Legacy estate intelligence

You cannot modernize what nobody can describe.

Before you migrate, rewrite, or point AI at forty years of COBOL, somebody has to know exactly what is in there. That is what we do, and it is all we do.

See how an engagement works What we actually build

Why this matters more than it used to

Benchmark is the language server COBOL never got.

Every AI coding tool your teams have evaluated depends on a language server: a component that reads the code and tells the model what is actually true about it. Where a function is defined. What type a field really is. Everywhere a variable is written. The model finds candidates, and the language server states the facts.

COBOL never got one. There is no service on a mainframe estate that will answer what writes this field with any authority. So the one architecture the entire industry agrees on is missing exactly where the risk and the money are.

That is the gap we fill. A language model reading raw COBOL has to infer everything that requires seeing the whole system at once: which fields overlap the same bytes, what a copybook expands to after substitution, whether a number is packed decimal with two implied places or a float that will round differently on the millionth transaction. It infers quietly, and the defects surface late.

Give the same model a resolved description of the estate and those stop being inferences. They become stated facts, checkable against the original rather than taken on trust.

The full argument →

The name is the idea

A benchmark is a fixed point of known position

In surveying, a benchmark is a mark of established height. Every other measurement on the site is taken from it. Without one, you can still build, you just cannot be sure of anything you build.

Legacy estates almost never have one. The documentation drifted out of date a decade ago. The people who wrote the system have retired. The only truthful description of what the business actually does is the code itself, and nobody can read eight million lines of it fast enough to matter.

We establish the benchmark. We build one complete, precise, standards-based model of your estate: every program, job, transaction, file and database, the data flowing between them, and the business rules buried inside them. Then the questions that used to get shrugs start getting answers.

Everything the model covers

You own the result, and it is not in our format

The model we build is structured on ISO/IEC 19506, the Knowledge Discovery Meta-Model, an international standard published by the Object Management Group. We did not invent it and we do not control it.

That matters for a practical reason rather than a philosophical one. What you receive is a description of your own systems in a documented, published form. Your team can read it. Your migration partner can consume it. Another supplier could pick it up if you replaced us. There is no format lock-in, because there is no proprietary format.

Traceable to source

Every element in the model carries a reference back to the exact line of source it was derived from. In a regulated business, that is the difference between a finding and an opinion.

What we read
COBOL, JCL and CICS, together, as one system rather than three separate inventories.

Where it runs
From our transformation center, or on your own server over VPN if your code cannot leave the building.

What becomes possible once it exists

Every engagement below starts from the same model. Most clients begin with one and come back for others.

Start here

Estate assessment

What you have, how it hangs together, where the risk is concentrated, and what it would genuinely take to change it. The document you can fund a program from.

New

Migration readiness

The resolved facts an AI-assisted migration needs in order to be accurate, with traceability from every generated line back to the source it came from.

Recover

Business rule recovery

Decades of accumulated business logic pulled out of code that outlived its documentation and its authors, in a form people can read and systems can run.

Prove

Data lineage

Where data comes from, every operation performed on it in transit, and where it ends up. For compliance, impact assessment and governance.

Convert

Datacom conversion

A turnkey, highly automated route from CA Datacom/DB to DB2 or Oracle, including schema, SQL, call refactoring and JCL updates.

Move

Cloud porting

Unix to Linux in the cloud, with the analysis done before the migration starts rather than discovered during it.

Start with one application

Most engagements begin with a single system and a specific question. Tell us what is keeping you from moving, and we will tell you honestly whether we can help.

Email us