How Repligen Built an AI-Driven Central Trade Compliance System Across Seven Acquired ERPs

Repligen is a global bioprocessing manufacturer serving biopharma and CDMO customers across more than thirty countries. Its growth strategy is unusual for the sector, one to two manufacturer acquisitions a year, layered on top of an already deep catalog of bioreactors, filtration systems, chromatography media, and analytics instruments. By the time Repligen approached GingerControl, the actively imported catalog had grown past sixteen thousand SKUs.

Every acquired company arrived with its own ERP. Some had been built in-house years ago, others were standard packages configured by a long-departed integrator. Seven distinct ERPs were live in production, each holding part of the truth about what Repligen actually sold, where it was made, and how it was classified for customs. Pulling a single trade-compliance report meant pulling from seven systems, normalizing seven schemas, and reconciling fields that disagreed about basic facts.

Repligen’s brief to us was unusually direct. They did not want another reporting tool bolted onto the side. They wanted an AI-driven central trade compliance system that could read across every ERP, produce a single global view of classification, reporting, and audit, and absorb every future acquisition without rebuilding the system each time. The hardest part of the build, by a wide margin, was the data standardization that had to come first.

The engagement opened with a four-week assessment. The trade team walked us through a quarter of past reports and we walked them through where their hours were going. Three patterns emerged.

We price engagements the way we run them: scope first, fixed against a defined deliverable, no open hourly meter. A central trade-compliance system across seven ERPs is a large build, so the scope was frozen up front and the budget tracked it. Here is how the Repligen engagement was budgeted, sequenced, and delivered.

We scoped the engagement to a fixed deliverable, a central trade-compliance system reading across all seven ERPs into one canonical standard, with the global classification service on top, and priced it as a fixed fee paid against milestones, not an hourly meter. Repligen knew the number before the build began, and each payment tracked a milestone they could accept.

We did not boil the ocean. The MVP was the smallest deliverable that could produce a single, trustworthy cross-ERP compliance report from the canonical layer. Everything else was named and deferred, so scope could not creep across seven systems.

The year split into nine months of build and three of rollout. Development ran through the first four milestones, the data standard first because it was the hardest, then the unified layer, the first live report, and the classification service across all seven ERPs. The final three months were testing, training the trade team, and evaluating the system against the baseline before sign-off.

Every milestone had an acceptance test Repligen owned. The data standard had to be signed by the field owners. The MVP had to produce a quarterly report that matched a hand-built one. The classification service had to surface the right stale SKUs against a known USITC change. Payment followed acceptance, not the calendar.

The rollout was staged, not a big-bang. ERPs moved onto the layer in tranches, and the trade team kept its old process running until each tranche passed its acceptance test. A failed stage was fixable before the next ERP came on, which is the whole point of sequencing a seven-system migration this way.

In the twelve months after deployment, Repligen’s trade team stopped spending its quarters assembling reports and started spending them reviewing the contents. The annual HTS audit, previously treated as a multi-month project that required borrowed analyst capacity, finished inside two weeks with the standing team. Repligen has since absorbed two additional acquisitions through the new onboarding pattern, each connected to the unified layer in roughly six weeks, against the six months the pre-engagement process typically required.

What Repligen valued most was the operational shift. The trade team no longer functions as a reconciliation desk that happens to also handle classifications. The reconciliation tax was paid once at the data layer, so the team’s hours could go into the judgement work that classifications actually require.