The Soon AI Scheduling Assistant streamlines activity assignments within shifts, offering fast, secure, and private support. Think of it as a helpful person assisting with scheduling tasks, including bulk updates, rather than a mathematical solver optimizing against complex constraints. For top results, prompts must be clear and specific. Its effectiveness depends on users knowing when to use "Fast" or "Smart" mode and how to prompt clearly. This guide provides tips, examples, and mode explanations to simplify scheduling.
Before you start
Plan Availability: The AI Assistant feature is available to teams subscribed to the Business or Enterprise plan.
Access Rights: Only admins with "Can Edit" permissions in the "Workload" intraday management view can utilize the AI Assistant.
Why Specificity Matters
The assistant processes commands with precision, coding names and encrypting data behind the scenes. Vague prompts like "Schedule a break" leave ambiguity, such as who it’s for or when it occurs. Users should specify individuals (e.g., "Alex" or "Team A"), activity types (e.g., "Call", "Break"), and times (e.g., "10:00-10:30") for spot-on results. Prompts can be in any language, and the assistant will respond in that language, too.
Boost Precision with @ and #
Using @ for users or groups (e.g., @Alex, @TeamA) and # for activities (e.g., #Call, #Break) enhances security and accuracy. These tags pinpoint exactly who or what is intended, avoiding errors from typos or duplicate names like "Alex Smith" vs. "Alex Jones". This method ensures the assistant assigns the right person or group to the right task, keeping data protected.
Security and Privacy
The assistant prioritizes user privacy. All names and groups mentioned (e.g., @Alex, @TeamA) are coded into encrypted IDs before processing, ensuring no personal data is exposed to the AI. Activity details stay secure, and the system only acts on explicit user commands. This design keeps sensitive information locked down, making it a trusted tool for scheduling. See more at Anthropic https://www.anthropic.com/legal/#security and OpenAI https://trust.openai.com/.
Fast Mode vs. Smart Mode
The assistant offers two modes:
Fast Mode: Ideal for simple, single-task prompts requiring no reasoning.
When to Use: One quick task.
Example: "Schedule @Alex on #Call from 09:00 to 10:00."
Why: Swift execution if the slot’s free.
Limit: Struggles with overlaps or multi-tasks.
Smart Mode: Suited for multi-task prompts or when reasoning is needed (e.g., resolving overlaps).
When to Use: Multiple tasks or potential conflicts.
Example: "Schedule @Alex on #Call from 09:00 to 11:00 and @Maria on #Break from 10:00 to 10:30."
Why: Handles both tasks in one prompt.
Limit: Takes longer to process. Vague prompts may lead to overthinking, missing the intended outcome.
Inspirational Prompts to Spark Ideas
Explore these creative prompts—copy, paste, and tweak as needed (times, names, etc.):
Break Scheduling:
Schedule a break for each user whose shift runs from 08:00 to 17:00, with a duration of 30 minutes. Distribute the breaks between 11:30, 12:00, and 12:30.
Schedule a 30-minute break for all users. Distribute the breaks between 11:30, 12:00, and 12:30.
Schedule a 30-minute break for @usergroup. Distribute the breaks between 12:00 and 13:00.
Schedule a 30-minute break for all users. Distribute the breaks between 12:00 and 13:00, ensuring that no more than 50% overlap.
Schedule two breaks for all users: one lasting 30 minutes and the other 15 minutes. Distribute the longer break between 12:00 and 13:00, ensuring a maximum overlap of 50%. The second break should start between 1.5 and 2 hours before the shift ends, with a maximum of 3 people overlapping.
Schedule a 30-minute break for @usergroup. Distribute the breaks between 12:00 and 14:00 in 15-minute increments for start times.
Schedule a 30-minute break for @user starting at 12:00.
Schedule a 30-minute break for all users, starting 4 hours into their shift.
Schedule break based on the following:
If the shift < 5 hours → No break.
If the shift is between 5-8 hours → Assign 30 minutes of break.
If the shift > 8 hours → Assign a total of 45 minutes of break, divided into a 30-minute break and a 15-minute break.
The first break occurs at 50% after the shift start time.
Ensure there are at least 3 hours between consecutive breaks for a user with multiple breaks.
Pro Tips for Prompting
Name users or groups clearly (e.g., @Alex, @TeamA).
Use exact times (e.g., "09:00-10:00") or ranges if flexible.
Structuring your prompt as bullet points helps the AI understand it better and generate more accurate results.
Specify activity types (e.g., #Call, #Training).
Invest time in crafting reusable prompts—initial tweaks save time later.
Verify results, undoing with Ctrl/Cmd + Z if needed.
AI Intraday Assistant: Considerations and Best Practices
The AI Intraday Assistant is powered by Anthropic’s Claude https://www.anthropic.com/ for "Fast" mode and OpenAI https://openai.com/ for "Smart" mode. "Fast" mode, using Claude, excels at quick, simple scheduling tasks, while "Smart" mode, leveraging OpenAI, handles complex scenarios requiring reasoning. While accuracy is a priority, AI-generated results may vary. Privacy and security are taken seriously—user data is protected and handled with the utmost care.
Accuracy: AI-powered tools generate results by processing large datasets and patterns, but accuracy can vary. Due to AI limitations, there may be inaccuracies or omissions. Treat the information as a helpful guide, not a substitute for expert advice or thorough research.
Your Responsibilities: When using the AI Intraday Assistant in Soon, users are responsible for following the usage policies of both Anthropic and OpenAI. Always verify AI-generated output before making decisions, especially for critical scheduling matters.
Anthropic Usage Policies: https://www.anthropic.com/legal/acceptable-use-policy/
OpenAI Usage Policies: https://openai.com/policies/usage-policies/
Data Privacy: Sensitive information about users or teams is not shared. By default, data processed by Anthropic and OpenAI is used only for generating scheduling results and is not used for model training. All data remains within Soon’s controlled environment and is securely stored.
Visit the following for more on data retention and security practices:
Anthropic Data Usage: https://privacy.anthropic.com/en/articles/10023580-is-my-data-used-for-model-training
Anthropic Security Information: https://www.anthropic.com/legal/#security
OpenAI Data Usage: https://openai.com/policies/api-data-usage/
OpenAI Security Information: https://trust.openai.com/