Support mail in by nine o'clock: who decides who gets it?
Tuesday morning, quarter past nine. The support inbox holds 23 new tickets. One is about an invoice, two are about the same login error, one comes from a customer who has been waiting for three days, and the rest are questions whose answers are already in your help center. Before a single reply has been sent, half an hour has gone on sorting alone.
That is triage. It isn't difficult work, but it has to be done again every day, and it always comes at the expense of the work your customer is actually asking for.
How triage usually works today
In a small software company, or a web shop with a support team of one or two people, it goes roughly like this. Someone opens each ticket, reads it, decides what it is about and how urgent it is, and assigns it to themselves or to a colleague. Only then does the reply begin.
As a simple calculation, take a company with 90 tickets a month and two minutes of sorting per ticket. That is three hours a month, just to decide what goes where. The figure is an assumption, not a measurement. Check in your own helpdesk how long it takes you; keeping track for a week is enough to find out.
The bigger problem isn't those three hours. It's the mistakes that creep in. The ticket from a customer who is thinking of cancelling stays buried among the rest. Two tickets about the same outage are answered separately by two people, with two different answers.
What an agent takes over
The Supportdesk assistant proposes ticket triage and writes customer-facing knowledge base articles. In the Stilops catalog this role costs 199 euros per month, with 90 tasks per month included. An extra task costs 2.50 euros.
In practice it comes down to this. The agent reads the incoming tickets and proposes a category and an urgency for each one. It groups tickets about the same error. For questions whose answer is already in your knowledge base, it prepares a draft reply. If the same question keeps coming back and there is no article about it, it proposes a knowledge base article.
If you use Zendesk or TOPdesk, the agent can read tickets there and help draft replies. How to connect it is described in the integration guides. If you work with something else, that is a conversation, not a promise.
What you approve is an outcome
This is where it differs from most tools that automate an inbox. With Stilops, you don't read an intention such as "the agent will sort and answer tickets". First you see a test run: a sorting of real tickets, with the draft replies alongside, before anything goes to a customer.
On that screen you look at three things:
- Is the waiting customer's ticket at the top, or has it sunk out of sight?
- Have the two tickets about the same login error been grouped together?
- Does the draft reply to the invoice question match what you agreed with that customer?
If it's right, you approve it. If one reply isn't how you would write it, you adjust it and the rest goes ahead. So you have always seen every reply that leaves the building. That makes the work a little slower than an agent that sends on its own. For customers you want to keep, that is a deliberate choice.
Where people stay in charge
Not everything belongs with an agent. A complaint from a customer who has lost trust, a refund, or a promise about a new feature is something you decide yourself. The agent prepares such a ticket with the relevant details and leaves it with you. On money and on what goes out to the world, a person decides.
You also see what each task costs. That makes a badly configured agent visible quickly, rather than only after a year.
Start with one morning
Tomorrow at nine o'clock, open your inbox and count how many minutes it takes before you write the first reply. Write the number down and multiply it by twenty working days. Those are the hours an agent can take over. What you do with them is up to you, for example calling back the customer who has been waiting for three days.
If you want to see what such a test run looks like, you can try Stilops without an account. You choose a role, give it a sample task and read what the agent would do, before you approve anything.