One queue, described so it can be measured.
Everyone can read a request. Nobody has time to decide where it goes.
Requests arrive in the operations mailbox and through the web form, in whatever words the sender chose. The team reads each one, works out what it is really about, checks whether it has come up before, and passes it to the person who can deal with it.
Most of them are routine. The day is spent on the ones that are not, and on the ones that went to the wrong person first. By the time anyone asks why a request took a week, the trail runs through three inboxes and a chat thread.
What it usually looks like
You are not short of people who can answer a request. You are short of time to get each one to the right person first.
One AI employee, hired on a job description the operations lead writes.
Nothing new is installed for the team. The AI employee is connected to the mailbox, the web form and the ticketing tool the team already uses, and only to those.
On day one it only reads and sorts. Every routing and every reply waits for a person. Before anything runs, the operations lead sets two things, and both are recorded as decisions with their name on them.
Complaints, anything with a legal or regulatory angle, and anything involving personal data are routed to a named person, at every trust level.
Nothing leaves the organisation without a person. Replies to customers and suppliers are drafted, never sent, whatever the level.
The ticketing tool stays the system of record. Xalios sorts, routes and drafts with the reasoning attached. Nothing leaves the organisation without a person.
“Read every request that arrives in the operations mailbox and the web form. Work out what it is about, how urgent it is and who should handle it. Draft a first reply for routine requests. Report to the operations lead.”
Four steps. One of them a person’s, and it takes a minute.
One request, from the moment it arrives to the moment it is answered. Nobody spends time on it until there is something to decide, and what reaches the operations lead is a decision, not a full inbox.
A customer emails the operations mailbox asking why an order shows as delivered when nothing has arrived.
Finds the order and the carrier’s delivery scan. Sees the same customer raised a similar request last month. Marks it urgent, routes it to the deliveries queue and drafts a reply that says what happens next.
Attaches the order, the delivery scan and last month’s request, and sends the draft to the operations lead with a recommendation: confirm the scan with the carrier before replying.
The operations lead reads the draft, changes one line and sends it. The request sits on the right queue with the evidence attached, and the customer has an answer before ten.
What lands with the team is a routed request with the evidence attached. Not an inbox to sort.
What the auditor would read, and what the operations lead reads on a Monday.
The same record, with no separate reporting layer. Both can be replayed. Six months later the request, the order and the earlier requests open as they were when the decision was made, not as they are today.
For the operations team, twelve weeks in.
Not fewer people. Sorting stops getting in the way of answering. The difficult requests, the complaints and the customers who need a call stay with the people who are good at them, and the queue stops growing every Monday.
Every request is read by a person before anyone knows what it is.
Every request arrives sorted, tagged and routed, with the context attached.
Urgent requests wait behind routine ones because nothing sorts them.
Urgent requests are at the top of the right queue within minutes of arriving.
The same question is answered from scratch each time.
Routine requests get a drafted reply from the record, and a person sends it.
Nobody can say why the oldest request is still open.
Every request carries its own record: what arrived, where it went, why, and who decided.
Trust in automation is all or nothing, so it stays at nothing.
Authority is earned for each kind of request, raised by the operations lead and withdrawn automatically when quality slips.
What stays with the operations team
The routing rules and the trust level
Both set and changed by the operations lead. Every change is itself a record with a name on it.
Judgement on the difficult ones
Complaints, anything with a legal angle and any customer who needs a call stay with the people who are good at them.
The ticketing tool
It stays the system of record. Xalios sorts, routes and drafts inside it, with the reasoning attached, and nothing leaves the organisation without a person.
Open a request from week three and get week three.
Most questions about requests are questions about handling: why did this go there, why did it wait, and who decided. A useful test at week twelve is simple. Pick ten requests at random and ask for the reasoning.
The answer to “why did this take a week” becomes a click, not a search through three inboxes.
One AI employee becomes an operating model.
The memory built on requests, which ones are routine, which customers need care, which team owns what, is the foundation the next role starts from. Each addition is faster than the last, and each one starts again at level 1 for its own kind of work.
Request triage
The pilot. Level 3 as the target, with drafting first replies as the next step you can choose.
Supplier chasing
Follows up late confirmations and acknowledgements, and escalates the ones that stay silent.
Order desk
Reads inbound orders, checks them against stock and confirms what can be delivered.
