Smart Paste
Smart Paste lets you drop raw, unformatted text — meeting notes, call summaries, text message threads, or shorthand jottings — into Policy Stack and have the AI extract a structured interaction record automatically.
How It Works
- Open a client's detail page and click Add Entry on the timeline
- Switch to the Smart Paste tab in the composer
- Paste your raw text into the input area
- Click Extract — the AI processes the text and produces a structured result
What the AI Extracts
From your pasted text, the AI identifies and populates:
| Field | Description | |-------|-------------| | Interaction type | Call, Text, Email, Meeting, or Internal Note | | Direction | Inbound or Outbound | | Content summary | A cleaned-up version of the key points | | Facts | Specific data points mentioned (e.g., policy numbers, dates, amounts) | | Suggested follow-ups | Action items or next steps identified in the text |
Side-by-Side Review
After extraction, you see a split view: your original raw text on the left, the structured output on the right. This lets you verify that the AI captured everything accurately before saving.
You can edit any extracted field directly in the right panel:
- Change the interaction type or direction if the AI guessed incorrectly
- Edit the content summary to add context or correct details
- Remove or modify suggested follow-ups
- Add facts the AI may have missed
Smart Paste works with informal shorthand too. Notes like "called john re: 2nd policy, wants to add PUA rider, f/u next tues" will be parsed into a structured call record with a follow-up date.
Accepting the Result
Click Accept to save the structured interaction to the client's timeline. When you accept:
- The interaction is added to the timeline at the detected date and time (or now, if no date was found)
- Any suggested follow-ups are automatically created as follow-up items for that client
- Facts are stored as part of the interaction record
Low-Confidence Fallback
If the AI cannot determine the interaction type with sufficient confidence — for example, if the pasted text is very brief or ambiguous — the entry is saved as an Internal Note rather than guessing incorrectly. You can always edit the type after saving.
Tips for Better Extraction
- Include dates and times when possible — "Met with Sarah on 4/10 at 2pm" gives the AI more to work with
- Mention the communication channel — "called," "texted," "emailed," or "met with" helps determine the interaction type
- List action items clearly — "follow up on," "send them," or "schedule" are strong signals for follow-up extraction
- Longer notes generally produce more accurate results than one-line entries