Forward deployed transformation hypotheses / Finance

# Finance: transformation hypotheses.

A process-led field guide for internal forward deployed engineers. Map the operation, investigate the causes of lost value, size the opportunity and inspect how the work could change.

By the Forward Deployed Team · Updated 24 October 2026 · Illustrative scenarios; not customer results.

Forward deployed transformation hypotheses

## Finance: the opportunity map

Start with the operating outcome. Follow the end-to-end process, identify a specific cause of lost value, then test the intervention. Eight process families below contain 24 working hypotheses; the visual investigations that follow illustrate selected mechanisms.

### Free working capital

Receivables, inventory, payment timing and recoverable balances. Measure one-time cash separately.

### Remove avoidable cost

Demand, specification, purchasing, process effort and loss. Distinguish spend removed from capacity released.

### Improve capital decisions

Asset use, investment discipline, liquidity and resilience. Preserve service and essential controls.

Evaluate every opportunity through: people and responsibilities · process and standard work · data and information · controls and decision rights · technology and adoption. AI is one possible enabler.

Process → cause → intervention → value → evidence

## The hypotheses to investigate

These are starting hypotheses, not diagnoses. Establish the baseline and the burden of proof before selecting a tool. All process families stay open as you scroll.

### Order to cash

Process stages: Customer setup → billing → cash application → collection

Missing order and delivery evidence may cause invoice disputes

Unapplied receipts may conceal collectible balances

Credit limits and inconsistent follow-up may extend overdue exposure

#### Proof or rejection: what evidence is needed?

Join orders, invoices, delivery confirmations, receipts and dispute timestamps; segment disputed and undisputed aging. Reject if customers cannot pay or commercial terms explain the delay.

#### Where AI could contribute

Extract invoice evidence, match receipts and classify dispute causes; AR owners approve settlements.

Size the opportunity / Cash released

Annual eligible credit sales ÷ 365 × verified days removed

Low / base / high illustration: $73m ÷ 365 × 2 / 5 / 8 days = $0.4m / $1.0m / $1.6m one-time cash release.

Cash release is not annual profit. Exclude taxes and balances outside the intervention scope.

### Procure to pay

Process stages: Supplier setup → purchase order → receipt → invoice → payment

Inconsistent receipt data may generate avoidable invoice exceptions

Duplicate supplier records may permit repeat payments

Missed payment terms may create late fees or sacrifice valid discounts

#### Proof or rejection: what evidence is needed?

Reconcile supplier masters, PO lines, receipts, payments and terms. Sample false duplicate matches; reject savings already recovered or prevented by existing controls.

#### Where AI could contribute

Match entities and documents, flag duplicates and prepare exception evidence; preserve payment authorization.

Size the opportunity / Annual processing capacity

Annual transactions × net minutes released ÷ 60 × loaded hourly cost

Low / base / high illustration: 120,000 transactions × 1 / 2 / 3 minutes ÷ 60 × $40 = $80k / $160k / $240k annual capacity value.

Subtract review and exception effort. Capacity is not automatically removed cost; do not also count it in another initiative.

#### Benchmark context: total AP process cost

APQC reports a cross-industry median of $6.00 per invoice across 5,846 organizations. It includes personnel, systems, overhead, other internal and outsourced costs. This is a process-cost reference, not an AI savings target.

Source: APQC · October 2026 ↗ . Compare the same cost scope, transaction definition and period before drawing a gap.

### Record to report

Process stages: Journal → reconciliation → intercompany → consolidation → close

Late upstream postings may drive close rework

Unexplained reconciling items may consume review time

Manual intercompany differences may delay consolidation

#### Proof or rejection: what evidence is needed?

Trace close tasks, journal sources and aged reconciling items across two closes. Reject if the critical path is a mandatory waiting period.

#### Where AI could contribute

Link reconciliations to source entries, propose matches and group recurring variance causes; accountants sign off.

Size the opportunity / Annual capacity value

Annual tasks × verified minutes released ÷ 60 × loaded hourly cost

Low / base / high illustration: 20,000 tasks × 2 / 5 / 8 minutes ÷ 60 × $45 = $30k / $75k / $120k annual capacity value.

Time released becomes cash savings only through a funded staffing, contractor or overtime change; check work transferred elsewhere.

### Planning and performance

Process stages: Plan → forecast → management reporting → business decision

Spreadsheet reconciliation may crowd out business analysis

Forecast errors may reflect missing operational drivers

Uniform allocations may hide unprofitable products or customers

#### Proof or rejection: what evidence is needed?

Reconcile management and ledger views, time reporting work and backtest forecasts at each historical cutoff. Reject if better predictions would not change a decision.

#### Where AI could contribute

Driver modelling, variance explanation and cost-to-serve analysis with traceable inputs; finance owns assumptions.

Example sizing model / Planning and reporting capacity

Annual reporting cycles × verified preparation hours released × loaded analyst cost

Low / base / high illustration: 1,000 reporting cycles × 1 / 2 / 3 hours × $75 = $75k / $150k / $225k annual analyst capacity value.

Define a reporting cycle consistently. Decision quality and forecast accuracy need separate validation; released analyst time is not automatically a budget reduction.

### Treasury and liquidity

Process stages: Cash position → liquidity forecast → funding → exposure

Unreported receipt timing may inflate liquidity buffers

Fragmented bank balances may create avoidable borrowing

Unreconciled exposures may distort funding and hedge decisions

#### Proof or rejection: what evidence is needed?

Use dated bank, AR/AP and debt schedules; backtest cash troughs and funding needs. Reject balances that are restricted or not available for pooling.

#### Where AI could contribute

Probabilistic forecasts and source-linked exposure summaries; treasury retains funding and hedge decisions.

Size the opportunity / Investment or carrying-cost effect

Verified avoidable capital × annual carrying-cost rate

Low / base / high illustration: $1m avoidable capital × 5% / 8% / 12% = $50k / $80k / $120k annual carrying-cost sensitivity.

Capital deferred, debt reduced and expenditure avoided are different outcomes. Rate and principal need treasury or investment-owner validation.

### Tax and statutory operations

Process stages: Transaction classification → filing → reconciliation → evidence

Poor transaction coding may increase correction effort

Fragmented filing evidence may cause repeated preparation

Unreconciled statutory balances may delay reporting

#### Proof or rejection: what evidence is needed?

Sample coding, filing adjustments and reconciliations with qualified owners. Reject opportunities that conflict with current local obligations; verify jurisdiction-specific treatment.

#### Where AI could contribute

Classify records and assemble review evidence; qualified specialists determine treatment and filings.

Size the opportunity / Annual capacity value

Annual tasks × verified minutes released ÷ 60 × loaded hourly cost

Low / base / high illustration: 20,000 tasks × 2 / 5 / 8 minutes ÷ 60 × $45 = $30k / $75k / $120k annual capacity value.

Time released becomes cash savings only through a funded staffing, contractor or overtime change; check work transferred elsewhere.

### Commercial and contract control

Process stages: Terms → consumption → billing → recovery

Invoices may not reflect negotiated rates or rebates

Unclaimed credits may persist because ownership is unclear

Incomplete cost attribution may hide margin leakage

#### Proof or rejection: what evidence is needed?

Reperform a sample of contract-to-invoice calculations and confirm recoverability. Reject amounts outside entitlement, limitation periods or dispute evidence.

#### Where AI could contribute

Extract rate clauses, reconcile usage and explain variances; owners validate contract interpretation.

Size the opportunity / Annual cost opportunity

Eligible annual spend × verified avoidable share

Low / base / high illustration: $2m × 2% / 5% / 8% = $40k / $100k / $160k gross annual opportunity.

Rates are assumptions, not typical savings. Confirm minimum commitments, switching costs and actual expenditure removal.

### Capital and benefit governance

Process stages: Appraise → prioritize → fund → track → realize

Optimistic project assumptions may overstate investment returns

Overlapping initiatives may count the same benefit twice

Released hours may be reported as removed expenditure

#### Proof or rejection: what evidence is needed?

Reconcile investment cases, dependency plans, ledger reductions and workload measures. Reject double counts and benefits without a realization owner.

#### Where AI could contribute

Compare scenarios and reconcile benefit evidence; investment decisions remain with accountable owners.

Example sizing model / Verified program benefit

Validated gross annual benefits × realization rate − recurring program cost

Low / base / high illustration: $500k × 60% / 75% / 90% − $100k = $200k / $275k / $350k annual net benefit sensitivity.

This reconciles the program total; it is not an additional benefit to add to the underlying initiatives. Validate benefits independently, deduct implementation costs and assign each benefit once.

Interactive tool / operating assumptions

## How large could the opportunity be?

Estimate released processing capacity from your own workload. This is an effort model, not a prediction of AI savings. Use the separate formulas above for cash, capital, margin, materials and energy.

Realized annual hours: 1,000

Annual capacity value after running cost: $30,000

First-year value after one-time cost: $0

Formula: tasks × minutes ÷ 60 × realization × hourly cost − running cost; subtract one-time cost in year one. Financial savings require an actual budget change. No inputs leave your browser.

Default calculation: 20,000 × 5 ÷ 60 × 60% = 1,000 hours; × $45 − $15,000 = $30,000 annual capacity value; less $30,000 implementation = $0 first year. Enable JavaScript to recalculate.

## Method and sizing notes

The opportunity structure combines three value lenses—cash release, cost optimization and capital resilience—with process, contract, working-capital, customer-management and investment analysis. Historical cost-transformation and finance operating-model references inform the process coverage. Every hypothesis requires evidence, a test that could reject it and accountable validation.

Historical transformation benefit ranges are not current benchmarks or AI effect sizes. This guide uses transparent low/base/high sensitivity examples instead. The eligible baseline, proposed change and realization route must be established for the organization. The examples are not additive, forecasts or customer results.

Prioritize by net value, strength of evidence, feasibility, time to value and service or control consequences. Deduct implementation, transition and recurring costs; assign one owner and one benefit record to prevent double counting.

All examples and evidence remain readable. Enable JavaScript to compare scenarios.

Finance / investigate the opportunity

## Invoice exceptions

The working hypothesis

Repeated invoice exceptions may come from inconsistent receipt and purchase-order data, rather than slow processing.

What would establish or reject the case?

Sample exceptions by root cause; compare matched supplier cohorts and inspect false matches before changing release rules.

### Measure the consequence

First-pass match rate; exception age; duplicate payments

AP interviews; purchase orders; receipts; invoice extraction; entity matching

### One receipt, two versions of an invoice

Document | Reference | Amount
--- | --- | ---
Invoice | INV-204 | $1,200
Invoice copy | INV 204 | $1,200
Receipt | RC-091 | 9 / 10 cartons

Document | Reference | Amount
--- | --- | ---
Invoice | INV-204 | $1,200
Invoice copy | INV 204 | $1,200
Receipt | RC-091 | 9 / 10 cartons

Review interpretation Route uncertain matches to an AP reviewer; retain payment authorization.

Route uncertain matches to an AP reviewer; retain payment authorization.

### AI use cases: how could the work change?

Two visual examples · inputs, interpretation and human decisions

An illustrative AI application: inspect the source records, the interpretation and the human decision.

AI / records & interpretation

#### Invoice exceptions: evidence assist

Duplicate candidate: INV-204 / INV 204 → same amount and supplier, review image.

Inputs: AP interviews; purchase orders; receipts; invoice extraction; entity matching.

Human action: Route uncertain matches to an AP reviewer; retain payment authorization.

AI / analysis & workflow

#### Exception root-cause grouping

Receipt quantities and exception notes

Validate: Compare the draft interpretation with a verified sample and record missing or incorrect conclusions before changing the process.

The proposed change fails the measured outcome, shifts work elsewhere, loses required controls or costs more than the verified benefit. Investigate those results before scaling.

Finance / investigate the opportunity

## Collections & disputes

The working hypothesis

Disputes that reach collections without delivery evidence may prolong cash conversion.

What would establish or reject the case?

Separate disputed from undisputed balances; compare resolution time and customer outcomes for a pilot cohort.

### Measure the consequence

Dispute resolution time; overdue balance; cash received

AR interviews; aging; dispute codes; delivery evidence; text classification

### An overdue balance is really a delivery dispute

Event | Day | Evidence
--- | --- | ---
Invoice due | 01 | Balance $8,400
Dispute opened | 03 | Missing proof of delivery
Reminder sent | 08 | Dispute still open

Event | Day | Evidence
--- | --- | ---
Invoice due | 01 | Balance $8,400
Dispute opened | 03 | Missing proof of delivery
Reminder sent | 08 | Dispute still open

Review interpretation Send the delivery evidence to the account owner before escalating collection.

Send the delivery evidence to the account owner before escalating collection.

### AI use cases: how could the work change?

Two visual examples · inputs, interpretation and human decisions

An illustrative AI application: inspect the source records, the interpretation and the human decision.

AI / records & interpretation

#### Collections & disputes: evidence assist

Dispute triage: missing delivery note → evidence request, not another reminder.

Inputs: AR interviews; aging; dispute codes; delivery evidence; text classification.

Human action: Send the delivery evidence to the account owner before escalating collection.

AI / analysis & workflow

#### Dispute evidence assembly

Order, delivery and customer correspondence

Validate: Compare the draft interpretation with a verified sample and record missing or incorrect conclusions before changing the process.

The proposed change fails the measured outcome, shifts work elsewhere, loses required controls or costs more than the verified benefit. Investigate those results before scaling.

Finance / investigate the opportunity

## Cash forecasting

The working hypothesis

Unrecorded timing changes in receipts and payments may explain forecast error better than aggregate seasonality.

What would establish or reject the case?

Backtest rolling forecasts using only information available at each forecast date; compare with the current method.

### Measure the consequence

Forecast error by week; liquidity buffer; unexplained variance

Treasury interviews; bank transactions; AR/AP schedules; probabilistic forecasting

### A late receipt changes the cash trough

Week | Expected receipt | Timing review
--- | --- | ---
Week 1 | $180k | $180k
Week 2 | $140k | $0
Week 4 | $90k | $230k

Week | Expected receipt | Timing review
--- | --- | ---
Week 1 | $180k | $180k
Week 2 | $140k | $0
Week 4 | $90k | $230k

Review interpretation Review a timing scenario and funding options with treasury; a lower trough is a risk signal.

Review a timing scenario and funding options with treasury; a lower trough is a risk signal.

### AI use cases: how could the work change?

Two visual examples · inputs, interpretation and human decisions

An illustrative AI application: inspect the source records, the interpretation and the human decision.

AI / records & interpretation

#### Cash forecasting: evidence assist

Receipt forecast: a disputed account moves from week 2 to week 4; show uncertainty.

Inputs: Treasury interviews; bank transactions; AR/AP schedules; probabilistic forecasting.

Human action: Review a timing scenario and funding options with treasury; a lower trough is a risk signal.

AI / analysis & workflow

#### Forecast backtesting

Historical forecasts and realized bank flows

Validate: Compare the draft interpretation with a verified sample and record missing or incorrect conclusions before changing the process.

The proposed change fails the measured outcome, shifts work elsewhere, loses required controls or costs more than the verified benefit. Investigate those results before scaling.

Finance / investigate the opportunity

## Financial close

The working hypothesis

Reconciliations may stay open because evidence and accountable owners are dispersed.

What would establish or reject the case?

Trace close tasks and reconciling items; pilot evidence bundles and check completeness, review quality and audit trail.

### Measure the consequence

Close duration; aged reconciling items; reviewer rework

Controller interviews; task logs; ledgers; supporting schedules; document retrieval

### A reconciliation waits across three handoffs

Reconciliation | Owner | Dependency
--- | --- | ---
Cash | Treasury | Bank statement linked
Inventory | Plant finance | Count evidence pending
Accrual | Controller | Receipt confirmation

Reconciliation | Owner | Dependency
--- | --- | ---
Cash | Treasury | Bank statement linked
Inventory | Plant finance | Count evidence pending
Accrual | Controller | Receipt confirmation

Review interpretation Assemble source-linked evidence and draft explanations; the controller signs off.

Assemble source-linked evidence and draft explanations; the controller signs off.

### AI use cases: how could the work change?

Two visual examples · inputs, interpretation and human decisions

An illustrative AI application: inspect the source records, the interpretation and the human decision.

AI / records & interpretation

#### Financial close: evidence assist

Reconciliation assist: two offsetting entries share an order reference; propose a match.

Inputs: Controller interviews; task logs; ledgers; supporting schedules; document retrieval.

Human action: Assemble source-linked evidence and draft explanations; the controller signs off.

AI / analysis & workflow

#### Close commentary draft

Ledger movements and reconciled schedules

Validate: Compare the draft interpretation with a verified sample and record missing or incorrect conclusions before changing the process.

The proposed change fails the measured outcome, shifts work elsewhere, loses required controls or costs more than the verified benefit. Investigate those results before scaling.

Finance / investigate the opportunity

## Spend & contract leakage

The working hypothesis

Paid rates may diverge from negotiated terms when agreements are difficult to reconcile with invoices.

What would establish or reject the case?

Match rate cards and invoices; validate units, effective dates and exceptions with category owners.

### Measure the consequence

Verified overcharge; recovery received; recurrence rate

Procurement interviews; contracts; invoices; clause extraction; normalized units

### The invoice uses the superseded rate

Contract / bill | Rate | Status
--- | --- | ---
Rate card v2 | $100 | Effective 01 October
Invoice 883 | $112 | October service
Difference | $12 | Check exclusions

Contract / bill | Rate | Status
--- | --- | ---
Rate card v2 | $100 | Effective 01 October
Invoice 883 | $112 | October service
Difference | $12 | Check exclusions

Review interpretation Confirm the effective rate and exclusions before disputing the charge.

Confirm the effective rate and exclusions before disputing the charge.

### AI use cases: how could the work change?

Two visual examples · inputs, interpretation and human decisions

An illustrative AI application: inspect the source records, the interpretation and the human decision.

AI / records & interpretation

#### Spend & contract leakage: evidence assist

Clause comparison: renewal date changed the rate; expose the relevant page and calculation.

Inputs: Procurement interviews; contracts; invoices; clause extraction; normalized units.

Human action: Confirm the effective rate and exclusions before disputing the charge.

AI / analysis & workflow

#### Recovery case preparation

Confirmed rate differences and payment history

Validate: Compare the draft interpretation with a verified sample and record missing or incorrect conclusions before changing the process.

The proposed change fails the measured outcome, shifts work elsewhere, loses required controls or costs more than the verified benefit. Investigate those results before scaling.

Finance / investigate the opportunity

## Cost to serve

The working hypothesis

Revenue reports may hide customers whose handling, delivery and support requirements consume margin.

What would establish or reject the case?

Build an activity-based cost sample; reconcile to the ledger and test service-level alternatives with account owners.

### Measure the consequence

Contribution after service cost; service level; cost per order

Finance and service interviews; order lines; time samples; freight; cost allocation

### Two equal orders require different work

Order | Order value | Handling
--- | --- | ---
Order 410 | $500 revenue | 2 picks / $14
Order 411 | $500 revenue | Split picks / $26
Difference | Same revenue | $12 extra handling

Order | Order value | Handling
--- | --- | ---
Order 410 | $500 revenue | 2 picks / $14
Order 411 | $500 revenue | Split picks / $26
Difference | Same revenue | $12 extra handling

Review interpretation Review order batching or service terms while monitoring retention and service quality.

Review order batching or service terms while monitoring retention and service quality.

### AI use cases: how could the work change?

Two visual examples · inputs, interpretation and human decisions

An illustrative AI application: inspect the source records, the interpretation and the human decision.

AI / records & interpretation

#### Cost to serve: evidence assist

Cost driver model: small split orders attract extra touches; show allocation sensitivity.

Inputs: Finance and service interviews; order lines; time samples; freight; cost allocation.

Human action: Review order batching or service terms while monitoring retention and service quality.

AI / analysis & workflow

#### Cost allocation sensitivity

Activity samples and alternative cost drivers

Validate: Compare the draft interpretation with a verified sample and record missing or incorrect conclusions before changing the process.

The proposed change fails the measured outcome, shifts work elsewhere, loses required controls or costs more than the verified benefit. Investigate those results before scaling.

Finance / investigate the opportunity

## Capital prioritization

The working hypothesis

Projects may be ranked on inconsistent assumptions about demand, uptime and implementation cost.

What would establish or reject the case?

Rebuild comparable cash flows; stress demand and downtime; independently validate costs and avoid double counting benefits.

### Measure the consequence

NPV sensitivity; cash requirement; delivery milestones

Investment owner interviews; project estimates; operating data; scenario modelling

### A low-cost project has a costly interruption

Assumption | Base case | Downside
--- | --- | ---
Installation stop | 1 day | 3 days
Ramp-up | 2 weeks | 5 weeks
Project NPV | $90k | $54k

Assumption | Base case | Downside
--- | --- | ---
Installation stop | 1 day | 3 days
Ramp-up | 2 weeks | 5 weeks
Project NPV | $90k | $54k

Review interpretation Use downside evidence in the investment review; simulated NPV is not an approval.

Use downside evidence in the investment review; simulated NPV is not an approval.

### AI use cases: how could the work change?

Two visual examples · inputs, interpretation and human decisions

An illustrative AI application: inspect the source records, the interpretation and the human decision.

AI / records & interpretation

#### Capital prioritization: evidence assist

Assumption review: check that volume growth and utilization do not count the same benefit twice.

Inputs: Investment owner interviews; project estimates; operating data; scenario modelling.

Human action: Use downside evidence in the investment review; simulated NPV is not an approval.

AI / analysis & workflow

#### Investment assumption checks

Project cash flows and operating assumptions

Validate: Compare the draft interpretation with a verified sample and record missing or incorrect conclusions before changing the process.

The proposed change fails the measured outcome, shifts work elsewhere, loses required controls or costs more than the verified benefit. Investigate those results before scaling.

Finance / investigate the opportunity

## Benefits realization

The working hypothesis

Time released by automation may be reported as savings before expenditure or throughput actually changes.

What would establish or reject the case?

Lock the baseline and classify cash release, recurring cost removal and redeployed capacity separately; reconcile with finance.

### Measure the consequence

Realized net savings; capacity used; cash released

Benefit owner interviews; GL; workload measures; implementation and recurring costs

### A gross opportunity becomes a net business case

Benefit bridge | Annual value | Evidence
--- | --- | ---
Gross opportunity | $100k | Owner estimate
Recurring operating cost | −$38k | Cost estimate
Net before one-time cost | $62k | Finance review pending

Benefit bridge | Annual value | Evidence
--- | --- | ---
Gross opportunity | $100k | Owner estimate
Recurring operating cost | −$38k | Cost estimate
Net before one-time cost | $62k | Finance review pending

Review interpretation Record 100k gross less 38k running cost; subtract one-time implementation cost separately.

Record 100k gross less 38k running cost; subtract one-time implementation cost separately.

### AI use cases: how could the work change?

Two visual examples · inputs, interpretation and human decisions

An illustrative AI application: inspect the source records, the interpretation and the human decision.

AI / records & interpretation

#### Benefits realization: evidence assist

Evidence check: match a cancelled contract to the ledger rather than counting a forecast twice.

Inputs: Benefit owner interviews; GL; workload measures; implementation and recurring costs.

Human action: Record 100k gross less 38k running cost; subtract one-time implementation cost separately.

AI / analysis & workflow

#### Benefit evidence reconciliation

Owner estimates, contracts and ledger entries

Validate: Compare the draft interpretation with a verified sample and record missing or incorrect conclusions before changing the process.

The proposed change fails the measured outcome, shifts work elsewhere, loses required controls or costs more than the verified benefit. Investigate those results before scaling.

Questions & answers

## Before testing an opportunity

### Where can AI help in finance?

Potential areas include invoice exceptions, collections & disputes, cash forecasting, financial close, spend & contract leakage, cost to serve, capital prioritization, benefits realization. Each is a hypothesis to test against operating evidence.

### Are these savings or measured results?

No. All figures, records and overlays are illustrative scenarios. Validate the baseline, net cost, feasibility and service outcomes before making a business case.

### Which enablers are needed?

Start with process interviews, reliable source records and a measurable baseline. Depending on the task, useful enablers include document extraction, retrieval, classification, forecasting and workflow integration. Assign a person to review exceptions.

### What should a first pilot establish?

A pilot should establish whether a specific intervention changes the measured outcome against a credible baseline, how often it fails, who reviews exceptions and what it costs to operate.

APQC benchmark source: https://www.apqc.org/What-Is-the-Average-Cost-to-Process-an-Invoice (October 2026).

Interactive value calculator: tasks × minutes ÷ 60 × realized capacity × loaded hourly cost − running cost. Default inputs: 20,000 tasks; 5 minutes; 60% realized; $45/hour; $15,000 annual running cost; $30,000 implementation. Result: 1,000 hours; $30,000 annual capacity value; $0 first year. This is not automatically removed expenditure.