Composite client. Details changed to protect confidentiality. The patterns are real.
You’ve read the AI Office page. You’ve seen the pricing. You want to know what actually happens after you sign. Here’s a real 90-day arc, in the operator’s own terms.
The company
A $12M revenue industrial services firm in East Texas — third-generation family ownership, 60 employees, mostly field crews servicing utility and industrial customers across the Pine Belt. EBITDA roughly $2M. PE-curious but not yet sold.
The CEO is 52. His operations director is 47. They both grew up in the field. They use computers but they don’t love them. The CEO had run two AI experiments in the prior 18 months — one with a vendor that promised “AI-powered dispatch,” one with a custom GPT his sales lead built. The first lasted 4 months and produced nothing. The second is still running but nobody uses it.
He told us on the discovery call: “I’m not sure I want to spend another dollar on AI. But I’m pretty sure I can’t afford not to.”
Discovery call (Day -7)
We did a 30-minute call. He asked what we did. We asked about his business — what was working, what was draining his team, where AI might fit.
He named three things:
- Quoting was killing them. 5–7 days to get a quote out, losing bids to competitors who turned around in 2.
- His senior estimator — 64 years old, planning to retire in 18 months — knew things nobody else knew. Pricing patterns. Risk flags. Vendor reliability. All in her head.
- Field-to-office reporting was a mess. Crews submitted photos and partial notes. The office spent days reconstructing what happened. Invoicing lagged 2–3 weeks.
We told him AI Office could probably tackle all three over 6–9 months. We sent him a 2-page Discovery Letter the next morning with a recommended Operator tier ($5,000/mo) and a rough 90-day arc.
He signed the SOW on day 5.
Day 1: Kickoff
We arrived at his office at 8 AM — our senior strategist (Sarah, a composite name) and our build engineer (Mike). The CEO had cleared 90 minutes. His ops director, his senior estimator, his office manager, and his sales lead were in the room.
For the first 30 minutes, the team walked us through the business in their own words. We asked questions. Mike took notes mostly about the systems — their ERP, their dispatching tool, their accounting platform. Sarah was listening for the harder thing: where the pain actually lived.
The harder thing turned out to be the senior estimator. She was sharp, fast, and quietly skeptical. Everyone in the room deferred to her on pricing. When we asked her what would have to be true for AI to help with quoting, she said: “You’d have to understand what I know. You won’t.”
We left the meeting at 11 AM with read-only access to the ERP, a walkthrough of their quote workflow, and a clear sense that the senior estimator was either going to be our biggest champion or our biggest blocker.
That afternoon, Sarah and Mike spent 4 hours at the hotel writing up what they’d heard. They had a draft inventory of 14 candidate AI use cases. Three were obviously the priorities; the rest were future-quarter material.
Days 2–9: The roadmap
Sarah worked async for a week. She had two more calls with the operations director, one with the senior estimator (the most important call), and one with the sales lead. She reviewed the past 30 quotes the firm had sent — wins and losses. She mapped the data flow from inbound RFQ to signed contract.
What she delivered on day 9 was a 1-page roadmap with three priorities:
- Quote acceleration (90-day priority) — AI parses inbound RFQs, pulls historical jobs with similar scope, drafts the quote with margin guidance. Human approves in 30 minutes vs. 4 hours. Projected impact: quote turnaround from 5 days to 24 hours, win rate improvement of 8 points, worth approximately $400K/year.
- Estimator knowledge capture (6-month build) — A workflow that captures the senior estimator’s pattern recognition through structured conversations and review of her past decisions. Codifies the patterns into a queryable knowledge layer.
- Field-to-office reporting (months 4–6) — Crew submits photos + voice notes from the field. AI structures them into the ERP and drafts the invoice line items. The office stops reconstructing.
The CEO pushed back on two things. He wanted the field reporting moved earlier — it was a more visible win to his team. And he didn’t want the knowledge capture to feel like “we’re going to replace her.” Both reasonable.
Sarah updated the roadmap. We agreed to start with quote acceleration in week 3.
Days 10–28: The first build
Week 3 kickoff (in-person): Sarah and Mike were back. The session was with the sales lead and the senior estimator. Two hours, focused on how a quote actually happens.
By lunch, Mike had sketched the workflow:
- Inbound RFQ arrives via email
- AI parses scope, equipment requirements, timeline, customer type
- AI pulls the 5 most similar historical jobs from the past 18 months
- AI drafts the quote in their standard format, with margin guidance
- Quote routes to a human review queue with red flags surfaced
- Human approves or revises, sends to customer
The senior estimator pushed back on three things. Mike adjusted. By 3 PM they had an agreed scope.
Mike worked the next two weeks from Frogslayer’s office, using anonymized historical RFQs and quotes the firm provided. He pinged the sales lead twice on Slack with questions — each exchange took 15 minutes. By day 25, he had a working prototype. He shared a Loom walking through it.
Day 28 demo (in-person): Mike and Sarah were back. The sales lead and the senior estimator ran the workflow on three real inbound RFQs. Two worked cleanly. One caught a real edge case Mike hadn’t anticipated — and the senior estimator caught it before Mike did. That mattered. She was now part of the build.
The CEO sat in for the last 30 minutes. He was visibly relieved. The senior estimator’s body language had shifted; she wasn’t a blocker anymore.
Days 29–60: Refinement and production
Mike fixed the edge case in week 5. He added two features the sales lead had requested in the demo. By week 7, the workflow was in production. The sales lead and senior estimator were using it for real quotes.
Sarah set up monitoring so any errors would flag Mike directly. She also set up a weekly 30-minute check-in with the CEO to keep him informed.
By week 8, the team had run 47 quotes through the system. Average turnaround: 22 hours, down from 5+ days. Win rate was too early to measure but anecdotally trending up.
Sarah documented the wins in the first monthly briefing to the CEO.
Days 61–90: The second track
With quote acceleration in production, we pivoted to the field-to-office reporting build. Same pattern: Mike scoped it in week 9, Sarah ran the kickoff with the operations director and two crew leads in week 10, build through weeks 11–12.
Quote acceleration meanwhile kept producing. By day 90, the firm had run 89 quotes through it. Turnaround was now averaging 18 hours. The team trusted it. The senior estimator had started telling her peers in the East Texas industrial services community.
The CEO told us at the day-90 review: “This is the first AI thing that’s actually felt like it works. And nobody got fired or felt threatened. That’s the part I didn’t expect.”
The year-1 picture
We’re still in flight with this client. Year 1 has more in store:
- Field-to-office reporting in production by month 5
- Estimator knowledge capture as a multi-quarter project starting month 6
- Likely tier upgrade to Embedded ($10,000/mo) around month 7 as the build queue grows
- A first Value Sprint around month 10 — likely a deeper integration with their dispatch tool: $40K, 5 weeks, single KPI with the 12-month guarantee
Documented ROI through day 90: roughly $80K in faster-quote win-rate improvement, ~5 hours/week of estimator time recovered, and 1 FTE-equivalent of office time saved on invoice reconstruction.
Total AI Office spend through day 90: $15K (3 months × $5,000 Operator tier).
ROI ratio: roughly 5X within the first quarter. The AI Office retainer targets at least 3X payback — it pays for itself — and here that target was already cleared by month 3.
What this costs you vs. the alternatives
A national consultancy would have scoped this engagement at a $150–250K project minimum, started in 6 months, and ended in 6 months with a deck and maybe one workflow.
An internal hire would have cost $300–400K loaded for the year, taken 6 months to ramp before shipping anything, and produced one workflow at a time.
Our AI Office Operator tier: $60K/year. Ships in 30 days. Quarterly rhythm. Senior team. Three workflows shipped or in flight by month 6. The senior estimator’s pattern recognition starts getting captured before she retires.
Different math.
Want to talk about a similar engagement?
If you’re an owner-led service business in our priority industries and the patterns in this case study sound familiar — quotes that take too long, knowledge concentrated in one or two senior people, field-to-office data that’s a mess — there’s a good chance a similar 90-day arc would work for you.
The first conversation is free. 30 minutes. We tell you straight whether AI Office fits and what the first 90 days would actually look like.