Before you start
You need three things for your first network run:
- Access to your VDF AI workspace. If you haven’t signed in yet, start with the Getting Started guide.
- A clear goal. Not a topic — a deliverable. “Produce a one-page competitor brief on Vendor X.” “Generate a Q3 status report from this quarter’s standup notes.”
- The source material. A folder, a few files, or a connected app — whatever the network should read.
Pick a real task for your first run. A task you'd otherwise do by hand makes the time savings obvious. Networks are at their best when the deliverable matters and the steps are well understood.
Step 1 — Open the network library
From your workspace, open the network area (sometimes called Workflows or Network templates). You’ll see a catalog of pre-built workflows organized by what they produce.
Browse by:
- Type of deliverable — reports, briefs, plans, comparisons.
- Function — product, ops, marketing, customer, executive.
- Stage count — short two-stage workflows for quick wins, deeper five- or six-stage ones for serious lifts.
Step 2 — Pick a template (or describe your goal)
You have two entry points:
Pick a template
Templates are pre-built networks the platform ships ready to use. They handle the most common shapes of work — quarterly summaries, vendor briefs, customer onboarding writeups, multi-doc analyses.
If you see a template that matches your task, pick it. You’ll skip the building step entirely.
Describe your goal
If no template fits, describe the goal in plain language. VDF AI will propose a workflow shape — a sequence of stages it thinks fits the task. You’ll see the proposed stages before anything runs, and you can adjust them.
“I need a one-page competitor brief on Vendor X covering pricing, security, and integration footprint. The output is for our exec team.”
That kind of plain-language goal is enough for the platform to suggest a network shape.
Templates are usually faster to start with. Even if you only use them as a starting point, seeing how an existing network is structured gives you a much better feel for how to brief one of your own.
Step 3 — Review the proposed stages
Whether you picked a template or described a goal, you’ll see the network’s proposed shape: a list of stages with one-sentence descriptions of what each does.
A typical first review:
- Read the stage list end to end. Does the order make sense? Are any stages obviously missing or redundant?
- Check the inputs each stage expects. Are they pulling from the right sources?
- Look at the final stage. Is the final output the shape you actually need?
If anything’s off, edit the stage description, swap it out, or remove it. The platform will re-propose connecting stages if you change something major.
Step 4 — Hand over the inputs
Every network needs inputs. At minimum:
- A goal sentence. What the final output should be.
- An audience. Who reads the final result.
- Source material. Files, folders, connected apps, or pasted text.
Some templates ask for more — a deadline, a tone, must-include sections. Fill in what’s asked.
Be honest about sources. If you only have one of the three source documents the network expects, say so. The network can run with less — but it should know what's available so it doesn't fabricate what isn't.
Step 5 — Run the network and watch it build
Hit run. The network starts executing stage by stage. You’ll see:
- Which stage is active right now.
- What the previous stages produced.
- A live status indicator showing progress.
Most templates take a few minutes to run end to end. You can step away and come back — when the final stage finishes, the result is waiting.
Stop and adjust mid-run
If you notice a stage producing the wrong direction — say, the research stage missed a key competitor — you can pause the run, adjust the input or the stage, and continue. You don’t have to start over.
Step 6 — Read the result, then refine
When the run finishes, you’ll see the polished final output plus every intermediate output along the way.
Read the result first. Then ask:
- Does the final output match the goal you set?
- Are the named facts (numbers, dates, customer names) verifiable?
- Does the tone match the audience?
If the result is 80–90% there, you’re in great shape — refine the final output the same way you’d refine an agent output (small, specific changes). If it’s not there, look at the intermediate outputs to find the stage where the work went off track. Adjust that stage, re-run from there, and continue.
Always verify named facts. Networks are powerful but they're not infallible. Especially for facts about real people, customers, and numbers — confirm before you ship.
Step 7 — Save what worked
If the network produced a result you love, save it. Two flavors:
- Save the run for reference. You can return to it, copy the result, or rerun with adjusted inputs.
- Save the network to your team’s library. The next person who needs this kind of deliverable starts from your win.
Common starter networks (each runs in under 15 minutes)
These are great first runs because they have clear inputs and a clear shape:
Weekly status report
Three stages: pull the week's activity, draft a report, polish for leadership.
Competitor brief from a URL
Three stages: gather public info, draft a one-pager, identify where we can win.
Multi-doc comparison
Four stages: read each doc, build a side-by-side, draft a recommendation, polish.
Customer onboarding writeup
Four stages: gather context, draft the welcome brief, critique for tone, finalize.
What to read after your first run
- How networks work — stages, intermediate outputs, and reviewing a run, in more depth.
- Building workflows — when to turn a recurring task into a saved network of your own.
- Use cases — six worked examples.