Nominal

The platform is ERP agnostic by design and plugs into any ERP, ledger or data source, with named solution pages for Workday, Microsoft Dynamics 365, Oracle, Sage Intacct and SAP alongside NetSuite and QuickBooks, plus national and regional banks and procurement systems, and a bi directional data ledger across entities, systems and currencies. Agents act rather than read: they post corrections as they occur and write adjustments back to the general ledger, and resolution agents solve discrepancies at source.

SourceNominal, nominal.so/our-platform and nominal.soread 2026-09-07

A network of autonomous agents runs accounting operations across systems, entities and time, executing reconciliation, transaction matching, variance analysis, consolidation and close continuously rather than waiting for instructions or for period end. Nominal's own account of the model states the agents do not just execute tasks but orchestrate them, understanding dependencies, monitoring progress and adapting as new data or exceptions emerge, and describes Agentic Performance Management as a coordinated system of specialized agents with a transaction level data foundation beneath them.

SourceNominal, nominal.so/our-platform and nominal.so/blog/agentic-workflowread 2026-09-07

The Shadow Ledger is a unified, bi directional data ledger across entities, systems and currencies, and Nominal names a transaction level data foundation as one of the three things autonomy requires, with agents reasoning against it and returning sourced answers to close questions. The limit: what persists is the transaction record itself, the operational data layer the product maintains, rather than a maintained retrieval structure over a knowledge corpus; no index, graph or embeddings layer is named, and nothing describes ingesting the customer's policies or documents into a queryable store.

SourceNominal, nominal.so and nominal.so/our-platformread 2026-09-07

Nominal documents its own review and approve surface with its mechanism: when an agent proposes an entry it does not post automatically but appears in a queue with full documentation showing the matching logic, source transactions and recommended treatment, and a controller reviews and approves before it reaches the general ledger.

Finance teams configure approval thresholds, define materiality levels and set the rules for when human review is required, with the system handling everything below those thresholds automatically. The platform page states the principle as control stays human, agents do the work, people guide, review and decide.

SourceNominal, nominal.so/blog/ai-agents-in-finance-accounting and nominal.so/our-platformread 2026-09-07

Nominal documents compliance but not an access model. Its security page states SOC 1 compliance, which it notes is the standard required for public accounting, and SOC 2 security standards, carrying an AICPA SOC 2 Type II badge, with the page titled as a SOC 1 certified platform.

The page names intelligent authentications and user management without naming a mechanism, and no SSO, SAML, SCIM, role model or permission scheme appears anywhere on the platform, security or legal pages. Read only mode is a data integrity control over the customer's ledger rather than an access model governing who inside the organization can do what.

SourceNominal, nominal.so/company/securityread 2026-09-07

Agents log every step they take, producing documentation of what was analyzed, what logic was applied and what conclusions were reached, and every proposed entry carries the matching logic, source transactions and recommended treatment through to the approval record.

The security page names a detailed audit trail alongside a read only mode as standing platform properties, and Nominal states every action includes documentation, traceability and human approval workflows sufficient to support internal controls and external audit.

That reconstructs why an agent acted rather than reporting that it did, and it is Nominal's own record rather than logs from the customer's ERP.

SourceNominal, nominal.so/company/security and nominal.so/blog/ai-agents-in-finance-accountingread 2026-09-07

The Shadow Ledger and the transaction level data foundation are the accounting record the product maintains, which counts toward Knowledge; deleting them would delete the ledger. Learning from historical data and user interactions across cycles is absorbed into the agents rather than kept as memory. No memory layer with a stated scope and lifetime is documented, and nothing states where memory lives or how a buyer deletes it for one customer.

SourceNominal, nominal.so/our-platformread 2026-09-07

No hosting region, customer environment option or selection surface is documented anywhere the disclosure would live.

The security page mentions hosting only in passing, saying that from hosting to user management private data stays private, and names no region, no customer cloud and no on premises option; the published privacy policy is website scoped by its own terms and states Nominal does not access or interact with customer systems through it, and no data processing addendum or subprocessor list is published. There is no trust center either, so this is an absence rather than a gated surface.

SourceNominal, nominal.so/company/security and nominal.so/legal/privacy-policyread 2026-09-07

Five agents ship as named products, each with its own page and its own job: Flux Analysis agents explain variance, Bank Reconciliation agents match bank activity continuously, Transaction Matching agents match across ledgers, entities and systems including intercompany, Trigger agents activate next steps when thresholds are crossed or anomalies appear, and Transaction Patrol agents monitor activity. Each is a whole product doing its own work when selected, and the others remain intact without it.

SourceNominal, nominal.so/our-platformread 2026-09-07

Triggers are a named product rather than an implied capability: Trigger agents monitor financial data in real time and activate next steps automatically across systems and workflows when thresholds are crossed, data shifts, or anomalies appear, and Transaction Patrol agents monitor account activity continuously to detect anomalies and risks early. Alongside those, agents run between period ends as transactions flow in from connected ERP and bank feeds, and continuously review close status across entities and flag delays before they compound, which is state and schedule driven work rather than user invocation.

SourceNominal, nominal.so/our-platform and nominal.so/blog/agentic-ai-accountingread 2026-09-07

No model provider is named on any page read, and no customer or admin choice of model or routing is documented; Nominal describes deterministic AI agents built for accuracy, auditability and control, which speaks to how the agents behave rather than which model runs them. The privacy policy is website scoped and publishes no subprocessor list, so the usual place a provider would be named carries nothing either.

SourceNominal, nominal.so/our-platform and nominal.so/legal/privacy-policyread 2026-09-07

No API, SDK or MCP server for calling Nominal from outside is documented; the site navigation runs Platform, Agents, Solutions, Company and Knowledge Base with no developer surface, and every connection named runs outward into the customer's ERPs, banks and procurement systems, which counts toward Integrations.

SourceNominal, nominal.so/our-platformread 2026-09-07

Nothing tests the agents. Nominal says the agents learn from historical data and user interactions to improve accuracy over time and improve every cycle, which is an outcome claim: no change is under test, no result is readable, and nothing describes a harness, scored cases, a release gate or a controlled loop the customer runs or receives output from. The confidence thresholds that clear high confidence matches automatically gate individual outputs and count toward Human oversight, and the audit trail toward Observability.

SourceNominal, nominal.so/blog/agentic-ai and nominal.so/our-platformread 2026-09-07

Agents work over structured financial data pulled from connected ERPs, ledgers and bank feeds and write adjustments back through those connections, so reach runs through programmatic integrations rather than an interface an agent operates. No browser, desktop or remote computer control is documented on any page read, and Nominal's read only mode over customer systems points the other way.

SourceNominal, nominal.so/our-platformread 2026-09-07