Automat AI
Also known as: Automat, Automat Workforce, Automat Core
AI agents that operate computers visually, sold as Automat Workforce (Ace, an AI coworker with its own Mac, inbox and phone) and Automat Core (managed agentic automation replacing legacy RPA on SAP, Citrix and portals without APIs).
Automat builds AI agents that operate computers the way people do. Its agents see screens with vision language models, use a mouse and keyboard, and adapt when interfaces change, so they run web and desktop software, SAP GUI, Citrix and legacy portals that never shipped an API. The company sells two product lines on the same execution engine.
Automat Core is a fully managed agentic automation platform built to replace legacy RPA. It combines UI based agents, AI document extraction and API based automations in one workflow, with a visual and code workflow builder, human in the loop approval steps, scheduled and durable runs, session replay and audit ready history, and a cloud, private or desktop runtime that can execute inside the customer's own infrastructure. Forward deployed engineers build and maintain the automations from a screen recording or SOP, and a REST API with personal access tokens creates, runs, schedules and inspects workflows.
Automat Workforce is an AI coworker named Ace with its own Mac, inbox, phone number and logins, reached over email, Slack, Teams, text, iMessage, WhatsApp, document comments and phone calls, priced per agent with a 14 day trial. Customers set which actions need approval, and every run is logged with the steps taken and the screens seen. Automat lists SOC 2 Type II, ISO 27001, HIPAA and GDPR, with role based access, SSO and SCIM.
Automat was founded by former Google engineers Lucas Ochoa and Gautam Bose and is based in San Francisco. It has raised $19M, including a $15.5M Series A led by Felicis with Khosla Ventures, Initialized and Y Combinator. Customers are concentrated in banking, lending, insurance and healthcare.
Vendor details
Canonical URL
https://runautomat.com
Category
Browser / computer-use agent
Subcategory
Agentic process automation / computer use
Funding status
$19M raised per the vendor, including a $15.5M Series A led by Felicis with Khosla Ventures, Initialized and Y Combinator
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Automat operates any application through the screen, including SAP GUI, Citrix, Salesforce, NetSuite, Workday, ServiceNow, EHRs, government and carrier portals and desktop apps, with no APIs required. API based automations call and route data between systems in the same workflow with managed credentials. A REST API with an OpenAPI 3.1 specification and personal access tokens, along with an Enterprise MCP server, exposes workflows, runs, schedules and sessions. The Workforce agent works inside the customer's own Slack, Teams, Google Workspace, Microsoft 365 and CRM with its own accounts.
In practice
An agent logs into a legacy carrier portal with no API, extracts policy data, and delivers it into a downstream system, adapting automatically when the portal UI changes
A mortgage lender automates initial disclosures and appraisal data extraction across an LOS without API access or a months long RPA build
A banking customer runs an automation twenty four hundred times a day with CAPTCHA solving, 2FA, and parallel browsers against a production API
Sources & related URLs
Agentic Index coverage score
11.5 / 14 capabilities · 82%
| Integrations & Tool Calling | Full |
|---|---|
|
API based automations call outside systems and route data between them inside a workflow, signing in with write only managed credentials. The Workforce agent, Ace, holds its own accounts and logins in the customer's tools. Sourcerunautomat.com/capabilitiesread 2026-09-27 |
|
| Workflow Orchestration | Full |
|
Automat Core runs versioned workflow definitions from a visual and code workflow builder, combining UI agents, document extraction and API calls in one process, with durable runs, a per node timeline, validation and cloning into development versions. Sourcedocs.runautomat.com/llms.txtread 2026-09-27 |
|
| Knowledge Grounding & RAG | Partial |
|
Agents work from documents supplied for each run and from a project library of curated files and data resources, and the Workforce agent reads a customer handbook. Automat keeps no retrieval index, embeddings layer or graph over customer knowledge, and document extraction parses each document on its own. Models trained on customer history learn from it in training and do not retrieve from it at run time. Sourcedocs.runautomat.com/llms.txtread 2026-09-27 |
|
| Human Oversight & Guardrails | Full |
|
Workflows carry human in the loop steps that raise approval tasks with actions, fields and an expiry, completed in the product or through the API before the run continues; Workforce customers set which actions need approval before they happen and can pause or redirect the agent. Sourcedocs.runautomat.com/api-reference/programmatic-access/programmatic-access-api/hitl/list-tasksread 2026-09-27 |
|
| Security, Identity & Governance | Full |
|
Access runs on role based controls with team level isolation, SSO and SCIM through the customer's identity provider, per tool scoped access and an organization scoped isolated environment. Automat lists SOC 2 Type II, with the report under NDA, ISO 27001, HIPAA and GDPR, and runs a trust center at trust.runautomat.com. Sourcerunautomat.com/capabilitiesread 2026-09-27 |
|
| Observability & Auditability | Full |
|
Every run keeps session replay, retained logs, a per node timeline with inputs and outputs, a recording and screenshots of each action, readable in the product and through the sessions API, with an audit ready history of what the agent did. Sourcedocs.runautomat.com/api-reference/programmatic-access/programmatic-access-api/sessions/replayread 2026-09-27 |
|
| Memory & State Persistence | Partial |
|
The Workforce agent keeps a handbook the customer adds notes to and works on a dedicated Mac whose state persists, and Automat sells tasks, memory and recall as an add on. Automat does not say how that memory is scoped or how long it lasts, and durable Core runs keep state only within a single run. Sourcerunautomat.com/pricingread 2026-09-27 |
|
| Deployment & Data Residency | Full |
|
A cloud, private or desktop runtime is offered, and agents can execute inside the customer's own infrastructure so sensitive data stays on premises, with eligible worker groups selectable per workflow; no region list is published. Sourcerunautomat.com/capabilitiesread 2026-09-27 |
|
| Prebuilt Agents, Templates & Packs | Partial |
|
Use cases on the industry pages include disclosures, claims, KYC and refunds, but forward deployed engineers build each automation for the customer, and there is no catalog of prebuilt agents or templates for customers to adopt. The packages API lists code libraries for workflow steps, which the customer assembles. Sourcerunautomat.com/capabilitiesread 2026-09-27 |
|
| Triggers & Channel Coverage | Full |
|
Workflows wake on their own through a schedules API that takes a recurrence rule and reports the next fire time, and they also start by webhook or API call. The Workforce agent is reached by email, Slack, Teams, text, iMessage, WhatsApp, document comments and phone calls. Sourcedocs.runautomat.com/api-reference/programmatic-access/programmatic-access-api/schedules/createread 2026-09-27 |
|
| Model Flexibility & Routing | Partial |
|
Automat assigns two providers by function, Anthropic for computer use and Google Gemini for video. For extraction the customer picks the Mini, Pro or Max tier, each backed by a named Gemini model, so the customer chooses a tier but not a provider. Sourcedocs.runautomat.com/guides/concepts/extractorsread 2026-09-27 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
A REST Programmatic Access API with an OpenAPI 3.1 specification and bearer personal access tokens creates, runs, schedules and inspects workflows, sessions, extractors, credentials and files, and an MCP server is listed for Enterprise. Sourcedocs.runautomat.com/llms.txtread 2026-09-27 |
|
| Testing, Debugging & Optimization | Partial |
|
A validate endpoint checks the active workflow definition before deployment without dispatching a run, and run outcome counts per version, session replay and logs support debugging. Testing against real data before production is done by Automat as part of its managed service, not by the customer in the product. Sourcedocs.runautomat.com/api-reference/programmatic-access/programmatic-access-api/workflows/validateread 2026-09-27 |
|
| Browser & Computer Use | Full |
|
Vision based agents see screens and drive mouse and keyboard across web and desktop applications including SAP GUI, Citrix and legacy portals, with unattended browser and desktop automation, stealth browsers and hosted browser sessions, and each Workforce agent operates its own dedicated Mac. Sourcerunautomat.com/capabilitiesread 2026-09-27 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
Automat made Ace, the AI teammate in its Automat Workforce product, generally available. Ace gets its own computer, email address, phone number and accounts, takes requests over Slack, Teams, email, SMS, WhatsApp and phone, and does the work inside browser and desktop tools rather than through integrations. It logs every action, handles small reversible tasks alone and asks a designated manager before anything consequential.
Bears on: Browser/computer use
View sourcePricing
Workforce is $2,000 a month per AI coworker, or $1,800 a month billed annually, with a 14 day free trial. Enterprise is custom.
Workforce bills per AI coworker, not per seat. Enterprise is quoted and scoped to outcomes.
What is public
The Workforce price per AI coworker, monthly and annual, and the 14 day trial are public. Enterprise is custom.
Billing mechanics
Workforce bills a flat fee per AI coworker (dedicated Mac, email and company account, browser and computer use) with no per seat pricing. Enterprise is a managed service with forward deployed engineers and custom SLAs, quoted on scope, and it is usage based with no per bot fees.
Cost watchouts
Workforce add ons (hosted coworker, real user accounts, phone, SMS and iMessage, tasks, memory and recall) are listed without prices. Enterprise is usage based, so high volume runs drive cost.
Variable cost rationale
Workforce meters AI coworkers, not usage. Each coworker is a flat $2,000 a month, or $1,800 billed annually, so the bill grows when another coworker is added. Add ons such as a hosted coworker, real user accounts, phone, SMS and iMessage, tasks, and memory and recall carry prices that are not published. Enterprise is usage based and scoped to outcomes, with no per bot fees, so heavy automation volume moves that bill. No Enterprise rate, cap or minimum is published.
Additional watchouts
Price is per AI coworker, so a second agent doubles the bill. Workforce add on prices and the Enterprise usage meter are not published.
Sales call required
Mixed (some tiers require a call)
Free / trial
14 day free trial of Workforce from the Studio sign in
Lowest paid plan
Workforce at $2,000 a month per AI coworker, or $1,800 a month billed annually
Key ambiguities
Enterprise rates, minimums and how managed engineering time is billed are not published. Several Workforce features (hosted coworker, phone and iMessage, tasks and memory) are listed as add ons without prices.
Support SLA / resale
Enterprise is fully managed, with forward deployed engineers and custom SLAs.
Missing data
Enterprise rates and minimums, and add on prices on Workforce.
Related vendors
- AgentQL — A self healing query language for the web, with natural language…
- AGI, Inc. — AGI, Inc
- Airtop — Cloud browsers for AI agents that compile plain English workflows…
- Asteroid — Healthcare portal integration platform whose supervised browser and…
- Autotab — General AI agent that learns a workflow from a demonstration and…
- Axiom.ai — No code browser automation tool and Chrome extension for building…
Alternatives to Automat AI
The closest documented capability profiles to Automat AI among browser and computer-use agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Skyvern11.5 / 14Fuller documented coverage on Model Flexibility & RoutingAutomat AI vs Skyvern →
- Asteroid11.5 / 14Fuller documented coverage on Prebuilt Agents, Templates & Packs and Testing, Debugging & Optimization
- Browser Use11.5 / 14Fuller documented coverage on Model Flexibility & Routing and Testing, Debugging & Optimization
- Grok Bot10.5 / 14Fuller documented coverage on Prebuilt Agents, Templates & Packs
- Narada10.5 / 14Fuller documented coverage on Knowledge Grounding & RAG
- OpenClaw13.5 / 14Fuller documented coverage on Knowledge Grounding & RAG and Memory & State Persistence
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded