Larkspur Arts Collective
How the organization's six tools connect, where they break, what to fix first, and where AI can safely help.
About this sample. Client reports are confidential, so this one describes a fictional organization. The people, systems, domains, and numbers are invented. The structure, depth, and kind of findings match what clients receive at the end of a diagnostic.
The tools work. They don't work together.
One page for the board and the executive director.
Larkspur runs on six tools. Each one does its own job well. But none of them shares data with the others, except the donation form. Staff fill the gaps by hand, every week.
One urgent risk
Medical and emergency-contact information for youth class students is copied into a shared spreadsheet. Eleven people can open it, including two former contractors. This should be fixed this month, whatever else the board decides.
Recommendation
Phase 1 (now, about $500–1,000): move youth medical data out of the spreadsheet and remove old access.
Phase 2 (one integration sprint, about $9,000–15,500): connect ticketing and class registration to the donor CRM, clean up duplicates, and sync the email list. This ends the weekly exports and gives Larkspur one record per patron.
AI (after Phase 1): start the three approved uses in section 6, under a one-page AI use policy. The patron chatbot waits until Phase 2 is done.
Six tools, three domains, no single sign-on
Every tool a patron or staff member touches, and how data moves between them today.
What the map shows
- Only one of five data paths is automatic: donations from the website form into the CRM.
- The CRM is meant to be the single record of every patron. In practice it's complete only on the Monday after the weekly export.
- Each public-facing tool has its own login. A patron who does all three things has three accounts and three passwords.
One patron, one season, before and after
A typical path we walked through using test accounts.
Today
After the recommended fixes
The class step stays partly separate on purpose. Replacing a tool that teaching artists like, just to remove one login, would cost more than it saves.
What we found, by severity
Each finding comes from the review, a staff interview, or the journey walk-through.
Youth medical data in a shared spreadsheet
Emergency contacts and medical notes for 140 youth students are copied each term into a spreadsheet shared with 11 accounts. Two of those accounts belong to contractors who left last year.
Weekly manual exports
Staff export ticket and class data every Monday, clean it in a spreadsheet, and import it into the CRM. This takes about 6 hours a week, about 300 hours a year, and fails when the person who does it is out.
Duplicate patron records
21% of CRM records duplicate another record. Most come from patrons who use different emails for tickets and classes. Duplicates split giving history and distort reports to funders.
Slow thank-you letters
Letters are written by hand once the weekly import finishes. The average wait from gift to letter is 19 days.
AI in use with no policy
Four staff members use free AI chat tools for grant drafts and donor letters. At least one has pasted donor names and gift amounts into a personal account.
Slow mobile pages
The events page loads slowly on phones, mostly because of uncompressed photos. Staff can fix this with a short guide.
What to fix, in order
Ranked by risk first, then by hours saved per dollar. Costs are ranges until the work is scoped.
| # | Fix | Effort | Cost range | What it unlocks |
|---|---|---|---|---|
| 1 | Move youth medical data into the class platform's protected fields. Delete the spreadsheet. Remove access for former contractors. | 1 day | $500–1,000 | Removes the largest risk. Phase 1 |
| 2 | Connect ticketing to the CRM with a nightly sync. Turn on guest checkout. | 3–5 days | $3,000–5,000 | Ends half of the weekly export. Removes one patron login. Phase 2 |
| 3 | Connect class registration to the CRM with a nightly sync. | 3–5 days | $3,000–5,000 | Ends the rest of the weekly export. Phase 2 |
| 4 | Merge duplicate records and set matching rules (email, then phone). | 2–3 days | $1,500–2,500 | One record per patron. Accurate funder reports. Phase 2 |
| 5 | Sync CRM segments to the email newsletter automatically. | 1–2 days | $1,000–2,000 | No more monthly uploads or double sends. Phase 2 |
| 6 | Export template from the CRM for the bookkeeper's monthly entries. | 1 day | $500–1,000 | Ends re-keying of donations. Phase 2 |
| 7 | Photo compression guide for the website. | 2 hours | Staff can do it | Faster event pages on phones. Anytime |
Phase 2 total (fixes 2–6): $9,000–15,500 as one fixed-scope integration sprint.
Where AI helps, and where it doesn't yet
Start these only after Phase 1, and after staff adopt the AI use policy below.
Safe to start
- Draft donor thank-you letters from CRM fields (name, gift, program supported). A staff member reviews each letter before it goes out. Goal: letters within 3 days.
- Summarize class feedback surveys each term into themes for the program director.
- First drafts of grant report sections from attendance and program data that contains no personal details.
Not yet
- A chatbot that answers patrons on the website. Today it would answer from scattered, out-of-date data, and give wrong dates and prices. Revisit after Phase 2.
- Anything that touches youth medical data. Not now, and not without legal review.
Policy and training
- One-page AI use policy: no donor, patron, or student personal data in personal or free AI accounts. Use one organization account with data-retention controls turned on.
- One 90-minute staff session: the policy, the three approved uses, and how to review AI drafts.
How this diagnostic was done
- Read-only access to all six tools. Nothing was changed during the review.
- Two staff interviews: the development director and the box office manager.
- One end-to-end journey walked with test accounts: ticket, donation, class registration.
- Review of the export spreadsheets and their sharing settings.
- A 60-minute walkthrough of these findings with the executive director, before this report was finalized.