Sample Build Sprint Pack
What the $10,000 Build Sprint delivers.
A sample Sprint Pack for Pathly on Northstar College, a fictional institution. Every Sprint ends with the same set of deliverables, built on your catalog and synthetic or de-identified data instead.
apply.northstarcampus.com / find-a-program
In · Applicant message, 11:12 pm
I do medical billing and want to move into healthcare IT. I can do two evenings a week and need to start before spring. Will my community college credits transfer?
Out · Pathly
Health Information Technology Certificate
Evenings, Tue and Thu · Starts Jan 13 · 2 terms · $3,840
IT Support Fundamentals
Online, evenings · Starts Jan 6 · 10 weeks · $1,950
Up to 15 credits can transfer. Transfer Credit Policy · §2
Advisor call booked · Thu 4:30 pm CT · lead in CRM
Then the transcripts: transfer form pre-filled, 12 credits suggested for the registrar.
Run it step by step on the Pathly page.
Try it step by stepIllustrative demo on a fictional institution. Not client work.
Before the clock starts
A setup week, so the two weeks of build are all build.
Access, data and test cases are the slow parts, and they sit on your side. We sort them out first; no build time is used until they are in place.
- 01Agree the buyer outcome, the 2–3 capabilities that deliver it, and what done looks like
- 02Agree the eval cases with your reviewer, 30 to 50 of them
- 03Choose the data: your catalog and published pages, plus synthetic or de-identified data
- 04Grant staging and repository access, and name a reviewer for about 2 to 3 hours a week
- 05In the setup week we map your integrations and agree what the Sprint touches
- 06We follow your coding standards, tests and CI; your engineers review and merge every pull request
- 07Half at signature, half when the Sprint Pack is delivered
- 08Your time: about 2 to 3 hours a week from a product owner, plus an engineer to review pull requests
01 · Quality evals
Quality is a test set your team reviews, not a feeling.
We build the test set with your team from real questions, run it before and after changes, and report the result case by case.
| Test set | What is checked | Pass criterion |
|---|---|---|
| Program recommendations | Each pick exists in the catalog, fits the stated goal, schedule and budget, and links its catalog page | Your reviewer marks each pick correct; the pass threshold is agreed before the Build Sprint |
| Cost and date answers | Every figure matches the published page it cites | No uncited figures |
| Policy answers (transfer credit, deadlines) | Answers quote the policy and point to it | No answer beyond the published policy |
| Advisor booking | Lead written to your staging CRM or sandbox with the applicant's stated goals | Every booked call recorded |
| Spanish conversations | Same checks, in Spanish | Same pass threshold as English |
02 · Safety tests
The questions your customers' security review will ask, tested.
Run on synthetic or de-identified data in a staging copy of your product.
| Test | How it is tested | Expected behavior |
|---|---|---|
| Cross-tenant access | Questions crafted to pull another institution's catalog or leads | Nothing returned from another tenant |
| Prompt injection | Instructions hidden in user messages and catalog text | Instructions ignored, attempt logged |
| Personal data | Requests for another applicant's details or aid status | Refused and handed to staff |
| Out of scope | Medical, legal or financial advice requests | Declined with a pointer to a person |
03 · Data-flow sheet
Where every piece of data goes, on one page.
Written so your team can hand it to a district or institution as part of a review.
- 01
Catalog and policy pages
Published pages from your product, indexed per institution
- 02
Retrieval
Only that institution's content, with page references
- 03
Model endpoint
Your chosen model and region; no training on the data
- 04
Answer with citations
Shown to the applicant; personal questions go to staff
- 05
CRM and calendar
Lead and booking written through your existing APIs
- 06
Logs
Conversations, citations, handoffs and approvals logged in your environment, under your retention policy
04 · Cost and quote
You know the margin and the price before you commit.
A cost-to-serve estimate per conversation on two models, with the assumptions shown; a packaging recommendation: include it, or sell it as a paid AI tier; and a fixed-price, fixed-duration quote for production. The Sprint Pack is yours whether or not you continue.
What you keep
- 01The feature working inside your product: in your staging environment, behind a feature flag, delivered as pull requests
- 02The eval report: quality and safety test sets with results
- 03A cost-to-serve estimate per conversation on two models, and a packaging recommendation: include it, or sell it as a paid AI tier
- 04The data-flow sheet for your customers' security reviews
- 05A fixed-price, fixed-duration quote to take it to production
- 06AI screens designed against WCAG 2.1 AA
- 07A customer notice you can send to your institutions
Get a Sprint Pack for your own product.
Bring one product and one workflow. In 30 minutes we agree which module fits and what the Build Sprint would test.
Calls run in US Central time. No generic AI presentation: we discuss your product, users and production constraints.