A two-month Salesforce rebuild, a new commercial model, and an agent layer across the stack
Turning the CRM from a database into the operating system for revenue: tens of tools wired together, agents doing the busywork, and a refreshed experience for every prospect.
VRIFYVP MarketingMay 2025 – Present4 min read- 2 mo
- HubSpot → Salesforce
- 0
- Operational blockers at cutover
- 4x+
- Thoughtful outreach capacity per month
- 90%+
- Newsletter open rate
- 40%+
- Newsletter click-through rate
Context
VRIFY ran on HubSpot, and it had become the ceiling. We were moving to a new commercial model that needed custom objects, fields and artifacts. Our stack (Gong, Vitally, Webflow, Amplitude and a handful of others) needed custom touchpoints and automated events and follow-ups built on what buyers and customers actually did.
The bigger goal was to make the CRM the place where work happens, not just where it gets logged.
The thesisEvery hour a rep spends on research, follow-ups and data entry is an hour not spent with a customer. The CRM should do that work, and people should do the selling.
How the migration ran
I led the program with an internal team of 4 and an agency team of 9, each building different parts of the CRM. I owned the roadmap, data handling and hygiene, the CRM design and wireframes, and every automation and integration.
- 01DesignRoadmap, data model and wireframes for the new commercial model: custom objects, fields and artifacts.
- 02BuildObjects, automations and integrations across the stack, including agreement generation in DocuSign parsed through to Xero for billing.
- 03Parallel runSales and customer success pressure-tested the new CRM daily. A bug board fed biweekly product and sprint kickoffs.
- 04CutoverA final bulk import of every record created in HubSpot after the initial migration, on the official swap date. Zero operational blockers.
The risk in any CRM migration is that the business stops while the plumbing changes, and the plan was built around making sure that never happened. Design came first, because the new commercial model required objects and fields HubSpot couldn’t support, and copying the old structure into Salesforce would have wasted the move.
The parallel run was the safety net. For a period both systems were live while sales and customer success used Salesforce for real work and logged every problem to a shared bug board. On cutover day, a final bulk import brought across anything created in HubSpot since the initial migration, so nothing was lost in the switch.
The architecture
Salesforce at the center, with every tool feeding it events and pulling context back out.
MCP servers connect Claude to Salesforce, Vitally and Slack, so sales and marketing can manage accounts by prompting instead of clicking. Confidential data is excluded from these pulls.
The diagram shows the principle behind the rebuild: one system of record, with every tool both feeding it and drawing from it. Product usage from Amplitude, customer health from Vitally, conversations from Gong and inbound from Webflow all land in Salesforce, so any automation or person can see the whole account in one place.
The most forward-looking piece is the MCP connection. With Claude connected to Salesforce, Vitally and Slack, the team can ask for what they need in plain language, such as summarizing an account and drafting the follow-up, instead of clicking through records. Confidential data is excluded from what the model can access.
The agent and automation layer
What was built after cutover, and what each piece takes off the team’s plate.
| System | Trigger | What it does |
|---|---|---|
| SDR agents | New target account or engaged contact | Sources and enriches leads in Clay and drafts personalized first-touch email |
| Gong enablement agent | Every recorded sales call | Turns the transcript into enablement material and follow-up actions on the lead’s Salesforce record |
| Sales deck generator | Form fill by a rep | Builds a tailored deck with logos, mineral types and nearby examples |
| Agreement-to-billing | Deal reaches contract | Generates the agreement in DocuSign and passes it to Xero for billing |
| Conference lead engine | Lead met at an event | Tags, tracks and schedules follow-ups automatically |
| Campaign monitor | Channel performance shifts | Texts me when a campaign should scale up or down, with copy and targeting recommendations |
| Competitive intelligence | Daily crawl | Updates battlecards and flags competitor positioning changes against our ICPs |
| Press release agent | Daily industry news | Splits releases into prospects and clients: outreach and a briefing for one, a congratulations workflow for the other |
| Programmatic ICP pages | Visit to an ICP page | Routes the inbound to Salesforce with an information packet attached |
Each of these automations removes a task that used to sit on someone’s list: researching an account, writing a first email, turning call notes into action items, building a deck, chasing a contract into billing, or reading the day’s industry news. Individually they save minutes; together they returned hundreds of hours a month to the team.
The common thread is that each one is triggered by something that already happens, like a call ending, a press release going out or a lead being tagged at an event, so nobody has to remember to run them.
Results
The point was never speed to lead. It was being able to reach far more of the right people thoughtfully, without broad template messaging that gets deleted.
Hundreds of hours a month returned to the team. Newsletter open rates above 90% with click-through above 40%, and first-touch open rates above 25%.
The goal was never to send more email; it was to reach more of the right people thoughtfully. Personalized outreach capacity grew more than fourfold, and engagement stayed high: newsletter open rates above 90%, click-through above 40%, and first-touch open rates above 25%. Numbers like these only hold when messages are relevant to the person receiving them.
The operating rhythm paid off on the sales side too. With clear stage gates and one shared scorecard, sales cycles shortened by 25 days and forecast accuracy reached 80%. July, the first month of Q3, delivered the biggest first-month sales lift of any quarter.
Design the CRM around the work you want people to stop doing. Migrations fail when they copy the old system; they succeed when they encode the new operating model.