Skip to content

Getting good answers

How to ask, what the assistant can filter and sort on, and the prompt patterns that consistently produce useful answers from your Vigilis data.

Updated View as Markdown

Most of the difference between a useful session and a frustrating one is in how the question is asked. None of this requires learning a syntax.

Ask for the shape you want

“Give me a table of expiring contracts with the provider and end date” produces a far better answer than “check contracts”, because it tells the assistant which fields to pull.

Start read-only

Explore with questions before you ask for changes. It costs nothing, and it builds an accurate picture in the conversation so the eventual change lands on the right record.

Read the preview

Every create, update, delete, and bandwidth change shows you what would happen before anything happens. It is the last checkpoint before a real write. See What your assistant can do.

Ask what fields exist when a name is fuzzy

“What fields does a service have?” is faster than guessing. An unknown field name is a hard error rather than something quietly ignored, so a guess costs you a round trip.

One task per thread

Long conversations that wander across sites, carriers, and record types make it easy for both you and the assistant to lose track of which record is in scope.

Prefer a report for a total

Spend by provider is one bounded query. Asking the assistant to list every invoice and add them up is slower and stops at 100 rows.

Do not paste secrets

Fields that look like credentials are stripped from everything the assistant can read, so it never sees them. Do not reintroduce one by typing it into the conversation.

What the assistant can filter and sort on

You do not have to write any of this. It is worth knowing it exists, because it is the difference between “check contracts” and a question that comes back already sorted and narrowed.

Capability What it does
Free text Matched against that record type’s searchable fields, such as an invoice number or a circuit’s service ID
Sort Any field, ascending or descending
Filter Conditions combined with AND and OR, using equals, not-equals, contains, greater than, less than, is empty, and is not empty
Relative dates This month, last month, next month, this quarter, last quarter, next quarter
Field selection Return only the columns you asked for
Creation date bounds Restrict results to records created between two dates
Paging 25 rows by default, 100 at most, and 200 for a report

Prompt patterns that work

Goal Ask it like this
Renewal exposure “Which contracts expire in the next 90 days, sorted by end date, with the provider name?”
Site audit “List every service and circuit at our Chicago location.”
Spend trend “Spend by provider for the first half of this year, biggest first.”
Invoice hunt “Find invoices from Lumen with status ‘disputed’.”
Ticket triage “Open tickets created this month, newest first.”
Careful bulk edit “Show me the three services missing a location, then set them to the Dallas site, one at a time, showing me each change first.”
Contract check “Compare the committed rate on the Lumen contract against what they actually billed last month.”

Common questions

Do I need to learn a query syntax?

No. Describe what you want in plain language and the assistant builds the filter, sort, and field selection for you. Knowing what is possible just helps you ask for more.

Why did my assistant only return 100 records?

Lists return 25 rows by default and 100 at most, and reports cap at 200. There are no unbounded queries. For totals across everything, ask for a report rather than a list.

What is the fastest way to get a spend total?

Ask for spend by provider. That is one bounded report. Asking the assistant to list every invoice and add them up is slower and stops at 100 rows.

Navigation

Type to search…

↑↓ navigate↵ selectEsc close