Daily operators
AP, AR, and expense staff: the queues they work, how to handle an exception, when to escalate, and how to correct something without creating a second problem.
Services
A customer who needs us to change a report, adjust a threshold, or investigate an escalation is a customer we have failed to enable. Consultancies with a different business model would call that recurring revenue. We would rather sell you something else.
Tell us your team and their roles. We will propose a programme.
A bill arrives; an agent codes it and routes it.
The work starts without a person. Requires an authority model, an audit trail, and a policy engine — which is why it cannot be retrofitted onto a screen-driven system.
The situation
By role, because a controller and an AP clerk need almost nothing in common.
AP, AR, and expense staff: the queues they work, how to handle an exception, when to escalate, and how to correct something without creating a second problem.
Close, reconciliation, adjustments, consolidation, and the audit trail — plus how to answer an auditor without raising a support ticket.
How to read a confidence score, what an escalation is telling you, when to override, and how to tell a model error from a data error. This is the genuinely new skill.
Whoever needs to answer questions without asking anyone: dimensions, drill-down, saved views, and scheduling.
Permissions, policies, thresholds, authority grants, and onboarding. Usually one or two people who become the internal authority.
Ninety minutes: what the system can answer, what it cannot, and how to tell whether a number in front of you is reliable.
Every finance professional knows how to review another person’s work. Almost nobody has been taught how to review an agent’s, and it is a different skill with different failure modes.
People are inconsistent in familiar ways: they get tired, they take shortcuts under pressure, they have blind spots you learn. Agents are consistent, including consistently wrong in one specific way that looks entirely plausible and repeats identically three hundred times.
So the review technique differs. Sampling works for people and misses systematic agent errors. What works for agents is looking at the distribution — where confidence clusters, which vendors escalate most, whether the reversal rate is moving. We teach that explicitly because it is not intuitive and it is not taught anywhere else yet.
A four-hour session covering everything teaches nobody anything. An AP clerk does not need consolidation and a controller does not need the expense queue, and sitting through the other person’s material is how training gets a reputation.
Sessions are ninety minutes to two hours, by role, on your data. Generic training on demo data does not transfer, because the hard cases are always yours.
Live training reaches whoever was in the room in that month. Everybody hired afterwards gets shown by a colleague, and by the third generation the process being taught has drifted from the one that was designed.
So every session is recorded and indexed against your configuration, and the library is updated when the system changes. New joiners get the same material rather than an oral tradition.
The measure is support tickets. A well-enabled customer raises them about genuine defects and new requirements, not about how to do something. If we are answering how-to questions six months in, the training failed and we will run more at no charge.
Where to start
Who does what, and therefore who needs which sessions. Usually four to six distinct groups, and most people attend one or two.
Live sessions on your configuration with your actual edge cases. Demo data teaches the demo rather than the job.
Every session recorded, indexed to your setup, updated when the configuration changes. New joiners get the same material.
Support tickets reviewed by category. How-to questions mean a gap, and we run the session again without charging for it.
Questions
Tell us your team and their roles. A well-enabled customer raises defects, not how-to questions.