Retiring Pivotal CRM without stopping the business.

A global workspace operator still ran its sales operation through Pivotal CRM, with years of automation nobody had fully written down. We moved every automated flow onto new .NET services, one at a time and proven against production, so the Pivotal application can be switched off in December 2026.

~10k Pivotal agents read and documented
50+ agents rebuilt as tested .NET services
100k+ messages a day on the new services
18 days from the start of the work to the first flow live

Nearly ten thousand agents, and nobody sure which ones mattered.

Pivotal CRM keeps its automation in agents: flowcharts saved inside the database in a format nothing but the platform itself can read. Over many years this operator's estate had grown to nearly ten thousand of them, alongside scheduled jobs and server-side rules. Most of the people who built them had moved on, and the lists of what actually ran disagreed with each other.

Pivotal still sat in the middle of the sales operation. Leads, accounts, contacts, orders and tasks flowed into it from newer systems all day, and the reporting estate read its tables directly. Whatever replaced it had to keep every one of those flows running, and keep the data in exactly the shape reporting expected.

  • Nearly 10,000 agents stored in a format no one could read outside the platform
  • Scheduled jobs and rules that had never been fully documented
  • Reporting and downstream systems reading Pivotal's tables directly
  • Tens of thousands of transactions a day that couldn't stop for a big-bang cutover

Read what actually runs, rebuild that, and prove it against production.

Our own tooling read the agents straight out of the database and turned each one into a flowchart and a written specification, with every statement traced back to its source and checked by a senior specialist. We took the answer to "what really runs" from the scheduler itself rather than from anyone's list. That cut the scope from nearly ten thousand agents to the fifty or so that run unattended and write data the business depends on. Everything a user triggered by hand retires with the platform.

Each of those flows was rebuilt as a tested .NET service on the operator's own servers, writing the same tables in the same shape, so reporting didn't need to change. Before every switch, real production traffic was replayed through the new code and every difference was explained. The pass mark was zero unexplained differences: the lead flow matched across 7,100 real messages.

Matching a legacy platform means matching its quirks. One pricing rule only lined up once we reproduced exactly how Pivotal rounded a number internally. After that, tens of thousands of contract prices matched to the penny.

Cutover went one flow at a time: stop the old job, start the new service, never both writing. Each flow began with a pilot of a handful of messages, then larger rounds, with a switch back to the old job ready if anything looked wrong. The first flow went live 18 days after the work began, and the rest followed over the next three weeks.

Pivotal CRM .NET SQL Server Azure Service Bus Windows Services Production Replay Automated Testing Phased Cutover

Every automated flow off Pivotal, one at a time.

Every automated flow now runs on the new services, handling more than 100,000 messages a day. The old integration jobs and the scheduled jobs behind them are switched off, and the Pivotal application itself is switched off in December 2026.

Along the way the work found problems nobody knew were there. One kind of update had silently stopped months earlier, and is now fixed. Data-fix jobs that existed only because nobody could find the cause are no longer needed, because the cause was fixed in the rebuild. When a backlog built up after one go-live, two missing indexes on a table of tens of millions of rows took a lookup from over ten seconds to under five milliseconds, and the backlog cleared the same day.

Off Pivotal one flow at a time: 50+ agents rebuilt, more than 100,000 messages a day, and no big-bang weekend.