Content review agent

Checks every generated asset, scores it on the Quality scorecard, ships high scores, sends the rest to a human in Assist and learns from their corrections.

1Start · WebhookContent studio

When the content studio posts a generated asset to the agent's webhook

Passes in: asset_id, campaign, channel, audience, brief

♦ nothing needed here

2Quick checksThunai

Load the brand kit, offer sheet, asset tags and this agent's learning set

Remembers: brand_kit, offers, rules

♦ nothing needed here

3AI decidesThunai

Score the asset on the Quality scorecard 'Content quality v1'

Brand kit 20 · Right asset 25 · Offer 25 · Audience 10 · Compliance 15 · Format 5

♦ nothing needed here

Then, depending on the score:

If score ≥ 85

4SendDistribution

Ship it to the channel and log the score

♦ nothing needed here

If score < 85

5ApprovalThunai

Ask a content reviewer in Assist › Approvals

Decides: approve · approve with edits · send back

While waiting: Escalate if not decided in 15 min

6ActionThunaiContent studio

If sent back: save the correction as a rule, regenerate and score again

♦ nothing needed here

7SendDistribution

If approved: ship it to the channel

♦ nothing needed here

Then, whichever path ran:

8LogThunai

Write the score, decision and reviewer to the audit log

♦ nothing needed here

Runs, learning set and human effort

What this agent has checked, the rules it learned from reviewers and how much human time it still needs.

Asset	Campaign	Channel	Score	Confidence	Attempt	Status
AS-2041	Electric line launch	Static ad	77	54%	First	Waiting review
AS-2042	Service reminder	SMS	90	66%	First	Waiting review
AS-2043	Summer offer	Short video	78	49%	First	Waiting review
AS-2038	Insurance renewal	Email	95	93%	First	Auto-shipped
AS-2039	Service reminder	SMS	99	95%	First	Auto-shipped
AS-2036	Electric line launch	Static ad	86	71%	First	Approved with edits