# Duet spotted 3 expiring trials, drafted nudges, then wrote welcome emails after conversion — founder only reviews and sends

Claude Duet · X (Twitter) · Amount undisclosed (workflow case, not a revenue case; 3 trials converting to paid is author-reported, no amount disclosed, no causal claim that nudges caused conversion)

## Business summary

AI2SQL founder Mustafa connected his product data to Claude Duet. Duet independently found three trials ending that night with no queries since Oct 21 and offered to draft nudges. After approval it produced the drafts. Once the trials converted, Duet checked the database again and drafted a welcome email for each new paid user. The founder only reviews and sends. No need to prompt first — useful for solo SaaS trial funnel and post-conversion onboarding.

## Business value

AI2SQL founder Mustafa connected his product data to Claude Duet. Duet independently found three trials ending that night with no queries since Oct 21 and offered to draft nudges. After approval it produced the drafts. Once the trials converted, Duet checked the database again and drafted a welcome email for each new paid user. The founder only reviews and sends. No need to prompt first — useful for solo SaaS trial funnel and post-conversion onboarding.

## Workflow

Duet proactively scans and identifies trials that are about to expire with low recent activity

Duet generates a personalized nudge draft for each account

Outbound messages go out after user review

Duet suggests a follow-up task → checks database → drafts a welcome email for each new paid user

## Preparation and limitations

Results and conversion outcome are as reported by the original post author; the original post does not establish that nudges caused conversion; no revenue amount disclosed. Workflow, prompts, and trial I/O are AgentFlow reconstructions and were not live-tested here.

## Starter prompt

````text
You are my SaaS ops partner (Claude Duet style).


- Product name: {{product_name}}
- Data source connection notes: {{data_source}}
- Trial-end rule: {{trial_end_rule}} (e.g. “ends tonight 23:59 and no query since {{inactive_since}}”)
- Welcome email goal: {{welcome_goal}} (e.g. confirm payment, guide first success, offer support)
- Brand tone: {{tone}}


1. Proactively scan the connected data source for trials that match the “about to expire + low activity” rule.
2. Produce a short finding summary + suggested action (draft nudges) and place it in Up next; wait for my confirmation before writing.
3. After I confirm, write one personalized nudge draft per account (do not send).
4. When a conversion is detected, re-check the latest user state and draft a welcome email for each new paid user.
5. Never send, pay, or delete without my explicit confirmation.


- Discovery summary (count + key signals)
- Suggested action + confirmation prompt
- List of email drafts (subject + body, one per user)
- Missing-info checklist if data is incomplete


- If last-activity or end-date cannot be confirmed, list the gaps and ask; do not guess.
- If a user has conflicting records, prefer the newest timestamp and flag the conflict.


- Never auto-send email, never auto-charge or change paid status.
- Execute outbound only after I reply “send” or “confirm”.

Untested reconstruction based on the original posts.
````

## Costs and pitfalls

Results and conversion outcome are as reported by the original post author; the original post does not establish that nudges caused conversion; no revenue amount disclosed. Workflow, prompts, and trial I/O are AgentFlow reconstructions and were not live-tested here.

## Source and attribution

mustafaergisi (@mustafaergisi)
https://x.com/mustafaergisi/status/2114968764143767734
2026-10-27
Results come from the source author. AgentFlow adds analysis and proposed workflows; outcomes are not guaranteed.