Rillet

Rillet is a cloud ERP that presents its own general ledger as the accounting system of record, with close management, revenue recognition, reporting, native integrations, and Aura AI in the same product. Rillet says Aura can answer questions against the ledger, prepare scheduled workflows, flag exceptions, and propose accruals; its close material describes reconciliations, audit logs, flux analysis, and controller review. [S1][S6]

This is a fit for a controller and finance-systems owner who are prepared to replace an accounting core, not merely add a task overlay. The potential benefit is fewer handoffs between operational inputs, ledger, close workflow, and evidence. The trade-off is correspondingly wider implementation risk: the buyer must validate data migration, policy and chart configuration, source integrations, posting authority, approval paths, audit exports, commercial scope, and a recoverable cutover. Treat every AI-produced result as reviewable preparation until the buyer has proven the configured controls in a parallel close.

Best for: Finance teams replacing a legacy ERP with a controlled cloud accounting core, reliable source systems, clear accounting policies, named approvers, and capacity to test migrations and close controls in parallel.

Not for: Teams that only need a lightweight close checklist over their retained ledger; cannot assign accountable accounting, systems, data, security, and migration owners; or expect unattended accounting judgment or posting before testing their specific tenant controls.

Reconcile imported finance activity

• Trigger: An approved source system, bank feed, billing system, or file delivers a period's data.
• Inputs: Permitted source fields, entity and account mapping, customer or vendor references, accounting policy, prior approved treatment, matching rules, period status, and an exception owner.
• Output: A prepared reconciliation with matched activity, source links, proposed accounting treatment, and unresolved differences. Rillet describes a perpetual ledger, native integrations, and automated reconciliation as parts of its finance workflow. [S1][S6]
• Human checkpoint: An authorized accountant verifies source completeness, mapping, cutoff, calculation, policy treatment, evidence, and every material exception before approval or posting.
• Success measure: Source-completeness rate, match accuracy, unresolved-item age, reviewer overrides, reconciliation breaks, post-close adjustments, evidence coverage, and preparation and review time.

Coordinate a controlled month-end close

• Trigger: Close tasks become due, inputs arrive late, a reconciliation fails, a reviewer rejects an item, or a period is ready for sign-off.
• Inputs: Close checklist, task owners, dependencies, source evidence, reconciliations, schedules, proposed entries, approval matrix, period-lock policy, and escalation rules.
• Output: A visible close state that connects ownership, workpapers, exceptions, approvals, and audit history. Rillet describes close management, an exportable immutable audit log, and AI-drafted flux analysis inside the ERP. [S6]
• Human checkpoint: The controller decides whether evidence is sufficient, resolves control breaks, approves final accounting decisions, and controls period close. A checklist or agent status is not a substitute for that judgment.
• Success measure: Close duration, overdue or blocked task age, unresolved exceptions, missing evidence, approval-cycle time, post-close changes, and audit-request response time.

Prepare a recurring accounting schedule

• Trigger: A scheduled accrual, revenue-recognition event, intercompany process, fixed-asset or prepaid schedule, or consolidation step reaches its agreed close point.
• Inputs: Approved source package, period and entity boundaries, contractual or transaction data, accounting policy, chart and dimension mapping, calculation rules, prior approved treatment, and a named reviewer.
• Output: A proposed schedule, journal, or exception queue. Rillet describes advanced revenue recognition, multi-entity capability, and Aura workflows that can process accounting data. [S1]
• Human checkpoint: An accountant checks source eligibility, policy, calculation, period, entity, currency, exception, and approval status before final downstream action.
• Success measure: Calculation accuracy, policy exceptions, reviewer correction rate, schedule completion time, recurring rework, posting errors, and reproducibility.

Review an AI-assisted finance exception

• Trigger: Aura flags an anomaly, produces a flux draft, proposes an accrual, or a user asks for an accounting analysis against the ledger.
• Inputs: Permitted ledger data, source documents, accounting policy, materiality threshold, user role, explanation or proposal, and linked activity history.
• Output: A reviewable exception, explanation, or proposed accounting action. Rillet says Aura can query the general ledger, schedule workflow agents, flag exceptions, and propose accruals. [S1]
• Human checkpoint: The assigned accountant determines whether the output is complete, correct, supported, authorized, and appropriate for the period; unsupported or ambiguous output remains an exception.
• Success measure: Material issue detection, false-positive and missed-exception rate, reviewer agreement, correction time, unapproved-action attempts, auditability, and repeat error rate.

1. The buyer defines the accounting boundary: entities, source systems, data owners, chart and dimension design, accounting policies, close calendar, approvals, exception rules, and evidence needed for each material workflow.
2. Rillet is deployed as the accounting core and connects permitted sources. Its help center says it can replace a legacy ERP and names several direct integrations; those public descriptions do not establish the buyer's available connector or authority. [S1][S2]
3. The system ingests and organizes finance data in the general ledger. Rillet says Aura can use live ledger context to answer questions, run scheduled workflows, flag exceptions, and propose accruals. [S1]
4. Accountants inspect reconciliations, schedules, flux explanations, source evidence, exceptions, and proposed actions. Rillet's close material says review and audit evidence are part of the close process; buyer configuration must determine who can approve, post, correct, or close. [S3][S6]
5. The buyer retains reconciliations, approval records, audit-log exports, source references, exceptions, migration evidence, and a manual fallback so it can reproduce, correct, or finish the close if an integration or workflow fails.

The decisive control is the full path from source data through mapping and accounting policy to reviewed posting or exception. A clean dashboard or a balanced proposed entry does not prove input completeness, appropriate cutoff, correct revenue treatment, valid authority, or a recoverable change.

Verified facts

• Rillet positions itself as an AI-native cloud ERP and describes a perpetual general ledger, revenue recognition, native integrations, close management, real-time reporting, and Aura AI. [S1]
• Rillet's help center says it is cloud-based, may replace a legacy ERP, names Salesforce, Stripe, Ramp, Brex, and Rippling as direct integrations, and says its implementation team performs the majority of the work. [S2]
• Rillet's security page states annual SOC 1 and SOC 2 Type II audits, encryption at rest and in transit, AWS hosting, logical segregation, SSO options, AI-action logging, and a no-auto-posts-without-approval statement. [S3]
• Rillet's public DPA annex lists AWS, Stytch, OpenAI, Anthropic, Langsmith, and other named subprocessors and functions. [S5]

Vendor claims

• Rillet says its implementation commonly takes four to six weeks depending on complexity and migration data; a separate close article says four to eight weeks. The apparent range needs a written plan for the buyer's scope. [S2][S6]
• Rillet says agents can automate accounting tasks and that the product can support a zero-day or continuous close. These are positioning and capability claims, not evidence that a particular migration or close will meet a buyer's timeline or control requirements. [S1][S6]
• Rillet's close article says customers save seven days on average on a monthly close. The public material does not establish population, baseline, conditions, or applicability to another buyer. [S6]

B2Bagents assessment

Rillet is best evaluated as an accounting-platform replacement with AI-assisted close work, not as a point solution. A buyer that needs its current ERP to remain the system of record should verify whether a less invasive close or reconciliation layer better matches the problem. A buyer whose existing ledger, integration maintenance, and spreadsheet reconciliation are the bottleneck may find the integrated architecture attractive, but should fund it as a migration and controls project with a parallel-close exit route.

Limitations and unknowns

• No public dollar pricing, included feature list, implementation price, usage limit, term, renewal, cancellation, export commitment, or transition assistance was found. [S2][S4]
• The public pages do not establish the buyer's specific source systems, connector edition, fields, cadence, error handling, entity or currency support, mapping, read/write rights, journal authority, approval rules, period locks, audit export, or rollback behavior.
• Public security and legal materials do not replace customer-specific evidence for the applicable SOC reports, service scope, DPA, model routing, training or reuse, data residency, retention, deletion, SSO/RBAC, incident response, support access, backup, and recovery. [S3][S4][S5][S7]
• Replacing the ledger can surface historical-data errors, incomplete source records, inconsistent policy, dormant entities, unsupported revenue rules, custom integrations, or audit evidence gaps. Those problems may be exposed by, but are not solved by, a faster accounting interface.

Good-fit conditions: a supported finance-stack pattern; documented accounting policies; a controllable chart, entity, and source migration; a clear approval model; willing finance and systems owners; and enough runway for a parallel close and recovery testing.

Poor-fit conditions: an accounting core that must remain unchanged; a near-term close or audit that leaves no migration capacity; unowned source systems; unresolved policy or revenue questions; no review coverage; or a procurement assumption that an AI agent can independently make accounting decisions.

Predictable failure modes and controls:

• A migration drops history, mappings, or context. Reconcile opening balances, historical periods, source totals, schedules, and reports against the old system; define acceptance thresholds and a cutover rollback before switching.
• An integration creates a complete-looking but incomplete record. Certify source systems and fields, track freshness and exceptions, reconcile feeds, use least-privilege identities, and retain a manual import or close path.
• An AI proposal treats a policy exception as routine. Keep materiality, accounting-policy interpretation, approval, and final posting with accountable people; test edge cases and sample outputs.
• Approval configuration differs from the intended control design. Demonstrate roles, maker-checker separation, delegation, period locks, rejected approvals, user removal, logs, and emergency disablement in the actual tenant.
• A faster close produces less reviewable evidence. Require source-to-entry traceability, exportable logs, exception ownership, workpaper retention, and an audit-request dry run.

Assign an executive sponsor, controller, accounting-policy owner, finance-systems owner, data owners for every source, security and privacy reviewer, implementation lead, internal audit or external-auditor liaison, and an incident and cutover owner. Rillet says it provides white-glove implementation with CPAs and estimates four to six weeks, but migration complexity and data history are explicitly variables. [S2]

Start with an inventory of entities, ledgers, charts, dimensions, source applications, field-level data flow, integrations, report and audit dependencies, custom close workbooks, roles, approvals, retention, and the old-system exit path. For each integration, demonstrate the actual service identity, reads, writes, refresh schedule, mapping, retry behavior, alert, change history, access removal, and manual fallback. The product's named integrations are a diligence starting point, not proof of a production-ready connector. [S1][S2]

Use a historical close and one live close in parallel. Do not decommission the previous process until source totals, opening balances, schedules, reconciliations, entries, approvals, reports, audit records, exceptions, correction behavior, and recovery meet pre-agreed acceptance criteria. The exit plan should preserve accessible historical records and exports, stop or revoke integrations, route unfinished work to the prior controlled process, and document responsibility for data deletion and transition support.

Rillet's security page states that it uses AWS, AES-256 encryption at rest, TLS 1.2 or later in transit, logical customer-data segregation, and SSO options for Google, Microsoft, and Rippling. It states annual SOC 1 and SOC 2 Type II audits and says AI actions are logged. [S3] Its July 2026 MSA says it will maintain a security-control audit in the form of a SOC 2 Type II report and/or HITRUST r2 certification, and a written security program aligned with a generally accepted framework. [S4]

Rillet's public DPA annex names subprocessors including AWS, Stytch, OpenAI, Anthropic, Langsmith, Honeycomb, Salesforce, Slack, and Pylon, with listed functions and locations. That is useful for initial diligence, but it does not prove which services, model providers, data categories, transfer mechanisms, retention rules, or controls apply to the buyer's order. [S5]

The public privacy policy says it does not cover information Rillet processes for business customers through the platform; customer agreements govern that processing. [S7] Before uploading production finance data, obtain the applicable DPA and current annex, data-flow diagram, all model and subprocessor terms, residency and transfer terms, retention and deletion, training or reuse terms, cryptographic and access-control detail, SSO/RBAC and approval behavior, audit-log export, vulnerability and incident process, backup and recovery evidence, support-access process, and contractual remedies. Test those answers in the selected tenant.

Rillet says it prices by features used and business complexity, rather than per seat or revenue. It does not publish a public dollar rate card. [S2] Its MSA says the order form details purchased features and that some modules can carry an additional charge. [S4]

Ask for a written proposal that separates platform features, entities, users and roles, source systems and integrations, historical-data migration, revenue or multi-entity requirements, implementation services, CPA support, training, AI or usage limits, support tier, added modules, service levels, term, renewal, cancellation, export, data deletion, and transition assistance. Compare the full cost with the buyer's migration, testing, parallel-close, controls, audit, exception, and operating effort—not only the software subscription.

1. Scope: Choose one entity or a representative limited group, one recurring close procedure, known source systems, prior approved output, a named reviewer, and a manual fallback. Keep policy changes and broad production write access out of the first AI-assisted workflow.
2. Baseline: Record close days, preparation and review time, source completeness, reconciliation breaks, post-close adjustments, audit-evidence retrieval time, exception age, correction effort, incidents, integration maintenance, and current total cost.
3. Test: Reconcile historical data, then run a live period in parallel. Include delayed, duplicate, missing, reversed, and corrected source activity; changed account or dimension mappings; foreign-currency and intercompany cases; revenue-rule exceptions; rejected approvals; role changes; connector failure; log export; access revocation; and cutover recovery.
4. Pass threshold: Agree before implementation on balance and source reconciliation, correct accounting treatment, complete evidence, exception visibility, reviewer workload, approval enforcement, report agreement, audit-log export, credential removal, recovery time, close impact, and total deployed cost.
5. Stop conditions: An unexplained material difference; unapproved or irreversible posting; missing source or audit evidence; inability to reproduce an entry; failed access revocation; incomplete recovery; terms that do not cover required data or support controls; or no net improvement after the buyer includes review and migration effort.

Compare Rillet with the directory's finance-close listings including Basis, Numeric, FloQast, Maxima, Puzzle, Campfire, and Truewind. Rillet's public materials position it as the system of record, while several alternatives may retain an existing ledger or serve a more bounded workflow. Compare accounting-core replacement versus overlay architecture, source-system coverage, entity and revenue complexity, migration burden, workflow depth, agent authority, approval controls, audit evidence, security terms, implementation model, commercial model, and the buyer's current bottleneck.

1. What exactly can Rillet and Aura read, prepare, submit, approve, post, or change in our tenant? Demonstrate a representative workflow with the real identities, fields, mappings, approval gates, logs, rejected cases, period locks, credential revocation, and emergency stop. [S1][S3]
2. What is the migration acceptance plan? Obtain the complete scope, data and entity inventory, historical period, opening-balance and report reconciliation, policy mapping, timeline, owners, parallel-close criteria, cutover decision, rollback, and audit evidence plan. [S2][S6]
3. Which integrations are supported in our exact configuration? Show the connector edition, data objects, field mapping, synchronization timing, read/write authority, errors, retries, alerts, support boundary, and manual fallback for each source. [S1][S2]
4. Which security and privacy claims become contractual commitments? Request the current SOC reports and scope, DPA, subprocessor and model list, data flows, residency, transfer terms, retention, deletion, training or reuse, SSO/RBAC, audit export, incident terms, and support access. [S3][S4][S5][S7]
5. What does feature-and-complexity pricing include and exclude? Get entities, modules, integrations, implementation, migration, support, usage, overages, additional services, term, renewal, cancellation, export, and transition in writing. [S2][S4]
6. What would make this pilot or migration fail? Agree in advance on material reconciliation differences, incorrect treatment, missing evidence, failed approvals, unsafe access, unacceptable reviewer workload, incomplete recovery, and total-cost thresholds.