How to QC a new template or agent in Assist 2.0

Who this is for: Anyone asked to check a new template or agent in Assist 2.0.

Goal: Run the template or agent 10-20 times with real/sensible inputs and log any snags on the QC Monday board.

1) Before you start (2 mins)

  • Open Assist 2.0 and find the template or agent you've been asked to check (either in your workspace's installed list, or the marketplace if it hasn't been installed yet).
  • Gather a few real or sample files (briefs, results/metrics, visuals, T&Cs, market lists).
  • Open the QC Monday board in another tab.

2) What "good" looks like

  • Client-ready document: clear headings (H2/H3), bullets, UK English.
  • Uses only uploaded materials and workspace knowledge; if info is missing, shows a clear placeholder (e.g., [MISSING: …]) and a one-line "What I still need."
  • On file conflicts, prefers the newest, most specific source and notes it briefly.
  • Tables (e.g., markets, KPIs) are accurate and tidy; images/screenshots are referenced clearly (e.g., "Figure 1").
  • Short Sources/Provenance list if the output includes one.
  • Plan Mode behaviour (if the template/agent uses it): the proposed plan is sensible, editable, and only runs once you accept it.

3) Run your tests (target: 10-20 uses)

Start a new chat in the workspace each time, and either call the template/agent directly or describe what you want in natural language (Just Start Talking) and confirm Assist 2.0 picks up the right existing template/agent rather than spinning up a new one.

Keep each run short (1-4 minutes). Mix these:

A. Complete inputs (happy path)

Provide all expected files. Check structure, tone, tables, KPI formatting (e.g., +27% uplift).

B. Missing info

Withhold a key input (e.g., results or markets list). Confirm [MISSING: …] appears with a helpful "What I still need."

C. Conflicting files

Upload two versions of the same file (dated). Confirm it uses the newest and briefly flags the choice.

D. Multi-market / multi-language (if relevant)

Provide 2-3 markets with differences (mechanics, prizes, T&Cs). Check the comparison table reflects this; language tags where applicable.

Add an image/screenshot. Check it's referenced (e.g., "Figure 1"). If none supplied, the output should leave a clear placeholder and (optionally) search suggestions.

F. Revision loop

Change one input (e.g., updated results file) and ask it to refresh just that section in the same chat. Confirm it updates cleanly without regenerating the whole document.

G. Reuse check (Just Start Talking)

In a fresh chat, describe the same task in natural language without naming the template/agent. Confirm Assist 2.0 finds and reuses the existing template/agent rather than creating a duplicate.

H. Plan Mode (if applicable)

If the template/agent runs in Plan Mode, check the plan is clear, accurate, and editable before you accept it. Try revising the plan and confirm the run reflects your changes.

4) Log snags on the QC Monday board

Use the same columns:

  • Name — short title of the issue
  • Link to chat — paste the Assist 2.0 chat URL
  • What task were you trying to do? — 1-2 lines (e.g., "Generate case study with 3 markets")
  • What happened? — actual behaviour (copy error, missing table, wrong tag, duplicate template created, etc.)
  • What was the expectation? — concise expected outcome
  • Severity — Low / Medium / High / Critical
  • Screenshots (Optional) — attach or link
  • Status (For Assist use) — leave blank

Severity guide

  • Critical: document incomplete/structurally broken; can't deliver to client. Or: template/agent duplication is happening on every run.
  • High: key section wrong/missing across runs.
  • Medium: formatting/logic issue with workaround.
  • Low: copy polish or small layout.

Example snag

  • Name: Markets table ignores prize differences
  • Link to chat: https://…/assist/chat/xyz789
  • What task were you trying to do? Generate a multi-market document (UK/DE/FR) where prizes differ.
  • What happened? Table shows identical prizes for all markets.
  • Expectation: Table lists correct prize differences and any T&Cs notes per market.
  • Severity: High
  • Screenshots: (attached)

5) Finishing up

  • If you hit no snags, let us know you're happy with the quality and that QC is complete.
  • Unsure if something is a snag? Log it anyway — we'll triage severity.
  • Need help? Reply to the rollout email with the template/agent name and your board link.
Did this answer your question? Thanks for the feedback There was a problem submitting your feedback. Please try again later.

Still need help? Contact Us Contact Us