Blog/Modernization

Modernization

Recording behavior nobody wrote down

Vijay SubramanianCo-founder, EnWithAI

3 min read

Talk to us

Where each rule’s evidence comes from

  1. CodeWhat the system can do
  2. RecordingsWhat it does, on real inputs
  3. Scheduled runsMonth-end, year-end and batch jobs
  4. HistoryTickets, audits and old records
  5. Needs evidenceNever assumed. Marked and chased

A behavior without evidence is a question, not a rule.

Before you rebuild a system, you need a list of what it does. Not what the documents say, and not what the team remembers. What it does, with evidence. That list almost never exists, so the first job is to make it.

The short answer

Record the legacy system in a test environment, replaying realistic requests and scheduled jobs, and write each rule down as a behavior with its evidence. Rules that never appear in recordings are marked as needing evidence and confirmed from historical records or business scenarios.

Why reading the code is not enough

Code tells you what a system can do. It does not tell you which paths matter, which ones run every day and which ones nobody has hit since 2011. A rule buried in a branch that fires twice a year looks exactly like dead code, until the month you remove it.

So we pair the code with recordings: the system running in a test environment, on realistic inputs, with every request and its result captured.

What to record

  1. Requests and responses

    Replayed or realistic traffic through the system’s interfaces, with the outputs it produced.

  2. Scheduled runs

    Batch jobs, month-end and year-end runs, run in the test environment on seeded data. This is where many forgotten rules live.

  3. Data changes

    What each run wrote to the database, so a rule that changes a record and returns nothing is still caught.

  4. Edge cases on purpose

    Inputs chosen to cross known limits: amounts at the cap, dates at period end, the largest customer.

Data handling comes first. Where recordings are kept, what is masked and for how long is agreed before anything is captured.

Turning recordings into behaviors

A recording is raw evidence. A behavior is the rule it shows, in a sentence the business owner can read, linked to the runs and the code that prove it.

BHV-0211

A payout never exceeds the policy’s payout cap.
Evidence: 41 recorded claims at or above the cap, and line 3 of PAYOUT-CALC.

Each behavior gets an id, a plain sentence, its evidence and a criticality. That inventory is what the spec is built from, and what the rebuild is later tested against.

When a rule never shows up

Some rules will not appear in any recording, because the conditions never came up in the window you captured. A reopened claim from three years ago. A leap-year adjustment. A customer category that no longer exists but still has open accounts.

Never count those as evidenced by assumption. Mark them as needing evidence and go and find it: historical records, an old ticket, a business scenario the owner walks you through, or a targeted run built to trigger the condition.

The rule you could not record is the one most likely to be lost.

How much is enough

You will not reach every rule. What matters is knowing which ones you have evidence for and which you do not, and letting the business owner decide what to do about the gaps before the rebuild starts, not after go-live.

In CrewPE

Discovery reads the code and records the running system in a test environment. Each rule becomes a behavior with its evidence, and rules without evidence are marked, never assumed.

See how modernization works

Questions teams ask

Can you capture legacy behavior without production access?
Yes. Discovery can start from the code and a test environment with seeded data. Gaps that only production would reveal are recorded and closed before any migration decision.
What if a business rule never appears in recorded traffic?
Mark it as needing evidence and confirm it another way: historical records, old tickets, business scenarios from the owner, or a run built to trigger it. It should never count as evidenced by assumption.
Is recording a legacy system safe for sensitive data?
It can be, if data handling is agreed first: where recordings are stored, which fields are masked and how long they are kept, decided before capture begins.

Vijay Subramanian

Co-founder of EnWithAI, working on modernization delivery: recording what a system does and proving the rebuild does the same.

Keep reading

Bring one legacy application. See what it does that nobody wrote down.

Start with a 45-minute call about your system: how it is built, where its business rules live, and a sensible first step.