Why legacy systems still run the back office
Legacy systems still run the back office because they work, they hold decades of business rules, and replacing them is risky. Order management on an AS/400 (IBM i), policy administration on a mainframe, a warehouse system on OpenVMS, payments on HP NonStop, an old web portal, a desktop application over Citrix: these remain the systems of record in many companies.
The cost is not the system itself. It is the work around it. Documents arrive by email, fax and SFTP, and someone has to read each one and key it into green screens or old forms, field by field. That work grows with volume, depends on a few experienced people, and creates typing errors that travel downstream.
What legacy system automation changes
Legacy system automation changes how data gets into the system, not the system. Modernization of the work happens without migration of the platform.
What stays and what changes
How to automate legacy systems without APIs
You automate a legacy system without APIs by working through the interface it already has: its screens. legacy bridge connects the way a terminal emulator or a browser does, so no new integration has to be built.
Documents arrive
By email, an SFTP/FTP folder or an API upload: PDF, scans, photos, faxes, Word and Excel. A file holding several records is split into one per record.
AI reads and checks
AI vision extracts the fields, no templates needed. Values are checked against the document, your plain-English process rules and what the system holds. Duplicates are caught.
The agent keys it in
It signs on, navigates the menus, fills the fields, handles lookups and function keys, and reads the system's own messages. By default a person approves each transaction before it is submitted.
It confirms the record
It reads the record back from the system and logs the result. Every keystroke and screen is recorded.
This is where it differs from a document AI tool on its own. Extraction tools stop at a data file someone still has to get into the system of record. legacy bridge does that last mile too: the data entry into the legacy system, with confirmation.
Legacy system modernization: four approaches compared
Legacy system modernization usually means one of three projects, and each leaves the daily retyping in place for months or years.
| Criterion | legacy bridge | Replace, API project or RPA |
|---|---|---|
| Time to a first result | A controlled proof on one real document and your own screens | Replace: a multi-year program API project: months of design and development RPA: weeks of scripting per workflow |
| Changes to the legacy system | None: no host software, no program changes, no database access | Replace: the whole system and its data move API project: new code and services on the host RPA: none, but bots on desktops or VMs |
| Reading the documents | Built in: AI vision, checks and duplicate detection | Not included in any of the three; a separate OCR or IDP tool and its integration |
| When a screen changes | Screens are read by meaning, not by position | RPA: recorded scripts break and need a developer API project: the interface must be updated too |
| Who maintains it | A finished product; your people's corrections teach it | Your developers, or a consultancy, for the life of the integration |
| Risk to daily operations | Low: a person approves, current work never stops | Replace: cut-over and data migration risk API project / RPA: new failure points in the path |
Legacy modernization by replacement
Rip-and-replace or re-platforming retires the old system. It can be the right long-term call, but the retyping continues until the last module goes live.
A legacy system integration project
New APIs need developers who know the host, and they are scarce. Even with an API, custom code is still needed to turn a PDF into a transaction.
Traditional RPA
RPA bots replay recorded clicks and keystrokes. They break when a screen shifts and leave document reading to another tool. See how AI agents compare with RPA.
Which legacy systems it works with
legacy bridge runs in production today on IBM i (AS/400). The same engine supports mainframe, VT, HP 6530, web and Citrix screens, and each of those is scoped per workflow with the same controlled proof.
IBM i (AS/400, iSeries)
In production. Custom RPG and COBOL applications and packaged ERPs, over TN5250. See TN5250 automation.
IBM Z mainframes
Supported over TN3270: z/OS with CICS, IMS and TSO screens. See mainframe automation.
OpenVMS, Unix, Pick
Supported over VT terminal protocols.
HP NonStop (Tandem)
Supported over HP 6530.
Legacy web portals
Supported: the agent drives the portal in a browser.
Citrix and RDP desktop apps
Supported: vision-driven, reading the published application the way a person sees it.
Hosts that are not exposed to the internet are reached through a secure SSH tunnel. Nothing is installed on the host and there is no database access.
Legacy workflow automation examples
The best legacy workflow automation candidates are documents that arrive every day and are keyed into the same screens every time.
Invoice processing
Supplier invoices checked and keyed into accounts payable.
Purchase orders
Orders entered line by line, lookups handled on screen.
Sales order entry
Emailed and PDF orders keyed in, duplicates caught.
Insurance claims
Claim forms checked against your rules and keyed in.
Medical and dental claims
Claims keyed with HIPAA-aware processing.
Bills of lading and delivery notes
Shipping paperwork entered as it arrives.
Who it is for
Back office automation for operations
Operations and finance teams whose people spend the day keying documents into old screens.
Legacy application automation IT can approve
IT teams that cannot take on another integration project or accept software on the host.
Leaders planning modernization
Teams that want results now while a migration is still being decided. Read about modernization without migration.
Control, security and audit
Your team stays in control: nothing is guessed, nothing is submitted without the rules you set, and every action is recorded.
- Approval first. By default every transaction waits for an employee to approve it in a web console or from an email link. A trusted workflow can run on its own, with people handling exceptions.
- Flagged, never guessed. An unclear, missing or hand-corrected value holds the record with the exact field and the document image for a person to decide.
- Host errors are shown. If the system returns an error, the employee sees the message and the screen, and nothing is submitted.
- Tamper-evident audit trail. Every keystroke and screen is recorded in a SHA-256 hash chain; sessions can be replayed.
- Certified company. legacy bridge is part of scandiweb, certified ISO 9001, ISO/IEC 27001 and ISO/IEC 27017, running on PCI DSS-compliant infrastructure and serving 500+ clients, including Fortune 500 companies.
