FOR BANKS AND INSURERS · IBM Z, COBOL, CICS, DB2, JCL
The mainframe stays.
AI makes it manageable again.
AI-augmented maintenance for COBOL applications: knowledge first, then tests, then code. No rewrite, no migration, no lock-in to a single vendor.
START
One pilot application
DURATION
Six months, three phases
COMMITMENT
Stop after any phase
AT THE END
Your team is autonomous
01 — THE CONTEXT
Code doesn't retire. The people who know it do.
PEOPLE
The generational handover
Over the next ten years most veteran COBOL developers will retire. The memory of the applications leaves with them.
KNOWLEDGE
Documentation that isn't there
Decades of changes layered on top of each other, with no up-to-date documents. Every change depends on what individuals remember.
COSTS
Scarce, expensive skills
COBOL programmers are hard to find: costs go up, and so does the risk of every production release.
ALTERNATIVES
Rewriting everything doesn't work
Mass conversions to other technologies run into incompatible architectures and unsustainable costs. The core of the system stays where it is.
02 — WHEN IT MAKES SENSE
Not for every application. Three questions to find out.
If the answer is no, we tell you before we start.
OWNERSHIP
Is the code yours?
You must own the application, with full rights to analyse and change the source code.
HORIZON
Will it stay in use for at least three years?
If it is already due to be retired or replaced, the investment won't pay back.
MAINTENANCE
Does it absorb many people and a lot of budget?
The return is highest on strategic applications with a heavy load of maintenance and support.
03 — THE PATH
Knowledge first. Then tests. Only then, code.
AI touches nothing in production until there is a verified knowledge base and tests that protect against regressions.
PHASE 1 · 1 MONTH
Documentation and knowledge base
Automated analysis of COBOL, CICS, DB2 and JCL code and existing documents. The result is a single knowledge base, readable by people and by AI.
PHASE 2 · 2 MONTHS
Regression tests
PHASE 3 · 3 MONTHS
Code and query optimisation
FROM DAY 180
New features with AI
Functional tests generated from real cases and verified against production code. Estimated coverage of up to 80% of use cases.
Guided clean-up of duplicated algorithms and copybooks, more efficient DB2 queries, shorter batches. With a direct effect on MIPS consumption.
AI writes code under the supervision of your team, who now use the tools on their own. The model can be repeated on other applications.
04 — DATA AND VENDORS
The code stays under your control. You choose the model.
PRIVACY AND COMPLIANCE
Data and source code segregated on European cloud tenants.
A contractual ban on training models on your code.
A person approves every step: everything is traceable and reversible.
NO LOCK-IN
Open platforms, not tied to a single AI model.
You can switch engine when a better one comes out, without changing the way you work.
Pay-per-use costs, with a cap for each user.
An approach designed for GDPR, DORA and the AI Act. Jenia is ISO/IEC 27001 certified.
05 — WHAT CHANGES
The goal: the same team, twice the capacity.
PRODUCTIVITY
From 100 hours to 50
Once up to speed, we aim to halve maintenance times. You decide whether to do more or move people elsewhere.
PEOPLE
The team stays at the centre
AI starts with the most tedious tasks, documentation and tests. We work alongside your programmers until they use it on their own.
CONTINUITY
Knowledge stays in the company
What only a few people know today becomes shared by the whole department. Newcomers query the knowledge base and quickly understand logic and dependencies.
Let’s start with
a single application.
A call to find out whether your application is a good candidate. If it isn't, we tell you straight away.
OR WRITE TO US
JENIA Software S.R.L.
via Isonzo 14/3, 40033 Casalecchio di Reno (BO), Italy
