| Metric | Before | After |
|---|---|---|
| Margin floor at submission | Knowledge-dependent (varied by who reviewed) | Structurally enforced (every submission, every path) |
| Time per proposal body | 20-25 min (manual, knowledge-intensive) | Under 60 seconds (form submission to email) |
| Proposal pricing confidence | Conservative, push-back prone | Justified, detailed, approved (50% higher labor value, no pushback) |
| Institutional knowledge access | Held by senior individuals | Encoded in system (all authorized users, always) |
| Proposal audit standard | Manual end-of-process (aspirational checklist) | Multi-point checklist at submission (structural, not optional) |
| Roles producing compliant output | Estimating team only | Any authorized team member |
| System ownership | No documentation | Team-owned (Code Logic Guide + Operator Card) |
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."
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.
The 37-second proposal is Step 4. It works because the three steps before it were solid.
"It captured the tribal knowledge, and that was the ultimate goal of everything we are doing."
"Instead of getting so many calls escalated to me, it should be: follow the system."
"I feel more confident in the work going out. The standard is in the system now, not just in my head."
"High-margin proposals leaving our door, and after the work is done, margins maintained or improved. That is what winning looks like."
"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."
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."
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."
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.
Six things made the difference, none of which are technical.
"Teams are no longer asking whether AI can help. They are asking where to build next."
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 touchResults in under 30 days · System owned by your team at handoff · Confidentiality guaranteed