What a person sees — the screens of an app with agents in it

The screens people meet in an app with agents: an approval in their inbox, a question from an agent, a business process told step by step, one page that explains a piece of an agent's work, the place to correct it, and the screens a person runs the agents from. Each is a real screenshot taken by a test of the kit that draws it. This page has no code.

You are here: Agents in an Osy# app › What a person sees

Summary#

Nobody "uses the agents". People use their app, and now and then an [[agent-guide-glossary#agent|agent]] has put something in front of them, or done something they want to understand. These are the six places that happens. Every picture on this page was taken by an automated test, in a real browser, of the screen it shows.

Description#

An approval, in the inbox#

An agent that may not do something alone proposes it. The [[agent-guide-glossary#proposal|proposal]] arrives where the person already works.

An approval in the inbox: what will happen, the supplier and the payment days, and what the agent looked at first

Read it from the top:

Who asks. "Vetter (an agent) asks for your approval."
Why. The agent's own reason, in a sentence.
What will happen. The exact action and what it will be done to. The small label says what kind of thing it is: this one changes data.
What it looked at first. So the person can judge whether the agent knew enough.
How it runs. "If you approve, it is done at once, as you."

There are three answers: approve, reject with a reason, or approve and let the agent do this alone from now on. The third is offered only to somebody the app allows to say it.

One click down: [[agent-requests#sheet]].

A question#

Sometimes an agent does not need permission. It needs to know something only a person knows.

A question from an agent in the inbox, with a box to type the answer

The agent's work stops there and waits, for a minute or for a week. Nothing runs meanwhile and nothing is spent. When the answer comes, the work goes on from the same point. One click down: An agent asks a person, and waits — the person rung of the ladder of help (Agents kit).

A business process, told step by step#

When agents take part in a process, each run of it has a page. It says where the run stands, draws the way it went, and tells what happened in each step, in order.

The first two steps of a supplier qualification: the screener finished; the assessor failed twice and the step was handed to a person

A red line is a problem. Here an agent's work failed twice, and the last line says what the process did about it: it sent the step to a person.

A reader who was not there can follow it without asking anybody. The whole of this run is a story of its own: A supplier is qualified — five days, four agents, one that failed, and a three-day wait. One click down: [[workflow-processes-kit#run]].

One piece of an agent's work, explained#

Any piece of work can be opened on one page, the [[agent-guide-glossary#inspector|inspector]]. It answers the questions a person asks afterwards.

The top of the inspector: who gave the work, for whom it was done, what it handed back, what it cost, the warning it proposed and who rejected it, the question it asked and the answer

A weather station's forecaster agent drafted a bulletin. It asked the meteorologist one question, and proposed a gale warning that she rejected: "The harbour master posts warnings himself."

Further down are all its steps in order, each as a sentence, with how long each wait lasted:

The fourteen steps of the same piece of work, from being given to handing back its result

And at the bottom, whether the record is still as it was written:

The record checks: all 14 lines are as they were written

Not everybody sees all of it. The person the work was for reads everything. A supervisor reads that each step happened, and not what was said in it. Somebody with no business there is told the work is not theirs to see. One click down: [[agent-inspector#who]].

The place to correct it#

When an agent gets something wrong, the person who sees it says so, on that piece of work: what was wrong, what it should have been, and why.

The "Correct this" form on a museum label: what was wrong, what it should have been, why

A [[agent-guide-glossary#correction|correction]] does not change what the agent does. It can be turned into an instruction, and an [[agent-guide-glossary#instruction|instruction]] reaches the agent only after somebody the app trusts with that has read it and made it active:

An agent's Instructions page: one instruction is active, with who proposed it, who made it active, and how much work has been sent it

Every change is numbered, and the agent can be put back to how it was before. One click down: Correct this, and an agent's Instructions page — the screens people teach an agent on (Agents.Teaching kit).

Where a person runs the agents from#

Somebody answers for every agent. This is where they look.

The roster: five agents, who answers for each, whether it is working, waiting or paused, and what it spent today

A film archive's five agents. One is paused, with the reason written beside it. One has spent most of what it may spend today.

From here a person in charge can pause an agent, call a piece of work off, try it again, or hand it to a person. There are screens for all the work, for everybody the agents are waiting on, for the standing permissions people have given, for spending, and for who may do what:

Spending: what was spent this month, by agent and by skill, and the work that was stopped at a limit