IT general controls: the foundation every application control stands on
IT general controls decide whether you can rely on anything a system tells you. Weak ITGC does not fail on its own — it invalidates the controls above it.
Segregation of duties splits a transaction so that committing fraud or hiding an error requires collusion. Most breaches of it are accidental, and invisible until tested.
Segregation of duties is the principle that no single person should control a transaction from beginning to end. The four functions that must be kept apart are authorising it, executing it, holding the asset, and recording it. Where one person holds two or more, the error or the theft can be committed and concealed by the same hand.
The classic case is the person who can add a supplier, approve an invoice and release the payment. Nothing about that is unusual in a small finance team, and nothing about it is safe: the fictitious supplier and the payment to it are created and blessed by the same person, and the record they produce is the only record.
The last one is the most commonly missed, because it looks like IT rather than finance. Someone who can grant themselves rights can defeat every other separation on the list, which is why administrator access belongs on the same matrix as payment approval.
A matrix that shows no conflicts in a team of five is a matrix that has not been tested against actual system permissions. Small teams always have conflicts; the useful output is not a clean grid but a short list of accepted conflicts, each with a named compensating control and a person who reviews it.
Below a certain headcount, separation is arithmetic rather than policy — there are not enough people. The substitute is oversight by someone outside the process: the owner opens the bank statement themselves rather than reading a prepared summary; a director reviews new suppliers monthly; payments above a threshold need a second approver from outside finance.
These compensating controls only work if they are performed on the source, not on a report produced by the person being overseen. A summary written by the one person who does everything proves nothing about what everything was.
Because duties are enforced by whatever the system lets people do, the matrix and the permissions have to agree. Ettex Teams keeps roles and their permissions explicit per workspace, so the answer to "who can approve this" is a setting rather than a convention — and the periodic access control policy review compares the same list that the matrix was built from.
Authorisation, custody of the asset, record keeping, and reconciliation or review. Any one person holding two of these can both cause and conceal a problem.
A different control that reduces the risk left by a conflict that cannot be removed — for example, an owner reviewing every payment over a threshold where one person prepares and releases payments.
At least annually, and after any reorganisation, system change or departure. Conflicts appear silently when people change roles and keep their old access.
Not usually on its own, but it is an expectation of most control frameworks, is tested in financial audits, and is explicitly required by some regimes for listed companies and regulated firms.
IT general controls decide whether you can rely on anything a system tells you. Weak ITGC does not fail on its own — it invalidates the controls above it.
A remote joiner misses everything that is normally absorbed rather than taught — who to ask, how things really work, whether their pace is right. All of it has to become explicit or the first month is guesswork.
Most hybrid policies say three days in the office and stop. The questions that cause the arguments — which days, who decides, what happens to meetings — are the ones left out.