Automations (AI Business Operations Manager)

Live

Build, schedule and audit business workflows that combine your Inaya data, an AI Operations Manager, rules and notifications, with human approval for anything that changes a record and evidence for every run.

What it is

Automations lets an authorized person build a workflow on a visual canvas: a trigger (a schedule, a button, an event, a webhook), steps that read your Inaya data, an AI Operations Manager that analyses it, rules that decide what happens, and notifications or requests for approval. It connects the parts of Inaya you already have (CRM, invoices, tasks, procurement, inventory, Business Brief, Trust Health, Digital Twin, Evidence Graph, notifications and the approval system). It does not replace any of them.

The idea is: data, then AI, then a decision, then governance, then action, then evidence.

What you can do

Area	What it gives you
Visual editor	Add, connect, move, disable and duplicate steps; zoom and pan; live validation; unsaved-changes marker; save a draft, test, dry-run, execute, publish, and roll back to an earlier version
Triggers	Manual, schedule (daily, weekly, monthly, interval, in your timezone), event, webhook (signed), API key, Evidence Graph event, Digital Twin completion, data change
Data	CRM and sales, overdue invoices, tasks, procurement, inventory, projects, documents, security events, backup status, Trust Health, Business Brief, Evidence Graph events, Digital Twin results, and your helpdesk through a controlled HTTP connector
Transform	Merge, join, filter, map, select, rename, sort, aggregate, group, deduplicate, derive. There is no code step.
KPI snapshot	A permission-scoped snapshot with its period, sources, organization and calculation notes, ready for the AI and for evidence
AI Operations Manager	Gemini (server-side only) returns a structured assessment: urgent or not, classification, confidence, findings, recommendations. It can only call tools you enable
Rules	Conditions with AND, OR, NOT, comparisons, thresholds and the AI's structured result
Notifications	Inaya notifications and email (through Inaya), plus Slack and Gmail. Repeats are deduplicated, so a retry never sends twice
Approvals	A step can only propose a business change. A person approves it, the normal delay applies, and the existing system performs it
Test and dry run	Test uses synthetic data; dry run reads real data. Both simulate every write and label every output
Evaluations	Save test cases with expected branches, notifications, tools and latency, and run them safely
Evidence	Every run is recorded in your tamper-evident audit trail and the Evidence Graph, with an exportable Evidence Passport and a plain "why did it do this" explanation
Health	Automation health, metrics, retention controls and failure notifications

How safety works

Permissions follow the person running it, not the author. A workflow reads only what the running person can see. Sharing a workflow does not lend anyone your access, and a scheduled run stops if its owner has lost access.
The AI cannot go beyond its tools. Tool names, arguments and permissions are checked by the server, not by the model. A tool that changes data can only ask for human approval.
Untrusted text stays data. Instructions hidden in a ticket or document are removed before the AI sees them and are recorded as an AI-security event.
Secrets never sit in a workflow. Credentials are stored encrypted and referenced by id; every use is audited and exports never contain them.
Outbound calls are locked down. HTTPS only, an allowlist of hosts, no private or metadata addresses, no redirects, size and time limits.
Runs are safe to repeat. A crash resumes where it stopped; nothing is sent or requested twice.

What is real and what is not yet verified

Gemini, the scheduler, the queue, permissions, evidence, approvals, the Digital Twin and Inaya notifications are exercised by automated tests against the real database.
Slack and Gmail sending have been verified live (2026-09-26): a real production run delivered through a Slack incoming webhook and the Gmail API. Slack uses an incoming-webhook URL; Gmail uses an OAuth refresh token you create once. A repeat run on the same day is deduplicated and does not send again.
Support tickets. Inaya now has its own Customer Support module (see Customer Portal & Customer Service), and Automations can read it directly through the native "Inaya support tickets" step, limited to what the person running the workflow may see. The older "Get Support Tickets" step, which reads an outside helpdesk through the HTTP connector, has only been tested against a local test server, not a real helpdesk product.
Schedules run from Inaya's five-minute background job, so a schedule fires within about five minutes of its time.

Set up Slack and Gmail

An owner or admin opens Automations, then Credentials.
For Slack, create an app in your workspace (a ready manifest is provided), turn on Incoming Webhooks, add the webhook to a channel, and store the URL as a Slack incoming webhook credential.
For Gmail, create a Google Cloud OAuth client of type Desktop app, run the helper script to obtain a refresh token for the sending mailbox, and store the client ID, client secret and refresh token as a Gmail credential.
Use the Notification Delivery Test template to send one message to the inbox, Slack and Gmail before relying on an alert.