1. ROI summary
The same shape you saw in the free scorecard, now carried into the full assessment. The scorecard estimates from your answers; this report works from your processes.
| Year | Value | Cost | Net | ROI |
|---|---|---|---|---|
| Year 1 | $2.1M | $0.2M | $1.9M | 9.2x |
| Year 2 | $4.2M | $0.5M | $3.8M | 9.0x |
| Year 3 | $6.5M | $0.7M | $5.9M | 8.8x |
2. Process audit scorecard
Current result vs benchmark. Benchmark equals target. Each gap below maps to one or more of the processes analyzed later in this report.
Gross margin
$24M revenue. Margin trails the sector benchmark.
Change-order capture
Higher is better. Field signals that reach a billed, collected change.
Payment administration
Lower is better. Pay-apps, waivers, invoices, follow-up.
Proposal cycle time
Lower is better. Qualified proposals out the door.
Start-ready backlog
Lower is better. Share of backlog not yet ready to staff and start.
Sources: NYU Stern margins, Rabbet payments, ABC backlog, HBR lead response. Proposal cycle is an internal operating standard. Full list in Part D.
3. The plan: sequence and first 90 days
Highest value at lowest risk goes first, then a 90-day window that proves the method before anything scales.
The order we would do this
Each step earns trust and data for the next. Nothing here requires ripping out a tool the team already uses.
The first 90 days
Focused on pilot one (change capture) and pilot two (cash), the two that fund and de-risk the rest.
Capture control
- Voice-to-field-note live on two jobs.
- Change-order draft and routing.
- Weekly capture funnel review.
Cash control
- Pay-app generated, not assembled.
- AR owner named per account.
- Live cash and aging view.
Proof and guardrails
- Before-and-after on pilot metrics.
- Human-at-the-gate rules written down.
- Decision on pilot three.
Days 1-15
Connect data, baseline the two pilots, agree the guardrails.
Days 16-30
Change capture live on two jobs. Pay-app generation built.
Days 31-60
AR cadence running. Capture funnel and cash view in use.
Days 61-90
Measured gains reviewed. Pilot three scoped and started.
4. The processes we analyzed
A full review looks across the whole company, not one department. We mapped the five highest-value workflows across five core business functions, scored each with the five dispositions, and flagged where AI actually earns its place. The full plan for each is in the deep dives below.
Proposal and bid turnaround
Estimate to signed proposal, plus follow-up on open bids.
Pay-app and AR collection cycle
Pay applications, lien waivers, invoicing, and collections follow-up.
Field-to-office change capture
Daily field notes, RFIs, and the change-order signal that starts in the field.
Crew and sub scheduling readiness
Keeping backlog start-ready and crews and subs sequenced.
Client comms and after-hours intake
Owner and client updates, plus calls that arrive after hours.
Numbers describe the problem
Four functions map to direct line items in the $2.1M case (proposal, pay-app/AR, change capture, scheduling). The fifth (client comms, customer service) protects margin and repeat work. We never promise a result.
The five dispositions
Every process is sorted into one or more of these before any tool is chosen. Most high-value wins need no AI at all.
5.1 Proposal and bid turnaround
Daily through bid season. Mostly automatable, with a human owner on price and scope.
The workflow today
- Invitation to bid lands in an estimator's inbox and is tracked in memory or a spreadsheet.
- Estimator pulls quantities and chases subs for pricing by phone and email.
- Numbers are assembled in a one-off spreadsheet, formatting redone each time.
- Proposal narrative is written from scratch, reusing last job's language by hand.
- Proposal waits for a principal to review and sign before it goes out.
- Follow-up on sent bids is inconsistent. Some are never chased.
What changes
- Report: a single proposal aging board shows every open bid, its age, and its owner.
- Simplify: a standard proposal template and sub-pricing request replace the one-off rebuilds.
- Automate: sub-pricing requests and reminders go out and chase themselves on a schedule.
- AI assist: a first-draft narrative is generated from the estimate and the standard template. The estimator edits, never starts from a blank page.
- Optimize: automatic follow-up sequence on every sent bid until it is won, lost, or withdrawn.
5.2 Pay-app and AR collection cycle
Monthly billing cycle plus weekly collections. Mostly automatable with almost no AI. This is the clean no-AI win.
The workflow today
- Each month, staff rebuild the pay-application package per project by hand.
- Lien waivers are requested from subs by individual emails and tracked on a list.
- Invoices are created, formatted, and sent one at a time.
- AR follow-up happens when someone remembers, not on a schedule.
- Owner or controller reconstructs cash position from several systems at month-end.
What changes
- Eliminate: manual package assembly. The pay-app is generated from project data on a schedule.
- Automate: lien-waiver requests and reminders to subs run as a rules-based sequence.
- Automate: AR follow-up fires on a fixed cadence with an owner named per account.
- Report: a live cash and AR-aging view replaces the month-end reconstruction.
5.3 Field-to-office change capture
Daily, on every active job. Partially automatable. This is the recommended first pilot: small, low-risk, real.
The workflow today
- A super notices a field change, a delay, or an owner request on site.
- It gets remembered, texted, or scribbled, and sometimes lost.
- Daily logs are written at night from memory, if at all.
- The change is priced and submitted days later, with weak backup.
- Owners dispute thinly documented changes, and some are written off.
Only 52% of field change signals reach a billed, collected change.
What changes
- Simplify: the super leaves a two-minute voice memo at end of day. No new app to learn.
- AI assist: the memo becomes a formatted daily log and a draft change-order with photos attached.
- Automate: draft changes route to the PM for pricing and to the owner-rep for sign-off, with reminders.
- Report: a change-capture funnel shows where signals are lost, week over week.
5.4 Crew and sub scheduling readiness
Weekly look-ahead plus daily adjustments. Partially automatable, mostly through visibility, not AI.
The workflow today
- Backlog is tracked as contract value, not as start-readiness.
- Permits, submittals, and long-lead items live in separate places.
- The three-week look-ahead is rebuilt manually in spreadsheets.
- Crew and sub conflicts surface on the morning they happen.
- Idle time and stalls are noticed after the fact in job-cost reports.
28% of backlog is not start-ready against a 10% benchmark.
What changes
- Report: a start-ready board scores each backlog job on permits, submittals, and long-lead status.
- Simplify: one look-ahead view replaces the rebuilt spreadsheets.
- Automate: readiness flags and reminders fire as gate items slip.
- Optimize: crew and sub sequencing is planned against readiness, not just contract dates.
5.5 Client comms and after-hours intake
Continuous. Partially automatable. Compounding value: it protects margin and repeat work rather than showing up as one line item.
The workflow today
- Project updates to owners are ad hoc, usually triggered by a problem.
- After-hours calls go to voicemail and are returned late or not at all.
- No record ties a missed call to a lost or delayed opportunity.
- The same status questions get answered by phone, over and over.
What changes
- Automate: a scheduled, plain-language project update goes to each owner on a set cadence.
- Automate: after-hours calls are captured, triaged, and routed, with emergencies escalated.
- Report: every inbound call and its outcome is logged so missed opportunities are visible.
- AI assist (optional): an after-hours assistant takes a structured message. A human handles anything urgent or unclear.
6. What we would not automate yet
A real assessment is as clear about what to leave alone as what to change. These stay human in this engagement.
Stays with people
- Pricing and scope on any proposal or change order.
- Releasing invoices, waivers, and any money document.
- Every hiring and screening decision (EEOC).
- Emergency and judgment calls from clients.
Not yet, and why
- Full estimating automation. Data is not clean enough to trust.
- Autonomous client replies. Tone and stakes are too high early.
- AI-driven scheduling decisions. Earn it after the readiness board is trusted.
- Anything that sends without a person, until the pilots prove out.
7. Source anchors
- NYU Stern margins: Engineering/Construction gross margin near 15.5%.
- Rabbet payments: GCs lose around 65 admin hours a month to payments.
- ABC backlog: roughly 8 months of backlog in early 2026.
- Autodesk/FMI: significant nonproductive time and data-linked rework.
- ASCE rework: field rework cost as a share of contract value.
- HBR lead response: response speed drives win rate.
Sample uses anonymized, modeled figures consistent with these public benchmarks. The audit produces the measured version from your real data.
This is what the audit delivers
What you just read is a worked sample. The free scorecard gives you this shape from your own answers in a few minutes. The audit turns it into the measured version: your processes, your numbers, your sequence.