Jabil's Integration-First Playbook: Why 100+ Global Sites Simplified Before Automating

If you've ever inherited a codebase where every team built its own workaround for the same problem, you already understand Jabil's core challenge. Just scale it up to more than 100 manufacturing sites across 30-plus countries, add 25 years of accumulated technical debt, and mix in local regulatory requirements. That's the environment Harish Manohar, SAP IT director at Jabil, has spent recent years trying to untangle.

Jabil is a global manufacturing company headquartered in St. Petersburg, Florida, with roughly 140,000 employees serving over 400 of the world's top brands. Like a lot of large enterprises that grew through decades of regional autonomy, each site ended up with its own tools, its own spreadsheets, and its own manual workarounds for processes that, in theory, should look identical from plant to plant. The result: data silos that made it hard to spot problems early or coordinate a response across the network.

Manohar's team settled on a principle that sounds simple but is genuinely hard to enforce at scale: simplify first, then innovate. "Any innovation without simplification is going to add more complexity," he says. It's the enterprise version of "don't add features to a mess, clean up the mess first."

The Architecture Behind the Simplification

The technical backbone of this effort is SAP's Integration Suite, running on SAP's Business Technology Platform (BTP), which Jabil is positioning as the central hub for connecting systems rather than letting integration platforms multiply site by site. A few architectural decisions stand out:

• API-driven, event-based integration as the default pattern. Instead of batch-moving data between systems, Jabil standardized on APIs and events, treating that as a hard best practice rather than a preference. Manohar is blunt about the alternative: constantly shuttling data from one place to another is "not good business practice in the IT world."
• Reuse over one-off tooling. When a specific site needs a capability, the team's default move is to check whether another site already solved it, then adapt that solution rather than buying a new point tool. Best-of-breed procurement still happens for major processes, but site-level needs get standardized wherever possible.
• A clean-core approach to SAP customization. After 25 years of heavy, customer-driven customization, Jabil is migrating systems into RISE with SAP, which is designed to support a cleaner core going forward. That means new governance and review processes that push back against the kind of deep customization that made past upgrades painful.
• Signavio for process standardization. Jabil is using SAP Signavio, a business process management tool, to help standardize workflows across plants that vary widely in maturity, legacy systems, and local regulatory context.

The clean-core piece is worth pausing on if you've worked in any large enterprise SAP environment. Heavy customization is often what happens when a system has to satisfy every customer's specific demand over decades. It works in the moment but it compounds: every future upgrade, integration, or migration has to account for years of one-off modifications. Jabil is essentially paying down that debt now, betting that a cleaner core makes every future integration and AI initiative faster to deploy.

Why start with integration specifically, rather than jumping straight to AI or automation, which is where a lot of the industry attention (and budget) currently sits? Manohar's answer is architectural: "the backbone of any contemporary or modern organization is data." Automation and AI models are only as useful as the data feeding them. If that data is fragmented across site-specific systems, any AI layered on top inherits the same fragmentation, just with more expensive tooling and less transparency into why it's failing.

Jabil frames this as building a single system of record, though Manohar is careful to note that doesn't mean everything lives in SAP. It means building consistent data pipelines across the systems that actually run operations: supply chain, planning, inventory, and the rest. Modernization, in this view, isn't about upgrading software for its own sake. "Any modernization or transformation should add measurable business value," Manohar says, and that value gets measured against end-to-end process connectivity across the supply chain, not feature checklists.

The rollout hasn't been uniform, and Jabil doesn't pretend otherwise. Sites differ in process maturity, legacy system age, and local workarounds, and regulated businesses add qualification and compliance requirements on top of all that. Standardization is meant to produce consistent workflows, data flows, and governance across sites, though Manohar is clear the company isn't at 100% yet. The near-term target is scaling shared processes across more than 40 plants, moving away from a site-by-site operating model toward something more consistent region to region.

One practical payoff of standardization: faster rollout of new technology. Previously, deploying something like BTP or Signavio at a new site meant accounting for heavy local customization first, which slowed everything down. Strip out that customization burden, and new tools move faster because they're landing on a more predictable foundation rather than a bespoke one.

Key Takeaways

The pattern here maps onto a lesson that shows up constantly in software engineering, just at industrial scale: integration and simplification aren't preludes to the interesting work, they are the interesting work. Jabil is explicitly building toward predictive supply chain insights, intelligent exception handling, and AI-driven planning and forecasting, but only after establishing trusted data flows and integrated systems as the foundation. For teams anywhere weighing whether to bolt AI onto existing infrastructure or fix the infrastructure first, Jabil's bet is instructive: "simplicity at scale is a very competitive advantage," as Manohar puts it, and that advantage compounds well before the AI even shows up.