Playbooks are ready-to-use, step-by-step prompts for bigger jobs than the short prompts on the Examples page. Each one includes the rules and definitions your AI assistant needs to give a consistent, reliable answer, so you get the same kind of result whichever assistant you use.

Playbooks describe the data they need, not specific tools. They keep working as the connector gains new tools. Each one shows which of its data is available today.
​

1. Copy the prompt and replace the text in brackets.

2. Paste it at the start of a new conversation with the Amilia SmartRec connector turned on.

3. Answer the assistant's scope questions. Playbooks ask you to confirm before doing the heavy work.

4. Check the numbers that matter before you act on them. AI assistants can make mistakes.

If you run a playbook often, save its prompt as reusable instructions in your AI assistant, where supported.

A short guided tour of your organization. It also works as a test script: each step reports whether it worked, so you can see at a glance whether the connector is activated and your administrator account can use it.
​

You are helping me get started with the Amilia SmartRec Connector for [organization]. Today's date is [today's date]. Program for step 4: [program name, or leave blank to use the program from step 3]. Work through these steps in order and keep each step short. 1. Connection check: list my programs. If you get an error, stop this step, show me the exact error message, and tell me to check that the connector is activated for my organization and that my administrator account has the connector permission, as described in the Set up page. 2. Organization overview: summarize my programs, including hidden and archived ones. Count the current, upcoming, and ended programs among those that are not archived. Report archived programs separately. Tell me which programs are hidden or password protected. 3. A closer look: pick the current program with the most activities that have a usable capacity and have not ended. Tell me how many activities it has in total, the price range, and how many are full, nearly full (80% or more), or under 60% full. Fill rate is reserved spots divided by capacity, so include each activity's reserved spots. Leave out activities that have ended and activities with missing, zero, or unlimited capacity, and tell me how many you left out of each kind. 4. People and households: for the program named above, count the unique participants and the unique households. - Unique participants: distinct person IDs among the people registered in the program. The roster can list the same person more than once, so do not use its row count. - Unique households: distinct account IDs on the same roster. An account is the billing household, so siblings in the same family count once. Count only the participants, not everyone on the account. - Also give the average number of participants per household. Read every page of the roster. 5. Next steps: based on what you found, suggest three questions I could ask next, one sentence each. Rules: - Read only. Do not change anything in SmartRec. - Show counts only. Do not show names, email addresses, or other personal details. - If a step fails, say which step failed and why, then continue with the next step. - If a step has no data to work with, such as no current program, no activity with a usable capacity, or nobody registered, say so and mark it Skipped, not Failed. - For every number, say what it counts. - End with a status line for each step: Worked, Failed (with the error), or Skipped.

A one-page health card for a program. It covers capacity and fill rates, enrollment, and cancellations, and it can compare two periods, such as this spring with last spring. It replaces the separate fill-rate monitor and period comparison prompts.

Build a program scorecard for org id [ORG_ID]. Scope - Program: [NAME] (id [PROGRAM_ID]) - Scope level: [category "X" | sub-category "Y"]. If I name only a program, the scope is all of its non-archived activities, including hidden and cancelled ones. - Today's date: [DATE]. Under-enrollment threshold: [60]%. - Single-scope scorecard. No comparisons with other programs, sub-categories, activities or periods. No new vs returning households, loyalty or retention metrics. Before you calculate anything 1. List every program whose name matches my request (including archived and hidden): ID, start/end dates, archived and visible status. Don't merge or swap programs with similar names. If my spelling doesn't match exactly, say which program you assume and why. 2. List the activities in scope (get_activities with showDetails, showCancelled and showHidden all true): ID, name, category, sub-category, capacity, reserved spots, status, start/end dates. 3. Flag anything odd: dates that don't match the category or sub-category name; duplicate-looking activities (same description, schedule, price and age range, or a prefix/suffix such as "*"); hidden activities; unlimited capacity (e.g. 2147483647); capacity that differs between sibling activities; activities that end after today or have already ended; activities with drop-ins enabled. 4. State how you will treat each flagged item (default: include everything in scope, exclude unlimited-capacity activities from fill rate only, footnote the rest). 5. STOP and ask me to confirm the scope. Do not calculate until I confirm. Data sources (summary only; no rosters) Make these four summary calls with the in-scope activity IDs, then stop calling: A. groupBy = the scope grain ("subcategory", "category" or "program"). This is the only source for scope totals AND for unique participants and unique households. B. groupBy = same grain, secondGroupBy = "activity". Per-activity detail only. C. groupBy = "registrationType", secondGroupBy = "activity". Gives sessions vs drop-ins and which registrations are priced. D. groupBy = "age" (scope level). Age bands only. - Seats: from get_activities (spotsReserved, maxAttendance), including cancelled and hidden activities. - Activities missing from the summary have 0 registrations. Say so. - Never sum per-activity or per-age uniqueParticipants / uniqueHouseholds. People repeat across activities. Scope-level uniques come only from call A. - The summary includes cancelled registrations in all counts and amounts. Active registrations = totalRegistrations − cancellations. Metrics 1. Volume: active registrations; participants = call A uniqueParticipants; households = call A uniqueHouseholds; registrations per participant (active registrations ÷ participants); participants per household. Label participants and households as "people / households with at least one registration, including people whose only registration was cancelled". The summary cannot separate them. 2. Capacity: sections and seats offered (activities evaluated, activities excluded); seats reserved; open seats; overall fill rate (flag if below the threshold); activities over capacity, under the threshold, and healthy. 3. Revenue: gross invoiced; net sales; net per priced registration (net sales ÷ active registrations with a price above zero); discount rate. Show gross → net as gross, less discounts, less cancelled, less credits. 4. Cancellations: cancelled registrations and the rate, cancelled ÷ (active + cancelled), shown as "5 of 39". Report cancelled activities (status Cancelled) separately, because the organization removed those, not families. 5. Audience: share of participants per age band (0–5, 6–10, 11–14, 15–17, 18+ unless I say otherwise), the dominant band with its share, and the median age. Use call D's uniqueParticipants per age. Rules - Registrations, participants and households are three different counts. Say which one each number is. One person can have many registrations, including drop-ins. - Seats reserved come from each activity's reserved spots, not from registration counts. - Open seats = capacity − reserved spots. Over-capacity activities count as 0 open seats. - Discount rate = discount amount ÷ gross invoiced. - Fill rate = reserved spots ÷ capacity, per activity. Exclude activities whose capacity is missing, zero or unlimited. Say how many you excluded and how many spots they hold. Never cap fill at 100%. - Over capacity = more reserved spots than capacity. - When duplicate-looking activities with zero registrations stay in scope, also give the fill rate without them as a sensitivity. - Priced registrations: registrations with no invoiced amount (e.g. drop-ins) are unpriced. State the assumption. - Show age bands only. Never show a date of birth, a name or an individual's age. Median age is allowed. Ages are completed years as of today, not at registration. - If any activity ends after today, say the scope is in progress next to the numbers. A scope in progress is not a decline. - This measures registrations, not attendance. Say so once. - No comparisons. Per-activity breakdowns inside the scope are fine. Checks (summary-only; report every mismatch) - Sum of per-activity registrations (call B) = scope total (call A); sessions + drop-ins = total registrations. - Sum of per-activity cancellations = scope cancellations. - Gross + discounts − cancelled − credits = net sales. (Discounts, cancelled and credits are negative in the summary.) - Sum of uniqueParticipants across age rows = call A uniqueParticipants. - Per activity: active registrations (registrations − cancellations) vs spotsReserved. Explain any difference (e.g. drop-ins don't hold seats). - Check that every registration type's cancellations are sessions, so the priced count (active sessions) is right. Output 1. One header line: organization, program, scope with date range, threshold, as-of date (with "in progress" if applicable). 2. Scorecard table (Metric | Value | Note) in this order: registrations, participants, households, registrations per participant, participants per household, sections / seats offered, seats reserved, fill rate (bold, with the threshold note), open seats, gross → net, net per priced registration, discount rate, cancellation rate, dominant age band. Add one line for activity health and one for cancelled activities. 3. Flagged activities: over capacity first, then under-enrolled, grouped by category and sub-category, with capacity, reserved and fill rate. Don't list healthy activities. 4. Three one-sentence findings that would change a decision (this scope only).

A ranked list of people to contact to fill under-enrolled classes. Each row explains why the person is a good fit, using only their own registration history. Contact always goes through the account owner: the parent for a child, or the person for an adult.

Build a targeted outreach list to fill under-enrolled classes in [organization]. Scope - Programs to fill: [program names, or "suggest them"]. - Feeder programs, if any: [programs whose participants often join the programs to fill]. - Today's date: [today's date]. Under-enrollment threshold: [60]%. Maximum contacts per class: [10]. Minimum score: [25]. Steps 1. Find the under-enrolled classes. Fill rate is reserved spots divided by capacity. Exclude classes with missing, zero, or unlimited capacity. Include only classes that have not started. Count a class once, even if it has several weekly occurrences. Rank the programs by open spots and show me the list. 2. Stop and ask me to confirm which programs to work on. Do not look up any people before I confirm. 3. Build the candidate pool from people registered in the confirmed programs, in current and past activities, and in the feeder programs I named. Use each activity's own roster to see which class a person took, because a program roster does not say which activity. Match people by person ID, not by name. A person can appear on several rows, so count each person once. 4. Score each person for each target class, using only facts from their own registration history: - Took the same class or subcategory before: +40 - Registered in the same category, in a different class: +30 - Registered in a feeder program I named: +20. If the person also earns +30, keep only the higher of the two. - Took part before, but not in the last 12 months: +10 - Recent activity: up to +15, halving for every 6 months since their last registration Keep people at or above the minimum score. Remove anyone already registered in the target class. 5. Look up contact details only for the people you kept, up to the maximum per class. Contact the account owner: the parent for a child, the person themselves for an adult. The roster gives each person's account owner ID, so take the email and phone from the owner's own record. Never use a child's own email or phone. Use the date of birth only to check the class's age range, then discard it. Remove people outside the age range. If a date of birth is missing, keep the person and show the age as "not on file". 6. Sort by class, then by score, highest first. Privacy rules - Never show a date of birth, health information, or financial details. Show age in years only. - Do not look up contact details for anyone who is not on the final list. - Reasons use only the person's own registration history, never a guess about them. - Label the output "Internal: outreach team only". - Read only. Do not change anything in SmartRec. Output 1. A header: organization, threshold, programs worked, today's date, and the internal-use label. 2. One table per target class with these columns: participant, age, account owner, email, phone, score, score breakdown (for example "Same subcategory +40; Recent +12 = 52"), and a one-line reason a caller could read aloud. 3. Leave the phone blank when there is none. If an account cannot be found, write "Look up in SmartRec" and keep the row. 4. If no one qualifies for a class, say so. Do not pad the list.

* Last updated in November 2026