Forward deployed transformation hypotheses / IT

# IT: 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

## IT: 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.

### Technology strategy and demand

Process stages: Business need → demand → portfolio → funding → roadmap

Duplicate requests may fund overlapping capabilities

Poor demand qualification may initiate low-value work

Unclear ownership may allow obsolete services to persist

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

Map business capabilities, demand evidence, service owners and total lifecycle cost. Reject consolidation where separation serves a real requirement.

#### Where AI could contribute

Cluster demand and link business evidence to portfolio scenarios; owners set priorities.

Example sizing model / Avoided duplicate delivery cost

Verified redundant funded initiatives avoided × avoidable delivery cost per initiative

Low / base / high illustration: 1 / 2 / 3 initiatives × $100,000 = $100k / $200k / $300k one-time delivery-cost opportunity.

Confirm business capability equivalence, commitments, cancellation and migration cost. Do not count a deferred project as expenditure removed or add this to the same project benefit twice.

### Product and software delivery

Process stages: Discover → design → build → test → release

Approval queues may dominate delivery lead time

Unstable requirements may drive repeated rework

Missing test context may increase escaped defects

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

Trace work-item waiting, rework and release outcomes; compare equivalent changes. Reject faster throughput that increases change failures.

#### Where AI could contribute

Assist requirement clarification, code and test preparation with review; teams retain release control.

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.

### Service desk and knowledge

Process stages: Request → diagnose → fulfill → resolve → learn

Missing request context may cause handoffs

Repeated questions may signal inaccessible runbooks

Incorrect guidance may create reopened tickets

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

Sample ticket journeys, verified answers and reopens by issue type. Reject containment that hides unresolved user problems.

#### Where AI could contribute

Source-linked retrieval, classification and approved diagnostic collection; technicians handle unsupported cases.

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.

### Reliability and service operations

Process stages: Observe → detect → diagnose → restore → prevent

Disconnected telemetry may delay diagnosis

Repeated incidents may share unresolved causes

Alert noise may distract responders from useful signals

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

Replay incidents with time-aligned logs, changes and dependencies. Reject correlations not supported by responder-confirmed causes.

#### Where AI could contribute

Rank candidate causes and assemble incident timelines; responders approve remediation.

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.

### Cloud, infrastructure and network

Process stages: Provision → operate → optimize → retire

Idle workloads may create avoidable spend

Oversizing may persist because peak requirements are unknown

Unused circuits may remain billed after site changes

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

Join billing, tags, utilization, peak SLOs and failover requirements. Reject changes that compromise performance, resilience or exit obligations.

#### Where AI could contribute

Usage analysis and reversible rightsizing scenarios; service owners approve resource changes.

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.

### Applications, licenses and vendors

Process stages: Inventory → entitlement → usage → renew → exit

Inactive seats may remain in renewal counts

Overlapping tools may duplicate support costs

Invoices may exceed contracted usage or rates

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

Reconcile contracts, usage, seasonal need and dependencies; validate retirement cost. Reject seats already covered by unavoidable minimum commitments.

#### Where AI could contribute

Entity matching, entitlement reconciliation and contract exception review.

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.

### Security, access and control operations

Process stages: Identify → protect → detect → respond → evidence

Disconnected identity events may delay access removal

Manual evidence gathering may duplicate effort

Unprioritized findings may consume effort without reducing exposure

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

Reconcile approvals, identity changes, control evidence and risk context; reject savings that weaken control outcomes or response coverage.

#### Where AI could contribute

Assemble evidence and prioritize review queues; owners authorize access and response actions.

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.

### Data and integration lifecycle

Process stages: Acquire → integrate → govern → consume → archive

Repeated transformations may duplicate engineering work

Poor lineage may cause reconciliation effort

Unused data copies may increase storage and governance cost

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

Map consumers, pipelines, quality incidents and retention obligations. Reject retirement if downstream uses or required records remain.

#### Where AI could contribute

Discover lineage, draft quality rules and explain discrepancies with source references.

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.

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.

IT / investigate the opportunity

## Application portfolio

The working hypothesis

Overlapping applications may create avoidable license and support cost, but apparent duplicates can hide critical dependencies.

What would establish or reject the case?

Map capabilities, users, integrations and retention needs; test a retirement candidate with its business owner.

### Measure the consequence

Retired spend; service continuity; migration cost

Business-owner interviews; application inventory; usage; dependency mapping

### Two tools support the same reporting workflow

Application | Capability | Dependency
--- | --- | ---
Tool A | Reporting | Finance exports
Tool B | Reporting | Plant interface
Tool C | Reporting | No owner confirmed

Application | Capability | Dependency
--- | --- | ---
Tool A | Reporting | Finance exports
Tool B | Reporting | Plant interface
Tool C | Reporting | No owner confirmed

Review interpretation Review dependencies and transition cost before approving retirement.

Review dependencies and transition cost before approving retirement.

### 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

#### Application portfolio: evidence assist

Capability map: cluster overlapping features and link each claim to user evidence.

Inputs: Business-owner interviews; application inventory; usage; dependency mapping.

Human action: Review dependencies and transition cost before approving retirement.

AI / analysis & workflow

#### Retirement impact analysis

Verified application dependencies and business usage

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.

IT / investigate the opportunity

## Cloud utilization

The working hypothesis

Idle and oversized workloads may explain part of cloud spend without contributing to service reliability.

What would establish or reject the case?

Replay utilization and service levels; test a reversible rightsizing change under representative peak demand.

### Measure the consequence

Cost per transaction; latency; error rate

Service-owner interviews; billing; telemetry; capacity forecasting

### A test environment runs all weekend

Workload | Runtime | Service context
--- | --- | ---
test-env-01 | 168 h/week | 60 h scheduled use
production-api | 168 h/week | Customer service
backup-job | 4 h/week | Recovery dependency

Workload | Runtime | Service context
--- | --- | ---
test-env-01 | 168 h/week | 60 h scheduled use
production-api | 168 h/week | Customer service
backup-job | 4 h/week | Recovery dependency

Review interpretation Schedule the test environment with an owner-approved exception path; protect production.

Schedule the test environment with an owner-approved exception path; protect production.

### 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

#### Cloud utilization: evidence assist

Usage anomaly: runtime continues after test completion; compare owner tags and planned jobs.

Inputs: Service-owner interviews; billing; telemetry; capacity forecasting.

Human action: Schedule the test environment with an owner-approved exception path; protect production.

AI / analysis & workflow

#### Cost anomaly explanation

Billing line items, tags and deployment events

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.

IT / investigate the opportunity

## Software licenses

The working hypothesis

Purchased seats may exceed genuine demand when joiner, mover and leaver events are disconnected from renewals.

What would establish or reject the case?

Reconcile entitlement and actual usage; confirm seasonal users and contract minimums before reclaiming seats.

### Measure the consequence

Paid active seats; verified reductions; access disruption

IT asset interviews; identity; usage; contracts; entity matching

### A renewal includes seats with no recent activity

Entitlement | Seats | Evidence
--- | --- | ---
Purchased | 250 | Renewal draft
Inactive candidates | 45 | Owner check required
Seasonal roles | 12 | Do not remove yet

Entitlement | Seats | Evidence
--- | --- | ---
Purchased | 250 | Renewal draft
Inactive candidates | 45 | Owner check required
Seasonal roles | 12 | Do not remove yet

Review interpretation Ask owners to confirm need; observed inactivity alone does not authorize removal.

Ask owners to confirm need; observed inactivity alone does not authorize removal.

### 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

#### Software licenses: evidence assist

Renewal review: identify 45 candidates, exclude 12 seasonal roles pending confirmation.

Inputs: IT asset interviews; identity; usage; contracts; entity matching.

Human action: Ask owners to confirm need; observed inactivity alone does not authorize removal.

AI / analysis & workflow

#### Contract renewal extraction

License terms and verified seat demand

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.

IT / investigate the opportunity

## Service desk & knowledge

The working hypothesis

Repeated tickets may arise from hard-to-find guidance and missing context in initial requests.

What would establish or reject the case?

Group historical cases; pilot source-linked guidance and measure correctness, repeat contacts and escalation.

### Measure the consequence

First-contact resolution; reopened cases; wait time

Support interviews; ticket logs; knowledge articles; retrieval and triage

### A familiar error arrives without diagnostic context

Ticket | Context | Next step
--- | --- | ---
IT-104 | Access denied | Role approval required
IT-105 | Known error code | Link current runbook
IT-106 | Unrecognized fault | Technician review

Ticket | Context | Next step
--- | --- | ---
IT-104 | Access denied | Role approval required
IT-105 | Known error code | Link current runbook
IT-106 | Unrecognized fault | Technician review

Review interpretation Gather approved diagnostic context and show documented steps; privileged actions require approval.

Gather approved diagnostic context and show documented steps; privileged actions require 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

#### Service desk & knowledge: evidence assist

Answer assist: cite the current runbook and route unsupported errors to a technician.

Inputs: Support interviews; ticket logs; knowledge articles; retrieval and triage.

Human action: Gather approved diagnostic context and show documented steps; privileged actions require approval.

AI / analysis & workflow

#### Knowledge gap detection

Resolved tickets and the current runbook library

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.

IT / investigate the opportunity

## Incident diagnosis

The working hypothesis

Telemetry may be abundant while the service dependency that explains an incident remains unclear.

What would establish or reject the case?

Replay known incidents with time-aligned evidence; test candidate ranking and missed causes with responders.

### Measure the consequence

Time to isolate; false leads; recurrence

Responder interviews; traces; logs; dependency graph; anomaly detection

### Three alerts share one upstream change

Time | Event | Relationship
--- | --- | ---
09:10 | Configuration change | Candidate cause
09:12 | Database latency | Upstream signal
09:14 | Application errors | Downstream impact

Time | Event | Relationship
--- | --- | ---
09:10 | Configuration change | Candidate cause
09:12 | Database latency | Upstream signal
09:14 | Application errors | Downstream impact

Review interpretation Present a source-linked timeline; the responder validates cause and rollback.

Present a source-linked timeline; the responder validates cause and rollback.

### 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

#### Incident diagnosis: evidence assist

Evidence graph: database latency starts after a configuration change and precedes application errors.

Inputs: Responder interviews; traces; logs; dependency graph; anomaly detection.

Human action: Present a source-linked timeline; the responder validates cause and rollback.

AI / analysis & workflow

#### Post-incident evidence synthesis

Validated event timeline and responder 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.

IT / investigate the opportunity

## Delivery flow & rework

The working hypothesis

Slow delivery may reflect waiting and unstable requirements rather than insufficient coding capacity.

What would establish or reject the case?

Map work from request to release; distinguish queue time from active work and pilot a smaller release batch.

### Measure the consequence

Lead time; escaped defects; change failure

Team interviews; work records; CI events; test evidence; process mining

### A two-day change waits ten days in queues

Stage | Active time | Queue
--- | --- | ---
Implementation | 2 days | 1 day
Review | 1 day | 5 days
Release | 1 day | 2 days

Stage | Active time | Queue
--- | --- | ---
Implementation | 2 days | 1 day
Review | 1 day | 5 days
Release | 1 day | 2 days

Review interpretation Reduce a verified queue while checking test coverage and operational readiness.

Reduce a verified queue while checking test coverage and operational readiness.

### 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

#### Delivery flow & rework: evidence assist

Change review: group recurring rework causes and trace each theme to an observed issue.

Inputs: Team interviews; work records; CI events; test evidence; process mining.

Human action: Reduce a verified queue while checking test coverage and operational readiness.

AI / analysis & workflow

#### Requirements ambiguity review

Approved change request and acceptance criteria

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.

IT / investigate the opportunity

## Vendor & network spend

The working hypothesis

Unused circuits and inconsistent service rates may persist because invoices, sites and ownership records do not reconcile.

What would establish or reject the case?

Match invoices to active services and sites; confirm failover and contractual obligations before any cancellation.

### Measure the consequence

Verified unused cost; availability; termination charges

Network-owner interviews; invoices; circuits; contracts; topology

### A closed office still has a billed connection

Circuit | Site | Check
--- | --- | ---
C-031 | Closed office | Termination terms
C-032 | Main warehouse | Primary connection
C-033 | Main warehouse | Low-use failover

Circuit | Site | Check
--- | --- | ---
C-031 | Closed office | Termination terms
C-032 | Main warehouse | Primary connection
C-033 | Main warehouse | Low-use failover

Review interpretation Confirm redundancy and exit cost; a low-traffic backup may still be essential.

Confirm redundancy and exit cost; a low-traffic backup may still be essential.

### 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

#### Vendor & network spend: evidence assist

Invoice audit: circuit 031 maps to a closed site; request owner verification.

Inputs: Network-owner interviews; invoices; circuits; contracts; topology.

Human action: Confirm redundancy and exit cost; a low-traffic backup may still be essential.

AI / analysis & workflow

#### Contract-rate reconciliation

Circuit inventory, invoice lines and rate cards

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.

IT / investigate the opportunity

## Data & control evidence

The working hypothesis

Repeated manual evidence gathering may consume effort because records are fragmented and controls lack clear ownership.

What would establish or reject the case?

Select a control sample; check completeness, freshness and reproducibility against the current evidence process.

### Measure the consequence

Evidence preparation time; missing evidence; reviewer findings

Control-owner interviews; access logs; approvals; document lineage

### One access review needs four separate exports

Control evidence | Source | Status
--- | --- | ---
Access list | Identity log | Current
Owner approval | Review record | One missing
Exception closure | Case record | Reviewer needed

Control evidence | Source | Status
--- | --- | ---
Access list | Identity log | Current
Owner approval | Review record | One missing
Exception closure | Case record | Reviewer needed

Review interpretation Assemble traceable records for reviewer sign-off; summaries do not establish control effectiveness.

Assemble traceable records for reviewer sign-off; summaries do not establish control effectiveness.

### 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

#### Data & control evidence: evidence assist

Evidence draft: show source date, owner and unresolved exception next to each control.

Inputs: Control-owner interviews; access logs; approvals; document lineage.

Human action: Assemble traceable records for reviewer sign-off; summaries do not establish control effectiveness.

AI / analysis & workflow

#### Evidence completeness checking

Control requirements and collected source records

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 it?

Potential areas include application portfolio, cloud utilization, software licenses, service desk & knowledge, incident diagnosis, delivery flow & rework, vendor & network spend, data & control evidence. 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.

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.