in Proto’s Security settings. Permission mode is a separate choice. Turn off Uninterrupted mode and leave Plan mode when the work should pause for action review, and use an appropriately scoped account for evaluation.

Decide what belongs in the app

A desktop task is useful for investigating a situation and preparing a result. An ERP•AI app is useful when several people need a shared record and a repeatable process. A saved workflow is useful when a defined event should initiate the same sequence again.

Keep that distinction clear in the handoff: what exists in the app, which records are examples, which workflows are active, and what still needs a person or provider connection. Proto remains in alpha, so validate the complete process with representative cases before relying on it for routine operations.

Need help applying this to your work?Get support