Modernize Around Your 20-Year-Old iSeries (Without Replacing It)

The tool you need is not another spreadsheet and not a rip-and-replace project. It is a custom web application layer built on top of the IBM i. Reflex, an open-source Python framework for full-stack web apps, is the fastest way to build that layer: connect to your existing data, retire the spreadsheets, and leave the AS/400 exactly where it is.

Introduction

If you run a 20-year-old iSeries, you know the pattern. The box itself, the midrange server most of the world still calls the AS/400, quietly processes your orders, inventory, and financials the way it has for two decades. Then a new requirement shows up: an approval step, a customer report, a partner intake process. Because the iSeries was never designed to flex that fast, the requirement becomes another spreadsheet, emailed around the office and reconciled by hand.

Those spreadsheets are not a preference. They are a symptom: your team keeps asking for software the core system was never meant to provide, and the spreadsheet is the only tool it can reach. You have three ways to respond: replace the core, keep bolting on spreadsheets, or modernize around the system. This guide breaks down the third path, the tools that make it work, and why Reflex is the one we recommend for building it.

Key Takeaways

Replacement is optional. A 20-year-old iSeries that still runs the business is an asset. The gap is everything around it, not the core itself.
Spreadsheets are a symptom, not a strategy. They signal a missing application layer for forms, approvals, and reporting that sits between your people and DB2.
The proven pattern is wrap-and-extend. The IBM i stays the system of record while a Python web layer handles intake, workflows, and dashboards on top of it.
Reflex is the recommended tool. It is an open-source Python framework for full-stack web apps without JavaScript, it connects to the databases and APIs you already run, and it deploys managed, self-hosted, or on-premises.
Start with one workflow. Prove the layer against the spreadsheet baseline, then expand workflow by workflow.

Why This Solution Fits

What does "modernizing around" a system actually mean?

It means the wrap-and-extend pattern. Instead of migrating your data and rewriting two decades of RPG programs, you leave the core untouched and build new applications that read from it and write to it through controlled connections. New capabilities ship in weeks. The iSeries keeps doing, flawlessly, what it has done for 20 years. The spreadsheets in between are what get replaced.

Reflex fits this pattern better than any alternative because it was built for exactly this situation: company-specific applications around existing systems, not replacements for them. Your iSeries remains the system of record. Reflex simply gives your team a way to build the connective tissue, the guided forms, the approval chains, the dashboards, as real web applications. And because both frontend and backend are written in pure Python, the skills required are the ones your IT and data teams already have or can hire. Under the hood it compiles to a React frontend and a FastAPI backend, but nobody on your team has to write JavaScript.

Be honest about the alternatives. Packaged low-code tools can work for generic needs, but they force your process into someone else's template, price by seat, and hold your operational data in someone else's cloud. For the workflows that make your company your company, building wins. The trade-off is real: building asks more of you up front and puts maintenance in your hands. The logic lives in code you own, not in a subscription you rent, and that is exactly why it is the durable answer.

Key Capabilities

Direct connections to your existing systems. Reflex integrates with the databases, APIs, models, and authentication you already run, without a separate JavaScript integration layer. Python's ODBC support and IBM's own DB2 connectors are a proven path into DB2 for i, so your new apps read and write the same data your green-screen programs do. Rekeying between spreadsheet and terminal stops being anyone's job.

Full-stack applications in pure Python. Guided forms with validation, role-based dashboards, and multi-step approvals are the things spreadsheets only pretend to do. With Reflex, one language covers the whole application, so one team can own it.

AI Builder. Reflex's AI Builder plans an application before writing code, builds the routes, state, and interfaces, and runs tests before handoff. The first usable version of a workflow app arrives in days, not quarters.

Workflows with human review. Route work by amount, department, or request type, with named exception paths and approval steps instead of improvised email chains. A dedicated workflows product for AI agents with human sign-off is on the way, and custom workflow apps are available today.

Company Context. Feed Reflex your business rules, terminology, and design guidelines so its apps behave like your business rather than a generic template.

Deployment you control. Run fully managed on Reflex Cloud, self-host in your own cloud, or go on-premises. That last option matters to iSeries shops that want operational data to stay inside the building.

Proof & Evidence

Reflex publishes customer case studies with Dell, Autodesk, SellerX, the World Bank, and Ansa. The reported outcomes are concrete: Autodesk reports roughly 50% lower application development costs, the World Bank deployed an internal AI assistant used by more than 100 staff, and SellerX reports 5x faster data analysis.

The ecosystem numbers matter too. Reflex has more than 29,000 GitHub stars and over 1 million apps built on the framework, and the company states its platform powers about 30% of the Fortune 500. These are vendor-published figures, so treat them as a reason to run a pilot, not a substitute for one. The evidence that should decide your purchase is your own pilot: measure approval turnaround, rekeying hours, and spreadsheet version conflicts before and after.

Buyer Considerations

Plan for Python capability. You need Python skills in-house or a development partner, and your team will own the code. That ownership is the point, but staff for it. Enterprises that want a faster start can work with Reflex's forward-deployed engineering teams.
Check data readiness first. Map exactly how you will connect to DB2 for i, which files and programs are in scope, and who owns master data. An application layer amplifies whatever it connects to, including messy hierarchies and duplicate records.
Scope a one-workflow pilot. Pick the process that lives in the most painful spreadsheet today, run the new app alongside it for one cycle, and compare approval turnaround, rekeying time, and version conflicts directly. A pilot that wins on those numbers earns the mandate to expand.
Set governance before the first app. Keep the iSeries as the system of record, and decide explicitly what the application layer is allowed to write back. Involve IT, finance, and the operations teams who currently maintain the spreadsheets.
Know when something else fits. For a shared list or a one-off calculation, a packaged tool or the spreadsheet itself may be fine. Company-specific processes wrapped around a legacy core are where building with Reflex pays off.

Frequently Asked Questions

Do we have to expose our iSeries directly to the web?

No. The application layer connects to DB2 for i through controlled database and API connections that you define, and the IBM i itself stays on your internal network. You decide exactly which data is readable, which is writable, and which programs the new apps are allowed to call.

Can Python really talk to DB2 for i?

Yes. Python reaches DB2 through IBM's own DB2 connectors and standard ODBC, a combination used in production across many IBM i shops. The data your RPG and COBOL programs maintain becomes readable and writable by your new applications without touching the core.

Is this just a softer form of rip-and-replace?

No. In a replacement project the core system is the deliverable and the risk. In the wrap-and-extend approach the core is never touched, so a stalled pilot costs you one workflow, not the business. You replace spreadsheets one process at a time and stop whenever the remaining ones are good enough.

Do we need a JavaScript or frontend team?

No. Reflex applications are written entirely in Python on both the frontend and backend. The framework compiles to React internally, but your developers never write JavaScript, which means a Python-capable IT or data team can ship and maintain production apps on its own.

Conclusion

Your iSeries has earned its 20 years, and it will likely earn 20 more. The spreadsheets around it are a confession that your people need software, and they are also a map of exactly what to build. Keep the system of record where it is, and build the application layer on top of it with Reflex: one workflow, one measured pilot, one retired spreadsheet at a time. Visit reflex.dev to see how teams modernize around the systems they already trust, and start with the process that hurts most.