AI-Assisted Estimating Case Study | Evangelia Leclaire
See the work and the results.View the Case Studies →
An AI Enablement & Workflow Case Study

How a Mechanical Field Operations Company Turned Its Estimating Function Into an AI-Assisted System in Under 30 Days

Challenge
Three problems, named by leadership. Margin that eroded because pricing discipline lived in one person. Critical technical knowledge that only existed in people, not playbooks. A quality standard written down but difficult to consistently enforce.
Outcome
In under 30 days: a purpose-built AI system that enforces all three. Proposals produced in under 60 seconds. A multi-point QA checklist enforced structurally at submission. $250K+ in annual team time recaptured.
Confidentiality: Company name, personnel names, and pricing data changed or omitted with client approval. Standard practice on every engagement.
Mechanical Field Operations
Industry
Under 30 Days
Full Build to Handoff
30+ / day
Quote Volume
10+ Team Members
Estimators, Account Managers, Assistants, Ops Director
The Results, Confirmed
<60s
Per Quote
Form to client-ready email
$250K+
Annual Time Savings
Proposal body work alone
QA Standards
Multi-point, enforced structurally
50%
Higher Labor Value
Approved without pushback
30+
Scope Types Encoded
Institutional knowledge in system
$250K+
Annual value in time savings on quote building
Estimators, account managers, assistants, Field Ops Director, and executive review time. Revenue upside from better proposals accumulates on top.
Run that number against your own team. Then ask how many functions look like this.
Every engagement starts here. Not with technology. With where the value is and what it costs to leave it.
Adoption & Impact

Before and After

MetricBeforeAfter
Margin floor at submissionKnowledge-dependent
(varied by who reviewed)
Structurally enforced
(every submission, every path)
Time per proposal body20-25 min
(manual, knowledge-intensive)
Under 60 seconds
(form submission to email)
Proposal pricing confidenceConservative, push-back proneJustified, detailed, approved
(50% higher labor value, no pushback)
Institutional knowledge accessHeld by senior individualsEncoded in system
(all authorized users, always)
Proposal audit standardManual end-of-process
(aspirational checklist)
Multi-point checklist at submission
(structural, not optional)
Roles producing compliant outputEstimating team onlyAny authorized team member
System ownershipNo documentationTeam-owned
(Code Logic Guide + Operator Card)
How It Started

Built in the Right Order. Adopted Before Anyone Asked.

The goal was consistent quality. Proposals that go out right every time, regardless of who submits them. Automation was only worth building once there was something worth automating.

The foundation came first. We captured 20+ files of institutional knowledge from senior leadership: playbooks, scope standards, pricing discipline, technical rules. Validated each one. Built it into Claude Projects and developed an AI-Assisted Estimating Skill.

After training the team on adoption, the Field Operations Director began using the tool on his own. The output exceeded expectations and increased efficiency. Then, we discovered a new part in the workflow for AI-assisted automations.

That is when the co-build began. Live working sessions with the Field Operations Director. We evaluated the workflow and the opportunity to integrate with existing field management systems. There were system constraints. We found workarounds. We built a multi-step agentic workflow integrated into the system: a form, APIs, Apps Scripts, system prompts, and compliance guardrails. We tested. We iterated. We shipped. The handoff was not a training session. It was the moment he re-authorized the system under his own account and it ran.

"The confidence in the work going out has changed. When a proposal is this detailed, it reads like what it is. Clients approve it."

F
Field Operations Director
On the AI Estimating Skill, before the automation was built
Why the sequence matters
The 37-second proposal is the last step, not the first. It works because the knowledge base was built before the skill, the skill proved its value before the automation, and the automation was built with the person who would own it. Most AI projects start at the end of that sequence. This one did not.
The Challenge

The Knowledge Was There. The System Was Not.

The team filters through a high volume of work. The standards existed. But critical knowledge lived in individuals, not systems. Proposals, pricing, and quality assurance were not consistently to standard. The knowledge was there. The system to deliver it consistently was not.

L
CEO & COO
What winning had to look like
Outcome 01 — Reliable execution at scale
Field and service operations execute reliably. Teams complete work correctly, on time, and to company standards with minimal supervision.
Outcome 02 — Guardrails that enable autonomy
Systems enable autonomy while providing early visibility when execution, performance, or outcomes begin to drift.
Outcome 03 — Margin protected through completion
Proposals, estimates, and job execution are financially protected. Margin for risk, labor variability, and operational uncertainty is consistently built into pricing and maintained through completion.
F
Field Operations Director
What he was dealing with every day
He was the quality filter on every proposal
After the team submitted quotes, he was rewriting each one before it reached the customer. Dozens of times a day. He was the quality filter and spent a lot of time reworking each proposal.
Better proposals meant fewer objections
Detailed proposals meant fewer objections. He had figured this out. The process had not caught up.
Critical rules that only he had heard said aloud
When a job requires a crane. How to calculate refrigerant. How many labor hours per scope type. None of it written down. A new estimator would not know.
Where they connected
The COO wanted knowledge in a system. The Field Operations Director was manually doing what that system was supposed to do. The foundation was already built. The path was clear.
How It Was Built

Four Steps. Each One Had to Come Before the Next.

The 37-second proposal is Step 4. It works because the three steps before it were solid.

Step 1
Map the workflow
  • The estimating process mapped end to end. Every stage, every role, every handoff, every breakdown.
  • The estimators were doing their technical job well. The gap was communication, not capability.
  • The pipeline stage map, QA checklist, and intake standard defined what "correct" looked like before any AI was asked to produce it.
Step 2
Capture the knowledge
  • Structured sessions with the COO captured what had never been written down: pricing rules, scope thresholds, labor hours by job type, when a crane is required.
  • One session extracted 30+ scope types. That session took the system from quote formatter to institutional knowledge engine.
  • 20+ files: playbooks, scope libraries, pricing guidelines, technical references. Validated before any AI was trained on it.
Step 3
Build the AI tool
  • An AI Estimating Tool built on the full knowledge base: scope standards, technical rules, pricing logic, the organization's own proposal voice.
  • Tiered access by role. Estimators see what they need. Managers see more. Sensitive pricing protected by design, not policy.
  • This is the tool the Field Operations Director adopted on his own. Dozens of times a day, by hand. That proved Step 4 was worth building.
Step 4
Automate the workflow
  • Built with the Field Operations Director as the day-to-day owner. Google Form, Apps Script, Anthropic API, finished proposal by email in under 60 seconds.
  • Sensitive pricing blocked from customer output by design. Multi-point QA checklist enforced at submission. A person reviews before anything goes to the customer.
  • Code Logic Guide, Operator Reference Card, and Handoff Memo. Written for the Field Operations Director, not a developer.
What was handed off
📋
Code Logic Guide
A plain-language guide. How every part works, when to change something, and what will break if you do it wrong. Written so the team can maintain and improve the system without outside help.
🃏
Operator Reference Card
One page. Five things to know: check if a submission worked, change a setting, update the scope library, run a test, diagnose a bad output.
📨
Handoff Memo
Who owns the system, what they're responsible for, and when to ask for outside help. The engagement ends when the team owns it.
Tech Stack
Intake
Field Management Software
+ Google Form
Orchestration
Google Apps Script
Intelligence
Anthropic API
Human in the Loop
HTML Email
For review, approval, and integration into the field management system.
Breakthrough Use Cases

What under 30 days of deliberate AI enablement looks like in practice.

20-25 min
Under 60 sec
Estimating
End-to-end proposal generation. Estimator inputs structured facts. System produces a complete, client-ready proposal drawing from the knowledge base. Confirmed in live session in under 60 seconds.
30+ quotes per day at consistent, audited standard.
Conservative
Justified
Revenue Capture
Proposals built on institutional knowledge command higher prices. Technical complexity and scope requirements are surfaced before submission, not after the client objects.
One confirmed example: 50% higher labor value. Approved without negotiation.
Manual catch
Structural
Quality Assurance
Quality standards moved from manual end-of-process review to structural enforcement at submission. Scope gaps are harder to produce, not just easier to catch.
Multi-point QA checklist. Every submission. Every path.
Tribal
Systematic
Knowledge Distribution
Technical thresholds and pricing rules encoded into the system. What previously required a senior expert in the room is now in the process, available to every authorized team member.
Every team member. Same knowledge base. Regardless of tenure.
Consultant
Team-owned
Handoff
In the final session, the Field Operations Director caught a quality issue live and fixed it himself. The Code Logic Guide was written for him, not a developer. He now runs, maintains, and improves the system independently.
Full team ownership. Improvement is continuous, not consultant-dependent.

"It captured the tribal knowledge, and that was the ultimate goal of everything we are doing."

F
Field Operations Director
Mechanical Field Operations Organization
In Their Own Words

What the people closest to the work said.

"Instead of getting so many calls escalated to me, it should be: follow the system."

C
Chief Operating Officer
On the knowledge standard

"I feel more confident in the work going out. The standard is in the system now, not just in my head."

F
Field Operations Director
After the first week of live use

"High-margin proposals leaving our door, and after the work is done, margins maintained or improved. That is what winning looks like."

C
Chief Operating Officer
On the definition of winning

"I'm fully on board. I see what we can automate, and I want to focus on the things we can't do with AI yet."

F
Field Operations Director
On championing the build
The People Who Changed

AI adoption is a people story before it is a technology story.

F
Field Operations Director
Operational Owner · Daily User · System Champion
Before
Manually running dozens of proposals through Claude every day, by himself, because he could not stop noticing the quality difference. Nobody asked him to.
After
Authorized trigger owner and email sender. The system runs. He reviews for quality and accuracy. The work that used to sit on his desk now happens before it reaches him.

Two weeks before the build began, he had already proven the case. Running thirty quotes a day through the AI Estimating Tool on his own, because he could not ignore the quality difference. In the final session, he caught a quality issue live and fixed it himself.

"I feel more confident in the work going out. The standard is in the system now, not just in my head."

E
Evangelia Leclaire
AI Enablement Lead · Architect · Trainer
Approach
Sit with the people doing the work before recommending anything. Map the workflow before building the tool. Build the knowledge base before automating anything.
Standard
Sustainable independence at handoff. The team runs, maintains, and improves the system on their own, without calling anyone.

Every engagement starts with listening. Sit with the people doing the work before recommending anything. Map before building. The Code Logic Guide was written for the Field Operations Director, not a developer. The goal is a team that owns and improves the system after the engagement closes.

"Sustainable independence was the goal. Not dependency on me."

Metrics, KPIs & ROI

What Is Confirmed. What Is Accumulating.

The organization already had a four-metric governance framework before this engagement began: turnaround time, win rate, gross margin quality, and change-order frequency. That framework is the measurement standard.

Confirmed Outcomes
Confirmed
Turnaround Time
How quickly complete, reviewable proposals are issued
Now
20-25 minutes reduced to under 60 seconds. Confirmed in live timed session.
Tracking For
Days from deficiency to client approval, accumulates with volume.
Confirmed
Team Time Recaptured
Proposal body work freed across the team
Now
$250K+ per year across multiple estimators, the Field Ops Director, and executive review time. Revenue upside from better proposals accumulates on top.
Tracking For
Whether freed capacity measurably improves escalation response and review quality.
Confirmed
QA Compliance Rate
Proposals meeting all governance standards at submission
Now
Multi-point quality checklist enforced structurally at submission. Compliance is architectural, not dependent on individual availability.
Tracking For
First-pass accuracy: proposals reaching the client without post-approval revision.
Accumulating with Volume
Accumulating
Win Rate & Approved Value
How often proposals convert, and at what labor value
Directional
One confirmed example: 50% higher labor value approved without negotiation. Fewer client clarification calls reported.
Tracking For
Approval rate and average approved labor value vs. historical baseline.
Accumulating
Change-Order Frequency
How often scope was unclear at time of quote
Directional
Pre-approved scope blocks and required exclusions reduce scope gaps at the source, before the job is dispatched.
Tracking For
Change orders per job vs. prior period, tracked at close.
Accumulating
Gross Margin Quality
How closely actual job economics align to the estimate
Built In
Financial and technical rules encoded and enforced at submission. Margin risks are flagged before the quote advances.
Tracking For
Gross margin variance between estimate and actual close, tracked against pre-engagement baseline.
Confirmed Now
$250K+ in recaptured proposal time annually. Multi-point QA checklist structural at submission. Full team ownership. System running and maintained independently.
Building Over Time
Win rate delta, higher approved labor values, fewer change orders, tighter margin alignment. One example already shows 50% higher approved labor value. Scale that across 30 quotes per day, the revenue upside is the larger number.
The Critical Insights

Why This Worked

Six things made the difference, none of which are technical.

1
The problem was understood before anything was built
The first step was not building anything. It was understanding what the work was supposed to produce and where the process was failing. The tool was chosen for the problem, not the other way around.
2
The steps happened in the right order
Map. Capture. Build. Automate. The 37-second proposal is Step 4. Most AI projects start there, skip the earlier work, and wonder why the output sounds wrong. Sequence is not optional.
3
Sensitive information protected by design, not by policy
Pricing, labor rates, and negotiation thresholds stay out of customer-facing output because the system is built that way. Not because someone remembers to remove them.
4
The team was taught to own the system, not just use it
Before the engagement closed, the Field Operations Director caught a quality issue live and fixed it himself. The goal was always to end with a more capable team. That happened.
5
Adoption happened because it was earned, not mandated
Nobody told him to run thirty quotes a day through the AI tool. He did it because the output was better. That is the only adoption that lasts.
6
Better proposals are a revenue strategy, not just a time-saver
A proposal that explains the work clearly removes the reason to push back. One example: 50% more labor submitted and approved without negotiation. That is the revenue case. It is larger than the time savings.
The Goal

"Teams are no longer asking whether AI can help. They are asking where to build next."

C
Chief Operating Officer
Mechanical Field Operations Organization
Work With Evangelia

Your business has the same gap.
The only question is whether it stays there.

Every engagement starts with understanding your workflows, finding where knowledge lives only in people, and showing you exactly what to build and in what order. No tools recommended before the work is understood.

Get in touch

Results in under 30 days  ·  System owned by your team at handoff  ·  Confidentiality guaranteed

What comes next

From "Can we do this?" to "What do we build next?"

Where the team is now
The estimating system is live. The team owns it. The question has moved from whether this is possible to which function to build next.
The goal
Any authorized team member can produce a technically sound, client-ready proposal from structured inputs, without needing a senior expert present. The knowledge is in the system. People are the competitive advantage, not the bottleneck.