Revenue Trace · ROI for AI

01Opportunity · overview

Where additional revenue might start

Which demand already reaching us could become additional profitable work?Illustrative contractor case · no figures are given or implied

The working floor · where a request may wait

IntakeOffice coordinatorUnowned
RoutingCoordinator to estimatorWrong person
CallbackEstimatorNo reply
ScopeEstimator and property managerAwaiting scope
Crew scheduleSchedulerWaiting for crew

Each tray marks a location and a reason. Every tray is drawn the same size: it shows where a request may wait, never how many. Each one is a hypothesis to confirm against existing records.

Stalls have locations and owners

A request can wait at intake, with the wrong estimator, after a promised callback or for scope. Naming where it waits tells you which person and handoff an investment would actually need to change.

A stall is a hypothesis

An unanswered request is not automatically a lost sale. The buyer may have chosen elsewhere, paused the project or never intended to proceed. Confirm that relevant demand actually stalls before funding a fix.

Test demand already reaching the business

Where relevant inquiries already arrive, tracing them can expose a specific investment opportunity. Check whether ownership, response or follow-through limits commercial progress, then compare that opportunity with other constraints before funding it.

Decision for your leadership teamWhich request type do we suspect stalls, and which existing record would confirm or refute it?

01Opportunity · a closer look

Choose a new-work inquiry

Why does the trace follow a new-work estimate rather than a routine service request?Illustrative contractor case · no figures are given or implied

Request A · new work

A property manager asks for an estimate on a commercial renovation.

Owner
Scope
Price
Buyer decision

A new buyer decision: an opportunity to trace

Request B · owed service

A facilities team requests work already covered by its service agreement.

Agreement already won
Deliver the owed work

Revenue rests on a contract already won: not a new sale

Eligible requests, written down before any testChannelTradeTerritoryCustomer type

Only request A leads to a new buyer decision. Faster owed service may still improve satisfaction and retention; it is simply not new revenue to credit.

Owed service rests on a contract already won

Handling owed service faster may improve satisfaction and retention, but its revenue rests on an agreement already won. Counting it as new revenue would credit AI for a sale that already existed.

New work needs a buyer decision

A new estimate request has a clear commercial path: an owner, a scope, a price and a customer who can accept or decline. That decision point is what makes additional revenue traceable.

Define the eligible requests first

Before any test, write down which requests are in scope: channel, trade, territory or customer type. Without that boundary, assisted and comparison groups cannot be matched and results cannot be read fairly.

Decision for your leadership teamWhich new-work request type is distinct enough to define and trace on its own?

02AI & People · overview

What AI prepares and who decides

Where does AI help, and where must a person own the action and the record?Illustrative contractor case · no figures are given or implied

AIMatchConfirm matchOffice coordinator confirms the property or marks it newSystem recordProperty or account link, or a new record
AIBrief · suggestAccept ownershipCoordinator routes; the estimator accepts or reassignsSystem recordAssignment and acceptance of ownership
AIDraft questionsSet scope and priceEstimator decides; a manager approves where policy requiresSystem recordConfirmed scope; estimate version sent
AIDraft · remindApprove messageEstimator decides whether and when to contact the customerSystem recordWhat people actually sent
AIFlag conflictsCommit scheduleScheduler and operations lead commit crew, materials and datesSystem recordJob scheduled; materials and subcontracts ordered

Teal: AI preparation. Ochre: a human commitment. Dashed: an uncertain match flagged for review instead of merged. AI sends no price, promises no date and repeats no follow-up on its own.

Preparing is not deciding

AI can assemble property history, draft an intake brief and suggest the estimator by trade and territory. The coordinator confirms the route, and the estimator accepts or reassigns ownership. The commitment stays human.

Uncertain matches stay unconfirmed

When AI cannot confidently link a request to an existing property or account, it flags the match for review instead of merging records. A wrong link can misroute the request and corrupt the trace.

No automatic commitments or endless contact

Drafts and reminders go to staff, who decide whether and when to contact the customer. AI does not send prices, promise dates or repeat follow-ups on its own; the communication system records what people sent.

Decision for your leadership teamFor our chosen request type, which steps may AI prepare, and which person approves each commitment?

03Sale & Delivery · overview

From prepared quote to earned revenue

What separate events must happen after the AI action before any revenue is earned?Illustrative contractor case · no figures are given or implied

One request's road · each marker a separate event

Qualified opportunityEstimator confirms scopeRecord · Contact, site visit and confirmed scopeNot a booking
Prepared quoteEstimator sets price; a manager approves where requiredRecord · Estimate version sentNo revenue, no buyer commitment
Accepted bookingProperty manager acceptsRecord · Signed contract or purchase orderContract value, not recognized revenue
Scheduled jobScheduler and operations lead commitRecord · Job scheduled; materials and subcontracts orderedDelivery costs begin
Delivered work, recognized revenueCrew lead delivers; finance recognizesRecord · Progress or completion, revenue entries, job costsUnder contract terms, possibly over time
Collected cashFinance clerk invoices and collectsRecord · Invoice issued; payment receivedTiming only, no second benefit

Different shapes, different events, recorded by different people at different times. The road runs through people and a customer decision, not through the software event.

Each stage is a different event

A qualified opportunity, a sent quote, an accepted booking and recognized revenue are recorded by different people at different times. Treating an earlier stage as a later one overstates what the change achieved.

The sale runs through people

AI preparation only matters if the estimator acts on it and the property manager accepts. The commercial path runs through human follow-through and a customer decision, not through the software event.

Revenue follows delivery and contract terms

Revenue is recognized as the contract and applicable accounting terms allow, which for longer jobs may happen progressively over time. A booking alone is contract value, not earned revenue.

Decision for your leadership teamCan we link each request in scope to its quote, acceptance, delivered work and recognized revenue records?

03Sale & Delivery · a closer look

The customer's decision

What turns a prepared quote into an accepted booking, and what does a decline or silence mean?Illustrative contractor case · no figures are given or implied

Estimate sentThe estimator owns scope and price; AI may flag gaps but does not set or change a price

AcceptedRecord: signed contract or purchase order. A booking, not yet revenue.
DeclinedRecord: the recorded decline.
DeferredRecord: the recorded deferral.
No responseRecorded as no response. Not automatically a lost sale, or a won one.

Four distinct outcomes, each with its own record; none is drawn as more likely. A sales manager reviews open quotes on a set rhythm and decides which to pursue or close.

The estimator owns scope and price

The estimator decides whether to visit, confirms scope and sets price; approval follows company policy. AI may flag missing scope items or draft the narrative, but it does not set or change a price.

Follow-through is a human rhythm

Reminders help the estimator follow up while the buyer is still deciding. A sales manager reviews open quotes on a set rhythm and decides which to pursue or close, rather than letting software keep contacting.

Record every outcome honestly

Accepted, declined, deferred and no response are distinct outcomes. Recording each keeps the trace complete and stops a silent request from being counted as either a lost sale or a won one.

Decision for your leadership teamWho reviews open quotes, on what rhythm, and where is each customer decision recorded?

03Sale & Delivery · a closer look

Can delivery absorb more work?

What must be true downstream for an additional booking to become earned revenue at a viable margin?Illustrative contractor case · no figures are given or implied

Accepted bookings arrive

Scheduling trayFills as bookings wait for delivery room

CrewsCommitted by the scheduler and operations lead
SubcontractorsAvailability for the trade in scope
MaterialsLead-time notes flagged for people to review
ScheduleDates committed by people, not by AI

If delivery is fullWork waits · quality slips · overtime eats margin

No capacity level or count is shown: the board names possible constraints. Freed salaried estimator or office time is capacity, not a wage saving.

Front-end gains can move the bottleneck

Faster estimating can expose limits in crews, subcontractors, materials or scheduling. If those are full, additional bookings wait, quality slips or overtime eats margin, and the revenue case weakens.

Check room before funding the front end

Before investing in intake and estimating, confirm delivery has room for the request type in scope, or plan the capacity change and include its cost in the same evaluation period.

Capacity is not removed cost

Freed estimator or office time on salary is capacity, not a wage saving. It earns only when demand, adoption and delivery turn it into additional completed, recognized work.

Decision for your leadership teamIf more of this work is accepted, which crew, material or schedule limit would we reach first?

04Return · overview

What economic benefit remains

After added delivery costs and full change costs, what benefit, if any, is left for the period?Illustrative contractor case · no figures are given or implied

One chosen investment period · unscaled

Additional earned revenueRevenue recognized within the chosen investment period, under contract and accounting terms, from work that a fair comparison suggests would not otherwise have been won and delivered. Quotes and bookings are excluded.
minusAdded delivery costsIncremental crew, material, subcontract and overtime cost of delivering that same additional work in the same period. Subtracting it gives contribution.
= contribution for this work
minusFull AI/change costsCount setup, integration/data, training/adoption, management/internal effort, licenses/usage/infrastructure, review/exceptions/corrections and upkeep for the chosen period. Include required capacity changes only where not already counted in delivery cost; keep unknowns visible and count each cost once.
equalsRemaining economic benefitWhat remains in the chosen investment period after both cost categories, without assuming a positive result. Read it alongside evidence quality, cash requirements, delivery readiness and the investment horizon when making the funding decision.PositiveZeroNegative

A test can cost more than it earns. Show the cost sequence without amounts or proportions, and leave its outcome open to a positive, zero or negative result.

Start from additional earned revenue

Only revenue from work that would not otherwise have been won and delivered belongs in the bridge, recognized under the contract's terms within the chosen investment period. Quotes and bookings stay outside it.

Subtract delivery, then change costs

Subtracting added crew, material and subcontract costs gives contribution. Full AI and change costs come next. Count each cost once: never repeat a delivery cost as a change cost, or one internal hour twice.

The result can be negative

A bounded test can cost more than it earns, especially while setup and training are underway. The bridge shows what remains; it does not assume a positive bottom line, a target or a return rate.

Decision for your leadership teamHave we fixed one investment period and listed every benefit and cost inside it, marking unknowns as unknown?

04Return · a closer look

The full cost of the change

Which costs must be counted against the benefit for the same investment period?Illustrative contractor case · no figures are given or implied

Cost ledger for one chosen investment period · rows are not sized
Cost row	What it covers	Status
Setup, integration and data	Configuration, connections and data cleanup	Estimate, or mark unknown
Training and adoption	Teaching staff and changing routines	Estimate, or mark unknown
Management and internal effort	The manager's supervision and staff time spent on the change	Estimate, or mark unknown
Licenses, usage and infrastructure	Subscription and usage fees: one row among several	Estimate, or mark unknown
Human review, exceptions and corrections	Checking matches, correcting drafts, handling misroutes; counted once here, not again as delivery cost	Estimate, or mark unknown
Upkeep	Keeping the change working through the period	Estimate, or mark unknown
Downstream capacity, where required	Only where not already counted in delivery cost	Estimate, or mark unknown

Every row belongs in the same investment period as the benefit. Any row may still read unknown; leaving it visible is better than quietly omitting it.

Subscriptions are one row among several

License and usage fees are one line in the ledger. Setup, integration, data cleanup, training and the manager's supervision time are real costs even when no new invoice arrives for them.

Review and exceptions cost people time

Staff checking AI matches, correcting drafts, handling misroutes and resolving exceptions spend internal effort. That effort belongs in the change cost, counted once and not again as delivery cost.

Same period, unknowns visible

Every cost, including upkeep and any downstream capacity added to deliver the work, belongs in the same chosen investment period as the benefit. Mark uncertain items unknown rather than quietly omitting them.

Decision for your leadership teamWho owns the cost list, and which rows are still unknown for our proposed change?

04Return · a closer look

Cash timing is separate

How does earlier invoicing or collection relate to revenue and profit?Illustrative contractor case · no figures are given or implied

Time runs left to right · spacing is not to scale

EarnedFeeds the bridge

Work delivered
Revenue recognized under contract terms, possibly progressively
Job costs recorded against the work

CashIts own view

Crew wages paid
Materials and subcontractors paid
Invoice issued under contract terms
Payment arrives later

Earlier payment changes when cash arrives. It is never added to the bridge as a second revenue or profit benefit.

Profitable work can still strain cash when crews and materials are paid before the customer pays. Finance keeps recognized revenue, job costs and collections as separate records.

Faster invoicing changes when, not how much

Invoicing sooner or collecting faster can improve cash on hand and ease pressure. It does not create additional revenue or profit, so it never appears as a second benefit in the bridge.

Profitable work can still strain cash

Crew wages, materials and subcontractors are often paid before the customer pays. A job with healthy contribution can still create cash pressure, so cash timing deserves its own view.

Keep the finance records apart

Finance tracks recognized revenue, job costs and collections as distinct records. Keeping them apart lets leaders read margin and cash correctly instead of mixing a timing gain into the return.

Decision for your leadership teamIs the proposed change meant to add revenue, improve cash timing, or both, and are the two kept apart?

05Proof · overview

What justifies crediting the change

What evidence shows the gain was additional and attributable to this change?Illustrative contractor case · no figures are given or implied

Observed

Won jobAI used

Proves use. The customer may have bought anyway.

Tested

AssistedQuoteDecisionDeliveredCostsComparisonQuoteDecisionDeliveredCosts

Demand
Pricing
Season
Adoption
Would buy anyway

A difference between comparable groups, read only after confounds are checked.

Records linked end to end

Eligible request
Handoffs
Quote
Customer decision
Delivered work
Delivery costs

A gap anywhere in the chain leaves the claim open. When several AI changes work on one request, their combined gain is judged once against their combined cost.

Use is not cause

AI appearing on a winning job proves only that it was used. The customer may have bought anyway. An incremental claim needs evidence of what would likely have happened without the change.

Link the records end to end

Attribution depends on linking each eligible request to its handoffs, quote, customer decision, delivered work and delivery costs. Gaps in that chain leave the claim open, however good the totals look.

Count a shared gain once

When several AI changes work together on the same request, evaluate their combined benefit against their combined cost. Crediting the full sale to each tool counts one gain several times.

Decision for your leadership teamWhat evidence, agreed before the test, would we accept to call a gain additional?

05Proof · a closer look

Building a fair comparison

How can a comparison or staged trial support, without guaranteeing, an incremental claim?Illustrative contractor case · no figures are given or implied

Demand shiftsPricing changesSeason

Eligible requests only · the same period

AssistedQuoteDecisionDeliveredCostsComparisonQuoteDecisionDeliveredCosts

Which requests or staff were selectedAdoptionCustomers who would buy anyway

Before and afterObservation, not a comparison: season, pricing, a busier market or work brought forward can explain a rise.

A comparable territory, team or staged rollout gives a reference for what would otherwise happen. A credible design supports an incremental claim; it does not guarantee or isolate one.

Before-and-after is observation

A rise after the change may reflect season, pricing, a busier market or work brought forward from a later period. Before-and-after totals alone cannot establish how much additional business the AI change created.

A comparison needs a credible design

A comparable territory, team or staged rollout can provide a reference for what would otherwise happen. The claim depends on credible design, similar groups and periods, and checks for other changes; a comparison does not automatically identify an AI effect.

Watch selection and adoption

If the strongest estimator or easiest requests go into the assisted group, results flatter the change. If staff ignore handoffs, a weak result may reflect adoption rather than the tool itself.

Decision for your leadership teamWhich comparison group is realistic for us, and which confound concerns us most?

06First Test · overview

Choosing one first test

What bounded test should we fund first, and how will we decide what comes next?Illustrative contractor case · no figures are given or implied

One eligible request type
One accountable estimating lead
One explicit investment period

Evidence agreed before starting

Linked records
Comparison read
Full cost list
Delivery room

Decision ruleAgreed before starting

Positions · none preselected

ExpandThen trace the next bottleneck
Repair
Stop

No position is preselected. The comparison, the full cost list and the decision rule are agreed in advance, so the evidence cannot be read to fit the hope.

One request type, one owner, one period

Bound the test to one eligible request type with an accountable estimating lead and an explicit investment period. A narrow scope makes records linkable and the comparison readable.

Agree the decision rule in advance

Before starting, agree the comparison, the full cost list and what result would justify expanding, repairing or stopping. Choosing the rule afterward invites reading the evidence to fit the hope.

Expansion follows the next bottleneck

If the test supports expansion, check whether delivery, billing or another territory becomes the next constraint. Repeat the trace there rather than assuming the first result transfers unchanged.

Decision for your leadership teamWhich request type, owner and investment period would we put through a first trace?

Find the AI investment behind the next profitable sale.

Start with demand you already attract

AI prepares; people decide and commit

Follow the quote to a customer decision

Delivery has to have room

What remains after every cost

Was it additional, and was it the change?

Fund one bounded test, then decide

Trace one request before funding the next AI project