AI Template Discovery scans your team's sent emails for repetitive reply patterns and automatically generates templates from them. It is the fastest way to build a template library when you are starting with Slokoto or migrating from another CRM.
Where to find it
Go to AI Settings → Templates tab. The page has three sections:
- Email & SMS Templates — your existing template library (create, edit, delete)
- AI Template Discovery — configure and run a discovery scan
- Discovered Templates — review, edit, and approve or reject AI-found patterns
Only admins can access AI Settings and run discovery scans.
How it works
Template Discovery works in three phases:
Phase 1 — Email collection
The AI scans sent emails from the Gmail accounts you select. It reads outbound messages, strips signatures and quoted text, and extracts the actual reply body. The scan moves backward through your email history, one page at a time.
Phase 2 — Pattern detection
Collected emails are sent to AI in batches. The model groups similar replies — emails that say essentially the same thing to different recipients — and extracts a generalized template with {{placeholders}} for names, products, and company references.
Phase 3 — Deduplication
Each discovered pattern is compared against your existing templates. If a pattern is too similar to a template you already have (above 85% similarity), it is automatically skipped. Patterns that are genuinely new enter the moderation queue.
Running a scan
Step 1 — Select Gmail accounts
Check the boxes next to the Gmail accounts you want to scan. Only connected accounts appear in the list. You can select one account or several.
Step 2 — Set the budget
The budget controls how many sent emails the AI will read. A larger budget means the scan goes further back in email history and has more data to find patterns.
| Budget | Typical coverage | Best for |
|---|---|---|
| 50–100 | Last few weeks of sent email | Quick test or small team |
| 200–500 | Last 1–3 months | Standard onboarding |
| 500–2000 | Several months of history | Large team or thorough migration |
| 2000–5000 | Deep historical scan | Comprehensive template extraction |
Step 3 — Choose auto-approve (optional)
When auto-approve is on, discovered templates are added to your library immediately without review. When it is off (the default), they go into the moderation queue.
Step 4 — Start
Click Start Discovery. A progress bar shows how many emails have been scanned out of your budget.
Budget and resuming
The budget is the total number of emails to scan, not a monetary cost. If the scan pauses — either because it reached the time limit or because you cancelled it — it saves its place. You can:
- Resume from where it stopped by clicking Resume
- Increase the budget to scan further back — the AI picks up from its last position and continues backward through your email history
This makes it easy to start small and expand: run a quick scan with 100 emails, review the results, then increase to 500 to discover more patterns.
Reviewing discovered templates
When a scan finishes with auto-approve off, discovered templates appear in the Discovered Templates section. Each card shows:
| Element | Description |
|---|---|
| Template name | AI-generated name based on the pattern |
| Body preview | The extracted template text with {{placeholders}} |
| Occurrence count | How many similar emails the AI found (e.g., "5x found") |
| Product tag | Whether the template is product-specific or generic |
| Similarity warning | Shows if a similar template already exists |
| Pattern description | AI explanation of what the pattern represents |
Actions
Each discovered template has three actions:
- Approve — creates the template in your library as-is
- Edit & Approve — lets you modify the name, subject, and body before approving
- Reject — discards the discovered pattern
You can also use Approve All or Reject All buttons for bulk actions.
Product tagging
The AI automatically detects whether a template is specific to one of your products. It checks the email content against your Product Knowledge catalog. Product-specific templates are tagged accordingly; generic templates get no product association.
What makes a good discovery scan
- Connect multiple team members' Gmail accounts — more mailboxes means more diverse patterns
- Use a budget of at least 200 — small budgets may not contain enough repetition to detect patterns
- Run the scan after setting up Product Knowledge — the AI uses your product catalog to tag templates correctly
- Review before approving — even with good AI detection, you may want to adjust wording or merge similar patterns
Tips
- Start with a small budget (100–200) to see what the AI finds before committing to a large scan
- After approving templates, check them in the Automate tab to verify they work well with your automations
- If you add new team members with Gmail later, you can run another scan to discover their patterns
- Templates discovered from sent email are created as public so the whole team can use them
Next: Learn about Email & SMS Templates