Understand. Authorize. Execute. Prove. Learn.
Five movements, in the order they actually happen. Every VERIDOM Worker runs the same shape, against the same operating model, whatever the workflow.
Establish what actually governs.
We start with one consequential workflow, not the whole company. ALEX ingests the policies, contracts, systems, documented procedures and the practised reality that never quite matches the documentation.
Then it does the part that matters: separating authoritative from outdated, conflicting, inferred and unsupported. Where two sources disagree, the disagreement is shown to you rather than settled by guesswork. Where authority has been assumed rather than established, that gap is named.
What comes out is not a report. It is a governed model, inspectable by a person and consumable by a machine. You own it.
Explore ALEX →Establish who may decide, before anything is proposed.
Authority is modelled explicitly: who may approve, escalate or override, under which conditions, up to what value, on which account. A job title is not authority. A verified binding is.
The check happens before the action, not in a log afterwards. Where authority cannot be established, nothing executes, and the refusal is recorded with its reason, at the same standing as an approval.
A person being present is not the same fact as a person being authorised.
Act inside the boundary. Nowhere else.
A Worker takes ownership of the job end to end. It establishes the case, gathers what is missing, resolves what it can against the rules that currently govern, and takes each case to a real ending rather than to a queue.
It operates within what the model has actually established. Not the whole organisation, just the part that has been made explicit. Anything outside that boundary stops and routes to a named person.
Explore Workers →Create the evidence while the work happens.
Was a person supposed to be involved? Did an accountable person with the right authority actually attach? Was execution permitted?
Those are recorded as three separate facts. Most systems collapse them into one log line, which is exactly how an organisation ends up unable to prove which of the three actually happened.
Reconstructing an audit trail afterwards is a different activity from producing one at the moment of decision, and only the second survives contact with a regulator.
Explore Trust →Keep what applies beyond this case. Leave the rest with it.
Something that turns out to govern the whole account, or the whole business, whether a customer’s portal rule, a contract term or an approval path, is recorded once and available to every future workflow. Something true only of this case stays attached to it.
That distinction matters. A fact from one case quietly becoming company truth is how a system turns one customer’s exception into everyone’s rule. One case is not a pattern, so it is not treated as one.
And when something upstream changes, ALEX identifies exactly what downstream behaviour is affected and marks it stale, rather than letting decisions keep running on a version of the truth that expired.
A model that doesn’t survive a change is a report.
Two ways in. Both converge on the same model.
Start with ALEX when you know you want AI but the operating reality is too fragmented to delegate safely. Start with a Worker when the operational pain is already obvious and you want VERIDOM to own the outcome.
Choosing one doesn’t waste the other. Nothing gets rebuilt.