Continuous close use case introduction
Finance Digital Twin: Continuous Financial Position
What BP is really asking for: always know where the books stand right now
Purpose: show that Continuous Close is not “close faster.” It is the ability to always know where the books stand right now. The books are continuously created from actuals, deterministic forecasts, probabilistic estimates, reconciliations, accruals, provisions, controls, evidence, and certification readiness. Month-end becomes certification, not creation.
Transactions
Actuals
Forecasts
Estimates
Reconcile
Accrue
Provision
Confidence
Certify
Replay
Run the live demo from here
Click left-to-right: setup, command center, proof, replay, technical pack
This launchpad is the intended validation path. Start here, then use the linked runtime surfaces to show that Continuous Close is a governed finance runtime rather than a static dashboard.
1. Run Finance As Of NowOpen the Continuous Finance Command Center and answer the current financial position question.
2. Live Demo CockpitClick through event injection, KPI movement, evidence, approval, replay, and value proof.
3. Continuous Close RuntimeShow events, readiness, accruals, provisions, evidence, value, and runtime impact.
4. Close Command CenterDrill into mission readiness, open blockers, controller actions, and ownership.
5. Close IntegrityShow audit/control posture, source health, evidence coverage, and safe degradation.
6. Credibility PayloadExpose source lineage, calculation basis, agents, replay, and provenance.
7. Replay StudioReconstruct the decision path from signal through policy, evidence, action, and learning.
8. Live Impact SectionJump to the value-impact proof rows and before/after readiness movement.
North Star
0 Day Close: month-end becomes certification of books that are already being continuously created.
If we closed the books right now, what would BP's Balance Sheet, P&L, Cash Flow, accruals, provisions, exposure, confidence, exceptions, and expected variance be?
Current Close Status93%
Actuals87%
Deterministic8%
Estimated5%
Expected Close Variance0.3%
Financial Position Confidence97.4%
Outstanding Exceptions12
Expected Variance<$500K
Finance portfolio analogy
Like Robinhood for enterprise finance, but governed
Robinhood does not wait until month-end to tell a user portfolio value. It continuously calculates positions, prices, cash, dividends, and unsettled trades. BP Finance should work the same way: actual transactions plus known future charges plus estimated missing activity produce a current financial position with confidence ranges.
| Portfolio concept | Finance Digital Twin equivalent |
|---|
| Positions | Actual posted balances and subledger activity |
| Market prices | Latest operational drivers, rates, volumes, claims, trips, utilities, trading positions |
| Cash | Bank, treasury, payments, receipts, unsettled items |
| Unsettled trades | Accruals, provisions, estimates, unresolved exceptions |
| Current value | Continuous financial position with confidence and expected variance |
Three accounting categories
Actuals, deterministic, probabilistic
| Category | Meaning | Examples | Confidence |
|---|
| Actuals | Known posted activity | Invoice posted, payment posted, journal posted, bank transaction, goods receipt | 100% |
| Deterministic Forecasts | Known future accounting impact | Lease expense, depreciation, amortization, subscription fees, contractual charges | 99% |
| Probabilistic Estimates | Predicted activity not yet known | Travel claims, utilities, legal expenses, bonus accruals, inventory adjustments, revenue true-ups | Confidence range |
Architecture needed
Five runtime layers
| Layer | Role |
|---|
| Financial Event Fabric | Captures SAP ECC, S/4, CFIN, Ariba, Treasury, Payroll, Bank, Trading, and Asset events continuously. |
| Continuous Reconciliation Engine | Continuously reconciles AP, AR, GL, Bank, Subledger, and Intercompany into reconciled, exception, or pending states. |
| Accrual Intelligence Engine | Continuously estimates travel, utilities, bonuses, legal, inventory, and revenue impact with confidence and variance risk. |
| Provision Intelligence Engine | Continuously estimates bad debt, claims, environmental, decommissioning, tax, and legal provisions with confidence ranges. |
| Continuous Close Cockpit | Shows current financial position, close status, actuals, deterministic forecasts, estimates, exceptions, and expected close variance. |
New agent family
Agents that create the continuous position
| Agent | Purpose |
|---|
| Reconciliation Agent | Continuously validates balances and explains breaks. |
| Accrual Agent | Predicts missing expenses and drafts evidence-backed accruals. |
| Provision Agent | Predicts future liabilities and provision ranges. |
| Journal Recommendation Agent | Suggests journals with policy, evidence, and approval gates. |
| Variance Agent | Compares actuals to estimates and measures learning variance. |
| Confidence Agent | Calculates reliability of the financial position. |
| Certification Agent | Determines whether Balance Sheet, P&L, and Cash Flow are ready to certify. |
Continuous Close Agent Architecture
Seven specialized agent layers, not generic task agents
This is the architecture BP should see: event agents maintain the signal layer, reconciliation agents maintain accounting truth, accounting and forecasting agents maintain the financial position, control and certification agents maintain governance, and learning agents improve the runtime after actuals arrive.
| Layer | Purpose | Representative agents |
|---|
| Layer 1 - Financial Event Agents | Continuously monitor enterprise finance signals before period end. | AP Event Agent, AR Event Agent, Treasury Event Agent, Asset Event Agent, Payroll Event Agent |
| Layer 2 - Continuous Reconciliation Agents | Make close readiness depend on reconciled financial state, not dashboard refreshes. | AP Reconciliation Agent, AR Reconciliation Agent, Bank Reconciliation Agent, Intercompany Reconciliation Agent, Subledger Reconciliation Agent |
| Layer 3 - Accounting Intelligence Agents | Create accounting readiness from accepted activity, missing invoices, provisions and journal risks. | Accrual Agent, Provision Agent, Journal Recommendation Agent, Journal Risk Agent, Revenue Recognition Agent |
| Layer 4 - Forecasting Agents | Answer the operating question: if we close now, what do the books look like? | Financial Position Agent, Close Position Agent, Variance Forecast Agent, Cash Forecast Agent, Exposure Forecast Agent |
| Layer 5 - Control Agents | Keep autonomy inside policy, control, SOD, evidence and approval boundaries. | Policy Agent, Compliance Agent, Segregation of Duties Agent, Exception Management Agent |
| Layer 6 - Certification Agents | Move month-end from creation of books to certification of continuously maintained books. | Close Readiness Agent, Controller Certification Agent, CFO Certification Agent |
| Layer 7 - Learning Agents | Improve the financial position runtime after actuals, exceptions and human decisions are known. | Forecast Learning Agent, Reconciliation Learning Agent, Close Optimization Agent |
Bangalore demo set
The 10 agents to show
Do not show 25 agents. Show the minimum set that proves the runtime can continuously maintain financial position.
| Agent | Why BP cares |
|---|
| AP Event Agent | Real-time finance signals |
| AP Reconciliation Agent | Continuous close foundation |
| Accrual Agent | Travel/utilities estimation |
| Provision Agent | Finance judgment |
| Journal Recommendation Agent | Accounting automation |
| Policy Agent | Governance |
| Exception Management Agent | Human-in-loop |
| Financial Position Agent | Current Balance Sheet, P&L and Cash Flow |
| Close Readiness Agent | Continuous close KPI |
| Forecast Learning Agent | Continuous improvement |
Flagship agent
Financial Position Agent
This is the agent competitors usually do not have. It turns actual transactions, forecast accruals, forecast provisions, open risks, controls, and evidence into current Balance Sheet, P&L, Cash Flow, confidence, and exceptions.
| Runtime field | Value |
|---|
| Inputs | Actual transactions, Forecast accruals, Forecast provisions, Open risks, Control status |
| Outputs | Current Balance Sheet, Current P&L, Current Cash Flow, Confidence, Open exceptions |
| Policies | R2R-CLOSE-CERT-001, SOX-CLOSE-404, R2R-SOD-003 |
| Last run | 2026-06-12T12:55:00Z |
| Runs today | 8,442 |
| Average latency | 280ms |
CLI proof
bp-close-now should answer: if we closed the books right now, what would happen?
| Output | Value |
|---|
| Close Readiness | 94.2% |
| Balance Sheet | Ready |
| P&L | Ready |
| Cash Flow | Ready |
| Confidence | 97.3% |
| Outstanding Exceptions | 12 |
Architect-run runtime proof
Show the runtime from command line, not just the browser
Use this sequence when BP technology asks whether Continuous Close is real or only a UI: bp-now, bp-close-now, bp-financial-position, bp-digital-twin, bp-close-runtime, bp-runtime, bp-provenance, and bp-replay-close 2026-06-12T23:59:59.
| Badge | Meaning | Why BP should care |
|---|
| ● | Live API query at command execution time | Proves the command is calling the runtime, not a screenshot. |
| ◐ | BP-shaped seeded runtime baseline | Prevents overclaiming organic BP production volume before credentials exist. |
| ◇ | Estimated or model-derived output with disclosed basis | Shows forecast/accrual numbers are explainable. |
| ◆ | Deterministic replay, policy, or rules path | Shows auditability and reproducibility. |
Safe validation language: The runtime is live. The validation baseline is seeded and explicitly marked. We do not claim BP production SOR connectivity until BP credentials and feeds are supplied; production SOR access is intentionally fail-closed.
Pre-close cadence
90% complete before month end
Continuous close companies start close work before period end. The goal is not a faster scramble after Day 0; it is continuous preparation and exception resolution before certification.
| Timing | Continuous close activity |
|---|
| Day -7 | Begin reconciliations and identify open exceptions. |
| Day -6 | Refresh accrual calculations and evidence coverage. |
| Day -5 | Refresh provisions and material estimate ranges. |
| Day -4 | Validate exceptions, owners, controls, and approvals. |
| Day -3 | Review variance, expected close position, and certification blockers. |
| Month End | Certify the books instead of creating them. |
Probabilistic estimate example
India travel expense accrual
Probabilistic estimates are not guesses. They are evidence-backed predictions with confidence, variance tracking, and learning when actuals arrive.
| Item | Value |
|---|
| Estimate | India travel expense accrual |
| Historical pattern | $3,000 per employee |
| Open trips | 120 |
| Estimated accrual | $360,000 |
| Confidence | 92% |
| Learning | When actual claims arrive, variance is measured and the model is adjusted. |
Use case story
A financial signal appears before month-end
A financial signal appears before month-end. bp Sphere detects it, explains it, routes it, resolves it, proves it, and shows enterprise impact from analyst to CFO.
| Runtime object | Value |
|---|
| Signal | SIG-ACCRUAL-MAINT-8200 |
| Case | CASE-ACCRUAL-8200 · Missing accrual detected before month-end |
| Entity / amount | UK Retail · $8.2M |
| Policy | R2R-ACCR-014 v3.2 + DOA-R2R-017 + SOX-CLOSE-404 |
| Human boundary | Agent can draft and route; controller approves before any posting/write-back. |
| Trace | TRACE-CC-20260618-ACCR-8200 |
Why this matters
Executive “so what”
- CFO sees close health continuously, not after close packs are assembled.
- Tower leaders see blockers by mission, owner, materiality, and SLA.
- Analysts validate decisions with evidence instead of searching systems manually.
- Technology and audit see deterministic replay, policy version, and evidence lineage.
How to show live
Use the same close event and show KPI movement
The live proof comes from /api/continuous-close/live-impact and the Continuous Close page’s ValueEventRecord-backed live impact ledger. In the demo, open the CFO screen first, resolve the missing-accrual event, then return to CFO and show the readiness and confidence change.
Impact events4Canonical value events linked to the close case.
Reportable value$8,413,750Traceable, reportable value from the impact runtime.
Measured value$8,200,000Measured and attested value where available.
Avg confidence88.0%Evidence-backed value confidence.
| Metric | Before | After | Delta |
|---|
| Close Readiness | 82% | 86% | +4 pts |
| Material Exceptions | 17 | 16 | -1 |
| Predicted Close Delay | 2.5 days | 1.8 days | -0.7 day |
| Evidence-Backed Confidence | 91% | 94% | +3 pts |
| Manual Effort Avoided | 430 hours | 380 hours | -50 hours |
| Financial Exposure Controlled | $0 | $8.2M | +$8.2M |
Recommended 12-minute demo
Minute 1-2 · CFO view
Ask: are we ready to close? Show readiness, material exposure, confidence, and what changed in the last 24 hours.
Minute 3-5 · Tower mission control
Show which mission is blocked, who owns it, which agent acted, and what human action is required.
Minute 6-9 · Analyst workspace
Open the missing accrual case. Show summary, evidence, policy, recommendation, journal preview, and governed approval boundary.
Minute 10-11 · Runtime proof
Show signal, context, policy version, agent inputs/outputs, evidence lineage, fallback path, replay ID, and audit trace.
Minute 12 · Return to CFO
Show close readiness, material exceptions, predicted delay, confidence, and manual effort avoided changing in real time.
Screen sequence
| Screen | Purpose | What to prove |
|---|
| 1 CFO Command Center | So what? | Business value, close readiness, materiality, confidence. |
| 2 Tower Mission Control | Where is the issue? | Operational ownership, blockers, agent/human split. |
| 3 Analyst Workspace | How does work get done? | Evidence, policy, recommendation, action boundary. |
| 4 Runtime Proof | Why believe it? | Trace ID, policy version, agent I/O, evidence lineage, replay. |
Live value attribution
These are the value rows to narrate when BP asks whether the page is live or just a dashboard. The value is tied to case, evidence, policy, decision, outcome, replay, and calculation method.
| Impact | Value | Runtime basis | Provenance |
|---|
| Financial exposure controlled | $8,200,000 | Aggregated from accepted-service amount and controller-attested accrual ValueEventRecord rows | computed |
| Manual effort avoided | 50 hours | Derived from the human_effort_augmentation ValueEventRecord calculation_outputs for evidence-search and journal-prep effort. | ledger_derived_estimate |
| Close delay reduced | 0.7 day | Derived from the productivity_improvement ValueEventRecord calculation_outputs for critical-path delay movement. | ledger_derived_estimate |
| Control improvement | $120,000 | Aggregated from SOX-evidence / policy-pinning ValueEventRecord rows | computed |