Campfire

Campfire combines an ERP and close-management system with Ember agents. Ember can prepare accruals, match and reconcile transactions, draft flux commentary, and run configured close checks; accountants review the evidence, handle exceptions, and approve postings. This is a core accounting-system decision, not a lightweight add-on to any ERP. [S1][S2][S3]

Best for: Finance teams with governed close procedures, named reviewers, and a willingness to standardize accounting operations inside Campfire.

Not for: Teams that need a standalone close copilot over an existing ERP, cannot migrate or integrate their accounting data, lack reliable close controls, or expect AI-generated accounting entries to post without accountable review.

Prepare recurring accruals for review

• Trigger: A finance owner begins month-end accrual preparation for an entity.
• Inputs: Historical bills, payments, journal entries, recurring-vendor patterns, the close period, and accounting configuration. [S2]
• Output: Proposed accrual groups and a draft journal entry with line-level reasoning and source links. [S2]
• Human checkpoint: An accountant can edit the amount, account, vendor, department, or description; nothing posts until that user creates the journal entry. [S2]
• Pilot measure: Accuracy of proposed amounts and offset accounts, preparation minutes per accrual group, correction rate, and post-close adjustments.

Investigate material variances

• Trigger: A controller selects a GL account, department, or cost center for flux review.
• Inputs: GL transactions, dimensions, materiality rules, and the relevant reporting period. [S1]
• Output: Editable variance commentary with supporting data intended for review, board preparation, or audit preparation. [S1]
• Human checkpoint: The controller checks the cited transactions, tests material explanations, and approves any external or management reporting.
• Pilot measure: Time to complete a reviewed flux explanation, explanation accuracy, reviewer corrections, and unexplained-material-variance rate.

Run a scoped close control

• Trigger: An accountant runs a configured Ember Skill during the close.
• Inputs: The Skill's stated procedure and scoped accounting records; Campfire's example for unrecorded liabilities includes AP bills, purchase orders, receiving records, vendor patterns, bank/card activity, and payroll. [S3]
• Output: A classified list of proposals, investigations, and exclusions with reasoning; the documented example can draft journal entries but stops before posting. [S3]
• Human checkpoint: The procedure owner sets the period, entity, currency, materiality threshold, and evidence hierarchy, then reviews each proposed accounting action.
• Pilot measure: Exception precision, exceptions found versus the manual control, close-control completion time, and unsupported-entry rate.

1. The controller defines the close calendar, entities, materiality thresholds, roles, approval rules, and a narrow initial procedure.
2. Campfire brings the accounting records and supporting information into its accounting and close-management environment. [S1][S2]
3. An Ember agent or Skill prepares a proposed result, links its reasoning and underlying records, and flags items that need review. [S1][S2][S3]
4. An accountant validates the support, edits or rejects proposals, and authorizes the journal entry or other controlled action. [S1][S2]
5. The finance lead compares results with the established close baseline, fixes rules and source data, and expands only when accuracy and control evidence are acceptable.

The core trade-off is scope. Campfire's value proposition depends on operating inside its ERP/accounting platform, which can make repeatable work and provenance more coherent but creates a larger migration, integration, and governance decision than a point workflow tool.

Verified facts

• Campfire documents Ember agents in always-running and on-demand modes for accounting operations, close preparation, flux analysis, transaction review, and period-end workflows. [S1]
• Campfire documents an accrual process that scans historical bills, payments, and journals, presents source-linked reasoning, permits edits, and does not post until a user creates the journal entry. [S2]
• Campfire documents Skills for recurring accounting procedures. Its unrecorded-liabilities example classifies results, shows reasoning, drafts journal entries, and stops before posting. [S3]
• Campfire says its Ember page has AI confidence thresholds, logged/reversible actions, and review-and-approve controls before posting. [S1]

Vendor claims

• Campfire states that it is SOC 1 and SOC 2 Type 1/Type 2 certified and that it encrypts data at rest and in transit, provides RBAC and full audit trails, and does not train on proprietary data. Request the reports, covered services, dates, exceptions, and contract terms. [S1]
• In Campfire's published Advisor360° customer story, the customer reports a three-day close versus six days on its prior system, 50% faster close time, and eight-plus monthly hours saved on cash reconciliations. Those are vendor-published customer results, not independently verified outcomes. [S4]

B2Bagents assessment

Campfire appears best suited to a finance team that wants to turn stable accounting judgement into narrowly scoped, reviewable procedures. The accrual and Skill examples preserve a recognizable accounting control: configured scope, visible evidence, edit/reject capability, and a human posting decision. The decisive pilot question is not whether Ember can produce an answer, but whether accountants accept its evidence and exceptions at a useful rate without weakening the close's approvals or audit trail. [S1][S2][S3]

Material unknowns

• Public pricing, contract minimums, trial terms, implementation fees, migration scope, and cancellation terms.
• Exact supported integrations, data direction, API scopes, field-level writes, and what applies to the buyer's plan.
• Contractual retention, deletion, export, residency, model-provider, subprocessor, support-access, and incident-response terms.
• How the described SOC reports and security statements apply to Ember, each deployment region, and the buyer's intended data flows.
• Independent evidence of accuracy, financial-control impact, or close-time improvement for the buyer's accounting mix; no B2Bagents hands-on test was performed.

Good-fit conditions: A controlled accounting team owns the core ledger and close checklist, can define materiality and evidence rules, has clean historical records, and can give an accountant authority to review exceptions and entries.

Poor-fit conditions: The buyer needs software that sits entirely outside the ERP, has an unreconciled or poorly owned chart of accounts, cannot migrate accounting operations, or would treat generated commentary as approved financial reporting.

Predictable failure modes and controls:

• A recurring pattern can look plausible but be wrong in the new period. Set materiality thresholds, sample proposed entries against source documents, and require accountant approval before posting. [S1][S2]
• Incomplete or inconsistent source data can generate incomplete close checks. Test known bad, missing, duplicate, and cross-entity records before relying on a Skill. [S3]
• A highly connected ERP can widen the effect of bad permissions. Review user roles, approval limits, integration scopes, period locks, audit records, and reversals before production use. [S1]
• A successful case study can conceal migration and operating effort. Treat Advisor360°'s reported two-week onboarding and outcomes as a scoped vendor-published example; build the buyer's own migration plan and baseline. [S4]

Start with a controller as accountable owner and one staff-accountant reviewer. Choose one entity and one recurring procedure, such as vendor accrual preparation or unrecorded-liability review. Establish the baseline close duration, exceptions, adjustments, reviewer minutes, and data-quality defects before enabling the procedure.

The buyer should then map every input record, user role, approval limit, integration scope, reporting output, and correction path. Campfire's public materials establish that Ember operates on accounting records and that its vendor-published customer story involved Ramp, HSBC, and JPMorgan, but they do not prove the buyer's required connector, API permission, migration path, or write controls. [S1][S4]

Before expanding, obtain and test exports, journal-entry reversals, audit-log access, period-lock behavior, user deprovisioning, retention/deletion process, and the exit plan for accounting data.

Campfire's Ember page says all agent actions are logged, attributed, and reviewable; it also states encryption at rest and in transit, RBAC, full audit trails, no training on proprietary data, and SOC 1/SOC 2 Type 1/Type 2 certifications. This is useful vendor-provided evidence, but it does not replace a buyer-specific security review. [S1]

Ask for current reports and scope, the DPA and privacy terms, data residency and retention choices, subprocessors and model providers, support-access controls, incident commitments, export/deletion workflow, and confirmation of which records Ember may read, draft, change, or post. In the operating design, separate a user's ability to configure a Skill from their ability to approve or post an entry, and routinely inspect evidence and corrections for material accounts.

Public pricing was not documented in the primary sources reviewed. Request a written quote that explicitly covers the accounting platform, legal entities, users and roles, data migration, implementation, integration work, AI or usage limits, support level, minimum commitment, renewal, fees for growth, and the usable export/termination path.

Price the pilot around the one close procedure being tested. A broad ERP comparison can hide the cost of migration, parallel close, data cleanup, and integration controls that determine whether the AI workflow is reliable.

Run a parallel close on one entity and one procedure for two cycles. A suitable first scope is recurring accrual preparation or unrecorded-liability review. Before each run, lock a baseline for volume, prep time, reviewer time, correction rate, post-close adjustments, audit requests, and days to close.

Create a test set with known recurring vendors, a new vendor, a missing invoice, an incorrect department, a duplicate payment, a multi-entity item, a reversal, and a deliberately immaterial variance. Require that every proposed item has source-linked evidence and a named disposition: accept, edit, investigate, or reject.

Expand only if the reviewed proposal accuracy, preparation-time reduction, exception handling, and audit-trail completeness meet the finance owner's pre-agreed threshold without extra post-close adjustments. Stop, narrow, or revert if an entry posts without required approval, source evidence cannot be reconstructed, material exceptions are missed, or required data/permission controls are unavailable.

Compare Campfire with finance-close options on deployment scope first: whether the buyer is adopting a full accounting system or adding a close layer over an existing ERP. Then compare accounting-data coverage, migration burden, controlled drafting versus posting, evidence links, role/approval controls, audit export, entity handling, integrations, pricing, and results from a buyer-run parallel close.

1. Which Ember workflows can draft, change, or post accounting entries in our planned configuration? Review least-privilege roles, approval thresholds, logs, reversals, and period locks. [S1]
2. Can we validate an accrual proposal down to its source transactions and override every line before it posts? Demonstrate the documented reasoning panel and edit/review flow on buyer data. [S2]
3. Which close procedures should we encode as Skills first, and who can publish or share them? Start with a stable procedure and keep a named finance owner accountable for its instructions and exceptions. [S3]
4. What does implementation include for our entities, historical data, integrations, and chart of accounts? Treat the Advisor360° onboarding example as a lead for diligence, not a delivery commitment. [S4]
5. Which security reports, privacy terms, data handling commitments, and model-provider terms apply to Ember and our region? Obtain current contractual evidence before production data is connected. [S1]