Factorial

Authenticated action reaches systems Factorial does not own. The platform automates provisioning and deprovisioning with AI agents, creates and updates accounts across the customer's SaaS stack the moment a contract is signed, and states that AI driven workflows take action across apps even where there is no perfect integration. Those are writes into outside systems, not reads or suggestions.

The integration estate is published and spans several ecosystem classes. Factorial runs an Integration Marketplace and an integrations page in its own navigation, an Integrations Framework in the developer documentation with payroll, finance and time tracking integration guides, and an Integrations Partner Program.

Named integrations on the IT platform cover identity (Microsoft Entra, Okta, Google Workspace), productivity (Microsoft 365), operating systems and device platforms (macOS, Windows, Linux, iOS, Android), endpoint security (SentinelOne, ThreatDown) and compliance automation (Vanta, Drata). Custom SAML and manual tracking cover apps where provisioning is unavailable, and a ready to use SaaS catalog sits behind it. The public API is the add your own route, so this is not a fixed connector list.

Sourcefactorialhr.com/factorial-it, /saas-management and apidoc.factorialhr.comread 2026-09-08

Multi step execution across several systems is documented. A customer can describe any process in plain language and the agent builds and runs it, from onboarding to an access review.

Provisioning and deprovisioning are automated with AI agents; daily operations such as enrollment, patching, password resets and offboarding are set up once, approved in a click and repeated; and every workforce change triggers the devices, access and security a person needs.

Execution spans HR, IT, identity, endpoint security and procurement in one sequence, with several participants rather than one agent looping. A separate Automations product sits in the platform navigation.

The strongest orchestration language sits on the Factorial IT product page rather than a platform orchestration page, and no named runtime, branching model or visual builder is documented; what is published is described behavior rather than an exposed engine.

Sourcefactorialhr.com/factorial-it and /saas-managementread 2026-09-08

Factorial grounds its agent in each customer's own data and policies. It states that, unlike generic assistants, One understands each company's structure, data and workflows, and the organization facing side of the agent holds and applies the policies a company defines across HR, finance and IT.

The corpus is the customer's own: employee records, contracts, time and leave, payroll, expenses, devices, access and the policies layered over them. It persists between runs rather than being gathered per job, is read continuously across recruitment, shift planning, expense management and reporting, scales across a whole workforce, and stays queryable.

Sourcefactorialhr.com/blog/factorial-one-ai-agent and /factorial-itread 2026-09-08

Approval appears as a step in the shipped flow, with no general review and approve surface over agent actions. Daily operations are set up once, "approved in a click", and then repeat. The vendor positions the agent as handling the first 80 percent so a person perfects the final 20, an advisory and handoff model rather than autonomous action.

The platform describes real autonomy without a stated gate: the AI helpdesk closes routine requests on its own, and provisioning and deprovisioning run automatically on contract signature. The trust center's knowledge base poses the exact question, whether AI generated outputs are automatically actioned, with the answer behind a click.

Sourcefactorialhr.com/factorial-it, /blog/factorial-one-ai-agent and trust.factorial.coread 2026-09-08

Both a compliance posture and an access surface are published. The trust center at trust.factorial.co, reached from factorialhr.com/security, lists four certifications with evidence attached: ENS RD311/2022 High, ISO/IEC 27001:2022, SOC 2 Type 2 and SOC 3. Alongside them sit a downloadable SOC 3 report and penetration test report, an ISO 27001 Statement of Applicability, a SOC 2 bridge letter, a security policies pack and a platform DPA, with private documents released through a request flow. The site footer carries the ENAC accredited ISO 27001 mark and the AICPA SOC mark.

On access, Audit Logging is listed as a Product Security feature, the trust center's knowledge base carries platform scoped entries on whether the Factorial platform has SSO and whether it has MFA, a System Access Control Policy is published, and Permissions System is a named product page in the platform navigation.

The trust center's Access Control section, listing Internal Single-Sign-On, Least Privilege and Logging, describes Factorial's own corporate stack governing its staff rather than a model the buyer administers. The platform SSO and MFA entries are readable as questions, with their answers behind a click.

Sourcetrust.factorial.co and factorialhr.com/permissions-systemread 2026-09-08

Platform and access activity is recorded, but no record of agent behavior is documented. The trust center lists Audit Logging as a Product Security feature, a customer facing surface the vendor supplies, and the SaaS management product states it keeps a clear record of every access change to meet GDPR, ISO 27001 and SOC 2 requirements, with audit evidence collected as work happens. That is a dated, reconstructable record of what changed and who changed it.

Nothing documents a per run trace, a decision rationale, or any field distinguishing a change made by an agent from one made by an administrator, which matters on a platform whose AI helpdesk closes routine requests on its own.

Sourcetrust.factorial.co and factorialhr.com/saas-managementread 2026-09-08

Memory with a stated scope or lifetime, per user, per team or per workflow, is not described, and no retention or deletion control over agent memory, as distinct from customer data, is published.

The organization facing side of One is described as holding and applying the policies a company defines, but holding policy is configuration the customer authored. The employee, payroll and device records the agents read are the application's own data model: deleting either would delete the business record, which is what separates a data model from a memory layer. No conversational surface carrying session state is documented either.

Sourcefactorialhr.com/blog/factorial-one-ai-agent and trust.factorial.coread 2026-09-08

The hosting footprint is published, with no choice on top of it. The core platform subprocessor list on the trust center names Amazon Web Services, Microsoft Azure, Google Cloud and SendGrid. No named region list, customer selectable residency, VPC, on premises option or customer tenant is published.

The ENS RD311/2022 High certification signals suitability for Spanish public sector buyers, but it is certification against a control standard, not a residency option a customer chooses. A demo hostname in the developer documentation contains a region like token, but a customer selectable region cannot be read out of a hostname.

Sourcetrust.factorial.coread 2026-09-08

Factorial deliberately ships a small agent set rather than a catalog. One has an organization facing side that holds and applies company policy and an employee facing side that drafts and executes on the individual's behalf, and Factorial positions that against vendors deploying hundreds of specialized agents. There is no catalog, library or browsable set of prebuilt agents for a buyer to choose among, because the product is designed not to have one.

Three nearby features are not prebuilt agents. The ready to use SaaS catalog provisions third party applications, the hardware marketplace sells equipment, and a Templates page in the Resources navigation carries HR document downloads as marketing resources rather than platform assets that do work when adopted.

Sourcefactorialhr.com/blog/factorial-one-ai-agent, /saas-management and site navigationread 2026-09-08

Factorial documents event driven triggers with a governance layer. Its developer portal has a four page webhooks section: what webhooks are, managing subscriptions, webhook policies, and a worked example clocking in a shift. The product side matches: every change in the workforce triggers the devices, access and security each person needs, accounts are created and updated across the SaaS stack the moment a contract is signed, and a separate Automations product sits in the platform navigation.

Channel coverage runs through the web application, native iOS and Android apps published on both stores, an employee portal, and an AI helpdesk that closes routine requests. A marketing consent notice stating that Factorial uses AI agents to make calls describes the company's own sales outreach, not a product channel.

Sourceapidoc.factorialhr.com webhooks section and factorialhr.com/factorial-itread 2026-09-08

Which provider's models run underneath is not named, and no model choice is offered. There is no model picker, admin model setting, provider list, routing description or bring your own key option, including on the developer portal's AI agents and MCP server pages. The trust center's AI section lists entries for AI Overview, AI Feature and AI Training Data, with their detail behind a click.

Sourcetrust.factorial.co and factorialhr.com/blog/factorial-one-ai-agentread 2026-09-08

Factorial runs a public, ungated developer portal at apidoc.factorialhr.com, linked from the site footer as "Factorial API": Guides, an API Reference and a Changelog across five published API versions from v2025-10-01 to v2026-10-01, with a versioning guide and production and demo environments. Two authentication methods are documented, API Keys and OAuth 2, the latter end to end: create an application, request an authorization code, request, refresh and revoke an access token, steps in Postman, OAuth scopes, and a partner company token guide.

The portal also has an SDKs page, an AI agents page and an MCP server page; webhooks covering concepts, subscriptions, policies and a worked clock in example; an Integrations Framework with payroll, finance and time tracking integration guides; pagination and FAQ; a technical support form; and an llms.txt indexing every page in Markdown with endpoints in OpenAPI. A separate Integrations Partner Program and labs.factorialhr.com sit alongside it. Nothing is gated.

Sourceapidoc.factorialhr.com/docs/getting-startedread 2026-09-08

Nothing is described for evaluating agent behavior: no harness, scored test cases, simulation surface, quality gate in a release path or optimization loop.

The developer portal's separate production and demo environments are a real pre deployment test surface, but what they test is the customer's integration code, not the agent: they show whether a caller works, not whether the agent behaves acceptably. The published claims that the AI resolves 60 percent of routine tickets and cuts time to certification by 90 percent are marketing outcomes, not an evaluation surface a customer can run.

Sourceapidoc.factorialhr.com/docs/production-and-demo and factorialhr.com/factorial-itread 2026-09-08

Multi OS device management is what Factorial IT runs, across macOS, Windows, Linux, iOS and Android: zero touch deployment so laptops arrive configured, remote lock and wipe, encryption, antivirus and threat detection rolled out across the fleet, enrollment and patching. That is administration of an endpoint estate through a management plane, not an agent operating a computer's interface, and no browser, desktop session, or remote or local computer control is documented.

Where Factorial's agents reach outside systems they use the REST API, webhooks and provisioning connectors, programmatic interfaces rather than a screen.

Sourcefactorialhr.com/device-management-mdm and /factorial-itread 2026-09-08