Vaan's sales function worked. Deals closed and SOWs went out. But every SOW took about eight hours of senior time to draft, and the rules that made it accurate existed nowhere except in the head of the person writing it.
The Client
Vaan Group is a Shopify Platinum agency known for its Brand Conversion Design framework, serving luxury, lifestyle, and DTC brands. In early 2026, Vaan was acquired by a larger digital experience company. Deal flow grew, and so did the number of people and systems a deal had to pass through.
Vaan's sales motion runs on Salesforce, Google Workspace, and Slack. Discovery calls lead to scoped SOWs across three recurring archetypes: platform migrations, redesigns, and optimization work. Scoping those projects accurately is what protects Vaan's margin on the delivery side.
The Challenge
At roughly 10 SOWs a month and 8 hours each, SOW drafting consumed about 80 senior sales hours every month.
The scoping rules that kept SOWs accurate lived in one person's head: hours in multiples of 4, a 400-hour development minimum, and separate rules for each archetype.
Around the SOW: no version history across negotiation rounds, about an hour of manual research before every discovery call, and no follow-up path for closed-lost deals based on why they were lost.
Estimates were also never compared to delivered hours. Discovery surfaced a 21% overrun on beauty-category migrations that nothing had caught.
What We Found in Discovery
We mapped the sales cycle stage by stage and asked one question at each step: is this a rule, a judgment call, or something a tool Vaan already owns could handle? The answer shaped every recommendation.
Before the call. Prospect research was repeatable and rules-based: identify the merchant, check the stack, look up attendees, scan the news. A workflow job.
Writing the SOW. Scoping required real judgment about the project, so it was a fit for Claude. But the judgment rested on rules that could be written down. Encode the rules first, then let the model work inside them.
Negotiation rounds. Version tracking needed structure more than intelligence: a consistent changelog format attached to the opportunity in Salesforce.
Closed-lost. Routing by loss reason is pure logic, and Salesforce Flow already does it natively. No new infrastructure needed.
After delivery. The scoped-vs-delivered comparison needed clean, structured data from both sides first. That made it a second-phase build, sequenced behind the automations that generate the data.
The pattern held across all four teams in the assessment, but it was sharpest in sales: the most expensive person in the cycle was doing the most mechanical work in it.
What We Designed
Four sales systems, each placed where the discovery map said it belonged. Three are first-phase builds. The fourth depends on data the first three create.
1. SOW Generator with Version Tracking (Claude)
A Claude Project configured as the sales lead's personal SOW generator. Its permanent knowledge holds the documented scoping rules, archetype definitions for migrations, redesigns, and optimization work, and a curated library of 6 to 10 exemplar SOWs.
Every v2+ SOW ends with a structured changelog block: client, version, what changed, why, and the date. That block is pasted into a dedicated Salesforce field, so anyone opening the opportunity sees how the scope evolved. The rollout pilots on three to five live SOWs before the format is locked.
2. Prospect Brief Automation (n8n)
An n8n workflow triggered by the sales calendar. It pulls the merchant name and URL from the invite, identifies the tech stack with Wappalyzer, and compares it against Vaan's recommended stack. It retrieves attendee profiles through Apollo and checks for recent funding, acquisition, and leadership news.
The finished brief lands in Slack 45 minutes before the call. Internal meetings, existing-client calls, and personal events are filtered out, so the brief only shows up when there's a prospect on the other end.
3. Closed-Lost Re-engagement (Salesforce Flow)
Built in native Salesforce Flow, with no added infrastructure. When a deal moves to Closed-Lost, the Loss Reason field picks one of three paths:
Timing: 30, 60, and 90-day check-ins, mixing automated emails with rep tasks.
Budget: a low-touch monthly nurture, with a rep review flag every six months.
Bad fit: suppressed from outreach, with a note logged on the account for future reps.
Email templates for each path are approved by the sales lead before launch.
4. Scoped-vs-Delivered Feedback Loop (Phase 2)
Connects the SOW Generator to Harvest time data from the delivery side. It compares estimated hours to delivered hours by archetype, tech stack, and scope component. Patterns like the beauty-migration overrun feed back into the exemplar library, so the next SOW starts from calibrated numbers instead of last quarter's assumptions.
This is the piece that turns SOW drafting from a speed problem into an accuracy system. It comes second because it needs a run of clean SOW and delivery data to work with.
Projected Impact
The sales workstream is projected to return about 70 senior hours a month at current SOW volume. These are targets from the assessment, baselined against discovery findings, and are validated against production data once each system is live.
| Area | Today | Target |
|---|---|---|
| SOW drafting time | ~8 hours per SOW | ~2 hours in month 1, ~1 hour by month 3 |
| Senior sales capacity | ~80 hours/month on SOWs | ~70 hours/month returned at 10 SOWs/month |
| Pre-call research | ~1 hour per discovery call | Brief delivered in Slack 45 minutes before the call |
| Scope version history | None | Changelog on every v2+ SOW in Salesforce |
| Closed-lost follow-up | No path by loss reason | Three routed paths: timing, budget, bad fit |
Each system ships with a defined metric, baselined in week 1 of its build. For the SOW Generator, that metric is hours per SOW, starting from 8.
Client Quote
"VAAN has experienced rapid growth over the past 12 months, reaching a point where our old ways of working no longer served us or our clients as effectively as we desired. We engaged Richard and Kris because of their reputation in optimizing agency workflows via AI tooling. They were tremendously helpful in diagnosing the specific pain points various departments and functions at VAAN were experiencing and how thoughtful automated solutions could streamline them."
Chuong Nguyen-Thanh, COO, Vaan Group
Looking Ahead
The sales systems are the front end of a larger loop. Once SOW, delivery, and pipeline data are structured, Vaan's leadership can ask questions that used to take a spreadsheet and an afternoon: "Are we consistently underscoping Magento migrations?" or "What did we discuss with this client about warranty?"
The roadmap sequences that work so each stage produces the data the next one depends on. Sales goes first because it pays for itself fastest and feeds everything downstream.
The roadmap is Vaan's. The next move is theirs to time.