Improve

Improving a process is a conversation. Ask your agent, for example:

How do we improve this?
What could we remove or simplify before we automate anything?
Which steps could an agent help with?

Your agent answers in the chat with three to five plain steps, the most useful first. Each step names the change and the number glasstarn measured behind it, for example:

Review the standard purchasing route first: 1,059 of 1,500 orders follow it. Ask the process owner which checks this route needs before deciding whether any can be simplified.

A frequent route shows what happened, not which approvals are required or safe to remove.

At the same time, the page in your browser puts on the filter for the step it leads with, so you see exactly which items it means. Ask “show me the next one” to move on.

Why the advice stays in the chat

The browser shows facts only: what glasstarn measured from your data, and what documents or people told your agent. Advice is a different kind of thing. It depends on what you want, what your company allows and what you tell your agent next, so it belongs in the conversation, where you can push back, add a constraint and get a revised answer straight away.

Nothing your agent suggests is saved as approved, and nothing is deployed.

What good advice looks like

• Remove or simplify first. A check nobody reads or an approval that is always granted may not need automating at all.
• Then rules, then agents. A step that follows a fixed rule can go to a rule. A step that needs judgement but follows a pattern might go to an agent, with a person deciding. Some steps stay with a person.
• A number behind every step. If your agent cannot point to a measured number, it says so and labels the idea as a guess.
• A small pilot, not a big rollout, with what to measure and when a person takes over.

glasstarn never makes up timings or savings. A step you only know from an interview has an unknown time until it is measured.

Next steps

• Check against how it should run
• Working with your agent