AI Assistant Capabilities – Use Case Example¶
Multi-step Meeting Coordination with Calendar, Survey & Email Integration
Date: May 2026
Scenario¶
A project coordinator at a research institution needs to schedule a pre-kickoff meeting with contributors of a specific work package (WP) before an upcoming EU project kickoff event in another city. The meeting should cover two task-level agenda items. A scheduling poll needs to be created and distributed via email to all relevant colleagues.
What the Assistant Did¶
1. Context gathering (parallel)¶
The assistant simultaneously:
- Searched emails for the upcoming kickoff event to determine its date and location
- Searched emails for the most recent communication to the WP mailing list to identify all contributor addresses
- Retrieved the current date to anchor all relative date references
All three actions were executed in a single step, in parallel.
2. Calendar analysis¶
Using the identified kickoff date as the upper bound and "end of this week" as the lower bound, the assistant:
- Listed all calendar events across the full relevant period (~4 weeks)
- Computed free slots within business hours across all candidate days
- Identified days to exclude due to travel, fully blocked days, or events ending too close to lunch
3. Slot selection with constraints¶
The assistant applied a layered set of constraints to select poll options:
| Constraint | Source |
|---|---|
| Must fall before kickoff date | Extracted from calendar event |
| Must be free in organizer's calendar | Free-slot computation |
| No lunchtime slots (12:00–13:00) | User instruction |
| Specific date excluded entirely | User instruction |
| Two slots per day (morning + afternoon) | User instruction |
| Slots must be exactly 1 hour and fit within free windows | Logical inference |
The final poll contained 14 slots across 7 days.
4. Meeting poll¶
The assistant created and deployed a Doodle-style scheduling survey using a connected survey service:
- Name field (non-anonymized, to identify respondents)
- Matrix question: each time slot as a row, availability options (Available / Maybe / Not available) as columns
- Free-text comment field for alternative suggestions
- Survey activated immediately upon deployment
5. Email draft¶
The assistant composed a professional email draft addressed to all WP contributors (extracted from the email thread), including:
- Context: upcoming EU project kickoff event
- Meeting purpose: alignment on two task-level agenda items within the work package
- Direct link to the scheduling poll
- Human-readable summary of all proposed time slots
- Response deadline (end of current week)
- Sender signature
The draft was saved to the drafts folder (not sent).
Tools & Integrations Used¶
| Tool | Purpose |
|---|---|
get_current_time |
Anchor relative date references |
semos_agentura_email__search_emails |
Find kickoff event info and WP contributor addresses |
semos_agentura_email__read_email |
Extract full recipient list and thread context |
semos_agentura_email__list_events |
Retrieve calendar events over the relevant period |
semos_agentura_email__free_slots |
Compute available time windows per day |
survey_agent__deploy_survey |
Create and activate the scheduling poll |
survey_agent__delete_survey |
Clean up outdated poll versions |
semos_agentura_email__create_draft |
Compose and save the invitation email |
Key Capabilities Demonstrated¶
- Parallel tool execution – multiple searches and lookups fired simultaneously
- Cross-agent orchestration – email, calendar, survey, and file system agents combined in one workflow
- Constraint reasoning – applying overlapping rules (free windows, excluded dates, no-lunch policy) to generate valid options
- Context extraction from emails – identifying attendees, dates, and topics from existing correspondence
- End-to-end automation – from raw user request to deployed poll + ready-to-send email draft