Short answer: IT general controls are examined under four headings — who can get into the system and what they can do, how changes to it are approved, how new systems are implemented, and how it is operated day to day. If those are weak, the automated controls inside the system and the reports it produces cannot be relied on, so the auditor has to fall back to substantive testing.
If your auditor has started asking for user lists, change approvals and backup logs, this is what sits behind those requests.
Why these controls come first
There is a dependency chain that explains the whole exercise.
Accounting records are produced by a system. Inside that system sit application controls — an approval limit, a three-way match, a duplicate-invoice check. An auditor would like to rely on those, because a configured control that runs on every transaction is stronger evidence than testing a sample by hand.
But a configured control is only as good as the controls stopping someone from reconfiguring it. If a user can quietly raise their own approval limit, or edit a posted entry without trace, then the automated control proves nothing. That is what IT general controls establish: whether the environment around the application is trustworthy enough for anything inside it to be relied on.
Get this wrong and the consequence is not a minor observation. The auditor cannot rely on system-generated evidence, testing expands substantially, and the audit becomes slower and more intrusive.
The four domains
| Domain | The question | Typical evidence requested |
|---|---|---|
| Access to programs and data | Who can get in, and what can they do once in? | User list with roles, joiner/leaver records, privileged-account list, password settings, periodic access review |
| Change management | How do changes to the system get approved and deployed? | Change register, approvals, test evidence, separation of who develops and who deploys |
| Program development | How was a new system or module implemented? | Project approval, user acceptance testing, data-migration reconciliation, go-live sign-off |
| Computer operations | Is it run reliably day to day? | Job schedules and failure handling, backup logs, restoration test evidence, incident records |
Access to programs and data
Usually the largest source of findings, and the one management can fix fastest.
What is examined: whether every user has a named account rather than a shared one; whether joiners are provisioned on approval and — far more commonly missed — whether leavers are actually deprovisioned; who holds administrator or superuser rights and why; whether password and session settings are enforced by the system rather than by policy; and whether someone reviews the user list periodically instead of it accumulating for years.
Underneath all of it sits segregation of duties: whether the same person can create a vendor and approve a payment to it, or post an entry and approve it. Segregation on an org chart means nothing if the system permits both actions under one login.
The recurring finding in mid-size Indian companies is a shared administrator account, often known to the accounts team and the software vendor alike. It cannot be attributed to a person, so it undermines both segregation of duties and the evidentiary value of the audit trail. It is also usually fixed within a week by issuing named accounts.
Change management
What is examined: whether changes and customisations are requested and approved before being made, whether they are tested before going to production, and whether the person who writes a change is different from the person who deploys it.
Smaller companies frequently have no change register at all — the ERP partner makes changes over email or a call. That is a genuine weakness rather than a formality, because an undocumented change to a tax rate, an approval matrix or a posting rule can misstate accounts for months before anyone notices.
Program development
Relevant in a year where a new system went live or a major module was added. What is examined: whether the implementation was approved, whether users tested it before go-live, and — most importantly for audit — whether opening balances migrated from the old system were reconciled to it. Migration reconciliation is where the real audit risk sits, and where evidence is often thin.
Computer operations
What is examined: whether scheduled jobs run and failures are noticed, whether backups run and have actually been restored in a test, and whether incidents are recorded and resolved.
Backups deserve a specific note. Almost every company has backups. A materially smaller number has ever restored one to confirm it works. An untested backup is an assumption, not a control.