Dave Criswell Systems · Operations · Revenue
Playbook & AI

The Select Playbook

63 plays (and counting) · 6 areas · 10 run as skills

In any organization that has grown over decades, much of what keeps it running is knowledge that lives with experienced people. The Playbook writes it down, one repeatable “play” at a time, so the work doesn’t depend on who happens to be available. The most repetitive plays then become AI skills, so the documentation and the automation are the same thing.

The problem
Select has run for more than 30 years on the experience of the people who built it. As the club grew, more of that knowledge needed to be written down, so getting something done didn’t depend on finding the right person to ask.
What I built
An operations playbook of 63 plays across six areas: registration, scheduling, staffing, data, marketing, and leadership. Every play follows the same eight-part structure, and ten of them run as saved AI skills.
What it changes
Anyone can see how a process works, who owns it, and what starts it, without tracking down the one person who knows.
Built to hold up
Every play uses the same eight-part structure, and new plays come in through one update skill with a person reviewing, so the Playbook grows without turning into a pile of one-off documents.

Anatomy of a play

  1. 01 · PurposeWhy it exists
  2. 02 · OwnerWho’s accountable
  3. 03 · TriggerWhat starts it
  4. 04 · ToolsSystems involved
  5. 05 · StepsWhat happens, in order
  6. 06 · RulesWhat must always, or never, happen
  7. 07 · SourceWhere the data comes from
  8. 08 · StatusDraft, live, or automated

A sample play

Play · DataGame results dashboard
Live · runs as a skill
Purpose
One view of how every team is doing, across every league the club plays in.
Owner
Marketing & Systems Operations
Trigger
A new results export from the leagues.
Tools
League results exports, a team-mapping sheet, a saved Claude skill, and an interactive HTML dashboard.
Steps
  1. Collect the latest results from each league automatically through Claude skills.
  2. Show every result from the club’s perspective: our score first, then a win, loss, or draw.
  3. Match each game to a club team using the mapping sheet, and roll leagues up into groups such as “State” and “ECNL.”
  4. Rebuild the dashboard, with filters for league, region, age group, and team.
Rules
Never edit the source exports. Flag any team that won’t match instead of guessing.
Source
League results exports and the club’s team-mapping sheet.
Status
Live. Runs as a saved skill, so a new export rebuilds the dashboard in one step.

Plays that run as skills

Useful, not magic

Most of my AI work isn’t glamorous. It takes a recurring, messy input and turns it into a clean, checked output, so people can spend their time on players and families instead of spreadsheets. A few of the plays that run as saved skills:

  • League schedule importLeague export → registration-platform import file, plus a punch list of team names that need a human
  • Season schedule exportA league website with no download button → a full-season workbook of about 475 games, with a refresh mode
  • Coach shift importSchedule PDFs, grids, and screenshots → shift import file, merged and padded
  • Pre-payroll exception reportTimesheet export → missing clock-ins and large variances, by region
  • Registration follow-up trackerProgram exports → who has and hasn’t registered, by team, with a follow-up list
  • Process documentationA rough outline → a standardized process document, filed where people can find it
  • Camp family recapThe day’s notes → a draft family email and social post, reviewed before anything goes out
  • Inbox themesA full inbox export → the questions families actually ask most, grouped by theme

How it stays current

  1. Step 1Work happensEveryday conversations and tasks in Select’s Claude workspace.
  2. Step 2Claude spots a candidateIt’s set up to notice repeatable systems and suggest them for the Playbook.
  3. Step 3The update skill places itA Playbook update skill decides where it belongs, a new play or an update to an existing one.
  4. Step 4A person reviews itNothing goes into the Playbook without someone checking it first.
  5. Step 5The Playbook growsThe most repetitive plays become skills of their own.

The workspace behind it

I set up Select’s Claude Team account around a few deliberate decisions:

  • A written AI use policy first

    We agreed on the rules before anything else: only aggregated or pseudonymous player data, read-only lookups through connectors, and drafts that a person reviews before anything is sent.

  • Connected to where work happens

    Connectors to Google Workspace, HubSpot, ClickUp, and meeting notes, plus Zapier for other apps, so answers come from the club’s real systems instead of copy and paste.

  • Recurring work saved as skills

    The same job gets done the same way, whoever runs it.

  • Built to grow the Playbook

    Claude watches for repeatable systems and suggests them as Playbook candidates, so documentation happens during the work, not after it.

  • Leadership and admin first

    Access started with leadership and admin roles, so we could learn what works before widening it.

Guardrails I build in

  • Drafts only. Nothing goes to families automatically.
  • Never overwrite a person’s edits.
  • Flag conflicts instead of quietly resolving them.
  • Verification checks on every import.
  • Limit what personal data an AI tool can see.
  • Write the process down, so it outlives the person who built it.