Pointee

Through UiPath's API and webhooks, Pointee connects to UiPath Orchestrator in the cloud or on premises, back to Orchestrator 2021.10, and brings Blue Prism, Automation Anywhere and Power Automate robots under the same layer, where Pointee for RPA reassigns tasks, restarts stuck virtual machines and recovers failed runs.

Workflows reach SAP, Oracle, Microsoft Dynamics and most other ERP systems by API or file, and a team can plug in any API, AI platform, RPA bot or other tool it already uses. Once a request is approved, the agent acts on its own in those systems, posting an invoice to the ERP, provisioning access or creating a vendor record, and agents built in Pointee read invoices, forms, emails and contracts before passing the results on.

Sourcemarketplace.uipath.com/listings/pointeeread 2026-10-13

Workflow automation runs a process as a live workflow across human tasks, AI agents and the APIs and bots a company already has, starting from the process as it runs today and replacing manual steps with AI one task at a time. The canvas models conditions, parallel work, races, loops and sequences, and a single workflow can run asynchronously for weeks or months.

A patent pending engine routes work on each agent's live telemetry and source code as well as its outputs and reroutes human tasks when they stall, and for RPA estates dynamic orchestration reorders and reassigns tasks across the bot fleet so the most critical work finishes on time.

Sourcepointee.com/workflow-automationread 2026-10-13

Process discovery builds a lasting process map from AI interviews held by text, voice, a shared screen or a meeting bot in Teams calls, from documents and SOPs, from desktop task capture that records the clicks and steps nobody wrote down, and from process mining event logs, and every proposed step points back to the interview or file behind it.

Agents read each document they handle with built in document understanding, and a built in assistant answers plain language questions over Pointee's own operational data. Agents at work draw on the documents in front of them rather than on a searchable library of company knowledge.

Sourcepointee.com/discoveryread 2026-10-13

Teams choose which steps run on their own and which always wait for sign off, and can change that for each workflow, alongside safety checkpoints and rules for when a person steps in. Where an agent reads a field with high confidence it posts directly, and where confidence is lower it flags the field for review instead of guessing.

People review or approve agents' output and decisions in custom apps built for each workflow, every approval lands in an exportable audit trail, and when an agent misbehaves the team chooses whether a Guardian Agent quarantines it on its own or alerts people in Slack or Teams.

Sourcepointee.com/workflow-automationread 2026-10-13

Pointee holds SOC 2 and builds on Microsoft Azure's security mechanisms. Organizations get granular role based access control, with users managed through the company's own Active Directory, and process discovery adds OIDC single sign on with team level access policies on every process library.

Discovery data is encrypted with AES 256 at rest and TLS in transit, with keys managed in the customer's own cloud key vault, an outside specialist runs regular penetration tests, and the data processing terms commit Pointee to return and destroy personal data within 14 days of a contract ending. Pointee aligns its controls with GDPR, NIS2 and the EU AI Act.

Sourcepointee.com/pointee-for-rparead 2026-10-13

Every request, decision, reminder and timestamp, whether it comes from a person or an AI agent, goes into a single trail a team can search and export, and each case shows where it stands.

Automation teams get operational data to troubleshoot agents, see where processes slow down and measure the return on each automation, down to how each case's time splits between machine work, waiting on a person and sitting idle.

When something breaks, Pointee traces the root cause and keeps an incident record with the evidence attached, and Pointee for RPA adds real time alerts, trend forecasts and automatic reports for business owners.

Sourcepointee.comread 2026-10-13

Each case keeps its status and a full record of every action by a person or an agent, and the AI assisted inbox brings up the context a user needs for the task in front of them. Nothing carries an agent's lessons from one case to the next beyond that case record.

Sourcepointee.com/workflow-automationread 2026-10-13

Pointee for RPA deploys either on premises or in Pointee's cloud, so its data stays in the customer's own environment or on the Pointee platform, and an on premises utility collects UiPath Orchestrator data where a direct API connection is not possible. Pointee's own cloud runs on Microsoft Azure with data processed under Czech and EU law, and Workflow automation, Process discovery and Guardian Agents are sold as that hosted service. Which Azure region hosts them is not public.

Sourcepointee.com/pointee-for-rparead 2026-10-13

Guardian Agents come ready to watch every automation for drops in output quality, failures, cost spikes and security risks, and run remediation playbooks such as a version rollback followed by a health check and a notice to the team. Ready made solutions cover internal approvals for invoices, requisitions and access requests, invoices and orders posted to an ERP, and monthly reporting.

Sourcepointee.comread 2026-10-13

Work reaches Pointee as it arrives. Requests are picked up from email, a form or the ERP the moment they appear, incoming emails are classified, extracted and routed, and business users raise new automation requests through a guided intake form. Recurring work such as monthly reporting runs on a schedule once it is automated. Pointee for RPA listens to UiPath Orchestrator through webhooks, and Guardian Agents watch every automation around the clock, opening an incident the moment one of their rules trips.

Sourcepointee.com/workflow-automationread 2026-10-13

Agent work is metered in credits per case and routed on each agent's telemetry and code. A team can move a step to a cheaper model tier, as an insurer did when classification was spending most of each case's credits on a frontier model, but Pointee keeps the makers and models behind those tiers to itself.

Sourcepointee.com/workflow-automationread 2026-10-13

Teams use Pointee through its web app, and its own API calls run outward to UiPath Orchestrator and to a company's systems. A developer has no public API or SDK for driving Pointee itself.

Sourcepointee.comread 2026-10-13

Guardian Agents hold each agent to a running baseline after every release.

At a bank, a new version of a document verification agent pushed credits per case from a seven day baseline of 38 to a running mean of 340, and the Guardian Agent opened an incident after 47 cases, rolled the agent back within half an hour and notified the owner, and the fix went out behind a 5% canary with a cap of 60 credits per case.

At an insurer, Pointee read the time and credits behind each step and proposed three changes as one click actions, which cut credits per case from 41 to 24. KPIs set during discovery fill with live values once agents run and feed back into the next round of discovery.

Sourcepointee.com/case-study-banking-agent-regressionread 2026-10-13

The robots Pointee manages click through applications, while Pointee restarts their virtual machines and reassigns their work without touching a screen. During discovery the AI interviewer can watch a user share a screen and desktop task capture records clicks and steps, and both only observe.

Sourcepointee.com/pointee-for-rparead 2026-10-13