Getting started with VDF AI Agents

Before you begin

You only need three things for your first run:

1. Access to your VDF AI workspace. If you haven’t signed in yet, start with the Getting Started guide.
2. One real task you’d otherwise do by hand. A recurring update, a draft, a summary — anything where you already know the output you want.
3. One source. A file, a meeting transcript, a connected-app reference, or a few bullet points. Agents work best with something to read.

Step 1 — Open the agent library

From your workspace, open the agent area (sometimes called the Agent gallery or Assistants list). You’ll see a catalog of specialists organized by what they produce: drafts, plans, research, analysis, refinements.

You can browse by:

• Type of output — drafts, summaries, plans, comparisons, refinements.
• Type of work — product, ops, marketing, customer, executive.
• Source material — agents that prefer files, meeting transcripts, or connected-app references.

If you can’t see an agent your teammate mentioned, your workspace may not have it enabled yet — see the FAQ for what to check.

Step 2 — Pick the agent by outcome

Pick by what comes out, not what goes in.

You want to produce	Try an agent that’s described as
An exec-ready summary of a meeting	”Meeting summary” or “Stakeholder update”
A polished draft from rough notes	”Drafting assistant” or “Content drafter”
A refined version of a backlog item	”Backlog refiner” or “Story polisher”
A market or competitor brief	”Research assistant” or “Competitive brief”
A prioritized plan from a raw goal	”Planning assistant” or “Roadmap drafter”
A comparison between two documents	”Document comparison” or “Diff & summarize”

Step 3 — Brief the agent well

A great brief has four parts. Skip any and you’ll trade quality for back-and-forth.

1. The outcome you want.

   "A two-paragraph internal update for our leadership team." Not just "summarize this."

2. The audience.

   "For our exec team" reads differently than "for an external customer." Be specific.

3. The constraints.

   Tone, length, format, must-include sections, must-not-include points. One sentence here saves three revisions later.

4. The source material.

   Attach files. Reference connected apps. Paste the transcript. Give the agent something concrete to work from.

A real example

Outcome: A short Monday update for our product leadership team.

Audience: Internal — VP of Product and engineering leads.

Constraints: Two paragraphs. No more than 250 words. Plain tone. Skip launch dates that aren’t confirmed.

Source: I’m attaching last week’s standup notes and the customer-call transcript from Thursday.

That brief — five minutes to write — gives you a first draft you can ship with five edits, not fifty.

Step 4 — Read the first draft, then refine

Resist the urge to read once and start over.

Good refinement prompts are small and specific:

• “Shorten paragraph two by half.”
• “Replace the second bullet with a customer quote from the transcript.”
• “Make the tone slightly more confident.”
• “Add a one-line risk callout at the bottom.”

After two or three refinements, you’ll have an output you can ship.

Step 5 — Save what worked

If your brief produced a great result, save it. Most workspaces let you save a brief as a starting template — your own or for the team.

The next person who needs to write a Monday update starts from your win instead of a blank page. Over time, your team builds a library of known-good prompts that compound.

Common starter tasks (each takes under 10 minutes)

These are great first runs because they have clear inputs, a clear shape, and a clear “yes, this is good” gut check.

What to read after your first run

• Working with assistants — the brief-context-output loop, in more depth.
• Use cases — six concrete worked examples.
• Tips & best practices — patterns that compound across your team.