Quote-to-invoice, compressed from days to minutes — in nine days.

CARISAM Caribbean Sourcing isn't short on demand. Lance Cantrell has spent years building the relationships, product knowledge, vendor depth, and customer trust it takes to source construction materials reliably across the islands. The constraint isn't capability. The constraint is operating capacity.

Quote requests arrive across email, text, WhatsApp, phone, PDFs, plans, and informal notes. Lance interprets, searches prior projects for pricing, picks a vendor path, requests supplier pricing, normalizes the responses in his head, prepares the customer quote, and decides what gets sent. The business has scaled to the point where too much of the workflow lives in one person's email inbox and memory.

Then Lance hired Jim — a logistics expert and the first full-time addition to the team. The faster Jim could become useful, the faster Lance could get out of the inbox and back into the field. But Jim didn't have a clear operating model to learn — because there wasn't one written down. The operating system was Lance.

The brief: build the first operating layer that gets the repeatable parts out of Lance's head without stripping away the judgment and relationships that make CARISAM valuable. Thirty days, $7,500 flat fee.

The first week was discovery: seven recorded interviews with Lance plus a full-day on-site session with Lance and Jim. The point wasn't generic process mapping. It was to find the exact place where the business was leaking time — and to build the operating layer around that single workflow first, not all of them at once.

That workflow turned out to be quote turnaround — the request-to-RFQ-to-comparison-to-decision-to-customer-quote sequence that drives every dollar through CARISAM. Map that one well, build a system around it, and the rest of the operating model (orders, payments, shipping, closeout) attaches to the same spine.

Then the diagnosis went into a 7-table relational data model with ~80 fields specified: customers, vendors, quotes, quote items, RFQs, vendor responses, vendor response items. The model wasn't a spec sitting in a slide deck. It was the schema for the working app that would be live the following week.

The build was paired: ThoughtLeap.ai owned the operating-model design, customer-facing architecture, and the front-end. Steve Dakh partnered on technical depth. Both worked from the same discovery artifacts, both lived in the schema, both pushed the system forward in parallel.

Beyond the three live sites, the engagement also produced the things that turn software into an operating system:

• New-hire operating playbook for Jim — what Jim owns immediately, what requires Lance's approval, daily rhythm, the interruption rule. People-process design, not just documentation.
• Six AI workflows with explicit guardrails — phone/meeting capture, email chronology reconstruction, supplier RFQ draft, vendor response comparison, quote review checklist, vendor research. Each with output specs and "what AI never owns" rules.
• 30-day next-phase roadmap — six phases covering customer PDF generation, order conversion, email integration, PDF / email extraction, multi-user, hardening.
• Detailed technical brief for the incoming engineer — Steve Dakh's continuity document, with architecture decisions, security model, and Phase 2 questions to answer first.
• Quote / pro-forma workbook structure — optimized Excel pro-forma the app exports to directly, preserving CARISAM's existing format.

• Quote requests have one home. Every new request, regardless of channel, lands in one record with customer, project, destination, source, priority, line items, and notes. Within one business day, every active quote is visible at a glance.
• Vendors respond through a structured link, not an email chain. Lance generates a per-vendor RFQ that produces a copy-pasteable email plus a unique unguessable response URL. The vendor fills in pricing per line item, lead time, freight, validity, confidence. No vendor account. Mobile-friendly.
• Comparison happens automatically. When two or more vendors respond, the app surfaces lowest price, fastest lead time, highest confidence side-by-side. The decision brief that used to live in Lance's head is now a visible artifact.
• Margin math runs once. Lance picks the winning vendor, enters margin or sell price, and the system computes the customer-facing pro-forma. CSV export drops into the existing CARISAM workbook unchanged.
• Jim has a job description that fits in software. The handoff playbook isn't a PDF in a drawer. The app's permissions, status fields, and approval gates encode it. Jim can prepare quotes; Lance approves what goes out.
• The system has a roadmap, not just a wish list. The next-phase plan is sequenced, technically brief'd to Steve Dakh, and waiting to be executed when Lance signs Phase 2.

• Time-boxed and fixed-fee. Every engagement has a hard finish line and a hard price. No hourly billing. No scope drift.
• Installed, not recommended. The deliverable is live working software plus the operating playbook that surrounds it — not a PDF report.
• Paired execution. Operating-model insight from Brandon Carson, engineering depth from Steve Dakh. Both inside the engagement from day one.
• Built on the client's infrastructure. Code runs on the client's own server, under their domain. They own the system after handoff. No vendor lock-in.
• One operating layer first. We don't build the whole CRM. We map the highest-value workflow, ship it, prove the data model on real work, then expand.