The approval flow
Nothing reaches a real inbox without a human reviewing it first. The Approvals tab is where you read every sequence Kai has written, edit anything that needs adjusting, and send it live. This is the most important step in the pipeline.
What you see in Approvals
Each card in the Approvals queue represents one contact. It shows the contact's name, title, company, and the sequence Kai wrote for them. The sequence is split across three tabs: Emails, LinkedIn, and Call brief.
Hi Allison,
Jellyfish just published your State of AI in Software Engineering report, and it is front and centre on the site. Smart move given how much interest there is in AI developer productivity right now.
One thing worth checking: if that report is being delivered as a PDF, the AI tools your buyers use daily to research engineering solutions cannot index or cite any of it.
Reviewing a sequence
For each card, read through all three tabs before approving. Here is what to look for in each one.
Emails tab
This shows every email in the sequence in order. Each card shows the subject line and the full body. Look for:
- The hook in Email 1 should reference something specific and real about this person or their company. If it is generic, hit Reject and regenerate.
- Each email should introduce a new angle. If two emails say the same thing in different words, that is a signal the research was thin.
- The tone should read like a well-prepared human rep, not a marketing email. No jargon, no bullet points, short sentences.
LinkedIn tab
This shows the connection note and any LinkedIn messages. The connection note must be under 200 characters and must reference something specific. If it reads like a generic "I'd love to connect," it will not land. Reject and regenerate.
Call brief tab
This is for the rep, not the contact. It shows the hook Kai recommends opening with, the key talking points, and a set of prep questions based on what Kai found in its research. The brief is also visible in the HubSpot task created when the contact reaches the call step. Review it to make sure the research hook is accurate and specific.
Editing before approving
Every field in the sequence is editable before you approve. Click Edit on the card, then click into any textarea to make changes. When you click Approve after editing, Kai records that the sequence was human-edited. This feeds back into the learning loop so Kai can understand what the approved version looked like versus what it originally wrote.
Approve all
The Approve all button at the top of the Approvals page approves every sequence in the queue at once without editing. Use this only when you have already spot-checked a sample and are confident in the quality. It is not recommended for the first batch from a new campaign or after any change to the Lemlist sequence structure.
Rejecting a sequence
If a sequence is not good enough to send, click Reject. The contact goes back to Pending status. You can regenerate it by clicking Run on the contact row, which will re-research and rewrite the sequence from scratch. This does not cost you anything extra, and the second attempt often produces better output because Kai uses any new data Firecrawl has found since the first crawl.
What happens when you approve
Approving a sequence immediately pushes it to Lemlist. The contact is created as a lead in your Lemlist campaign, all the personalised variables are injected, and the sequence starts running on its configured schedule. The contact moves to In Sequence status in Kai and you will see them appear in the In Sequence tab. See Pushing to Lemlist for what to check after a push.