In Mesh AI, AI-Generated Cells function as a high-level analytical tool, converting natural language prompts into graphs, tables, and structured data.

To achieve the highest fidelity in your reports, view the AI as a junior analyst newly integrated into your clinical team. It possesses the technical capability to process data but lacks the intrinsic clinical context of your department; therefore, precise "orders" are essential for accurate results.

Strategies for High-Fidelity Data Extraction

Use the following guidelines to ensure your clinical and administrative analyses are precise:

Define the Output Format: Explicitly state the required visualization, such as a pie chart, bar graph, or summary table (e.g., "Generate a pie chart of..." or "Tabulate a list of...").
Specify Granular Requirements: Avoid broad requests like "Show shift data". Use specific parameters, such as "Provide a bar graph illustrating total shift counts per associate".
Isolate Clinical Variables: Define exactly which positions, call types, or request categories to include (e.g., "Filter for Night shifts only" or "Include only weekend coverage").
Communicate Departmental Workflows: While the AI understands core concepts like Shifts and Associates, it requires guidance on your specific logic. For instance, specify if a "Conference" request should or should not impact an individual’s total shift quota.
Outline Procedural Logic: If your analysis involves multiple steps, describe your thought process to the AI (e.g., "First, identify all approved vacation requests, then calculate the remaining shift capacity per member").
Incorporate Operational Rules: The system is unaware of internal departmental constraints unless explicitly mentioned. Clearly state any relevant rules, such as identifying instances where a provider is scheduled for more than four consecutive night shifts.
Perform Iterative Refinement: If the initial output requires adjustment, refine your existing prompt rather than restarting. For example, modify "A pie chart of shifts" to "A pie chart of shifts, restricted to associates with the 'ER' position".

Examples

❌ Unclear Prompt	✔️ Clear & Effective Prompt
Show shifts numbers.	Create a graph of the number of shifts each associate has.
Who works too much?	Show a list of the people who have more than 9 shifts.
Show max shifts per week.	For each week, calculate the number of shifts each associate has that week; then show a table of the maximum number of shifts that associate has in any one week.
Find weeks where people are overbooked.	Give me a list of times that people have more than 2 shifts in any week, along with the shifts that contribute in order by start date.
Break down shifts by job.	Show a stacked bar graph of the number of shifts each associate has, separated by job.
Find out if scheduling rules are broken.	List all shifts where someone is scheduled for more than 5 consecutive days, along with the shift details.
Tell me the percentage of Prefer On - Day requests that were honored	Show the percentage of Prefer On - Day requests being honored. Having a Prefer On - Day request means the associate wants a shift that starts between 8am and 4pm during that request.
Shifts for each person vs. GME limit.	List the amount of shifts each person has worked compared to the GME limits. The GME limits for our team are: no more than 1 in 4 'In Hospital Call' days and no more than 1 in 3 'Home Call' days, and no more than two weekends in a 28 day period.

Advanced Configuration: Information for Developers

For complex data requirements or highly specific reporting needs, you can directly create or modify the code within an AI cell. This level of access allows you to manipulate raw data structures to produce customized clinical insights.

Core Data Arrays

The programming environment has access to the following primary data arrays. Note: Do not modify these terms, as they are essential for the system's analysis.

shifts: Individual scheduled blocks of time.
requests: Member submissions for time off or availability.
tours: Patterns of recurring shifts or rotations.
associates: The physicians, residents, or staff members in your team.
jobs: The specific duties or tasks assigned within the team.
positions: The professional roles or designations of your members.
requestTypes: The categories of submissions (e.g., Vacation, CME, On-Call).
holidays: Standard recognized public or religious holidays.
specialDates: Custom-defined dates relevant to your specific department.

Implementation Note: Ensure your code references these exact schemas to maintain data integrity across your analytical reports.

Associate:

Field	Type	Description
id	integer	Links associate to shifts or requests
name	string	Name of the person
email	string	Account owner’s email
organization_email	string	Organization-specific email
contact_info	object	Object with `primary_contact`, `secondary_contact`, `tertiary_contact`; blank = 'N/A'
role	string	Role in organization; blank = 'N/A'

Shift:

Field	Type	Description
id	integer	Unique identifier
job	string	Job name this shift is linked to
start	string	Start datetime (with offset)
end	string	End datetime (with offset)
allDay	boolean	True if all-day shift
weight_modifier	integer	—
has_issues	boolean	True if breaking rules
cluster	string	Common strings define clusters
view_priority	integer	—
metadata	object	Object of key-value pairs
associate_id	integer	Linked associate (null = unassigned)
cached_position_ids	array	IDs of positions held
is_special	boolean	True if holiday/special
specialness_cache	string	Name of holiday/special date
associate_name	string	Cached associate name
changed_since_publish	boolean	—
team_metadata	array	—

Request:

Field	Type	Description
id	integer	Unique identifier
request_type	string	Type of request
start	string	Start datetime (with offset)
end	string	End datetime (with offset)
allDay	boolean	True if all-day request
show_to_team_associates	boolean	True if visible to non-admins
metadata	object	Object of key-value pairs
associate_id	integer	Linked associate
all_teams	boolean	Applies to all teams
cached_position_ids	array	IDs of positions during the request
team_ids	array	—
request_targets	array	—
approval_state	enum	0=pending, 1=approved, 2=denied, 3=acknowledged

Job:

Field	Type	Description
id	integer	Unique identifier
name	string	Job name
team_id	integer	Linked team
weight_modifier	integer	—
color	string	—
can_take	integer	—
view_priority	integer	—
can_be_empty	boolean	—
postcall_period	integer	—
positions	array	Positions qualifying for this job

Position:

Field	Type	Description
id	integer	Unique identifier
name	string	Position name
team_id	integer	Linked team
jobs	array	Jobs qualified for this position

Tour:

Field	Type	Description
id	integer	Unique identifier
associate_id	integer	Linked associate
start_date	string	YYYY-MM-DD (nullable)
finish_date	string	YYYY-MM-DD (nullable, inclusive)
job_ids	array	Qualified jobs
position_ids	array	Held positions
metadata	object	Extra key-value data
weight_modifier	integer	Workload/FTE (0–1)
tour_block_name	string	Block name

Holiday:

Field	Type	Description
id	integer	Unique identifier
name	string	Holiday name (e.g., “Christmas”)
date	string	YYYY-MM-DD

SpecialDate:

Field	Type	Description
id	integer	Unique identifier
description	string	Label name
date	string	YYYY-MM-DD
is_holiday	boolean	True if holiday

AI Generated Cells in Reports