04 — QUALITY FRAMEWORK: Advisory Onboarding Kit v2

Level 3 design. This is a quality framework, not a checklist. Fix what you can. Escalate what you can't. The goal: Kathryn only sees things that require her judgment — everything else is resolved before it reaches her.

The Standard

A good output from this kit looks like this:

• The client reads it and knows exactly what's happening. No ambiguity, no consultant-speak, no sections that only make sense if someone explains them.
• The practice owner hands it over with confidence. It sounds like their practice, uses their names and tools, and reflects how they actually work — not how a generic advisory firm works.
• The three artifacts tell the same story. Same trigger, same cadence, same first win, same completion gate. A client who reads all three never hits a contradiction.
• The maturity layers are honest. Level 1 is documented. Level 2 names specific AI-assisted workflows that exist or are being built. Level 3 describes what's possible when the practice is ready — not what's happening today if it isn't.
• Nothing is made up. Every detail traces to the extraction conversation or the reference data. No fabricated examples, no invented team members, no assumed pricing.

A bad output looks like this:

• Bracket-filled. [Client Name] is fine (per-engagement). [TBD] or [to be determined] means the extraction was incomplete or the output skill didn't do its job.
• Generic. Reads like it could be any advisory practice. Nothing practice-specific. The member would say "this isn't mine."
• Wrong structure for the practice. An institutional firm's roadmap looks like a solo practitioner's. A coaching practice's cadence is seasonal when it should be activation-based.
• Advisor-internal content in client-facing output. Pricing details, internal success metrics, constraint data, emotional notes from extraction.
• Consultant language. "Leverage," "synergy," "circle back," "touch base." Or AI-pattern language: "It's not X — it's Y."

What to Fix Yourself

These are issues you catch during production. Don't flag them. Fix them.

Proper noun misspellings. Check against reference data. ATOM not Adam. Karbon not Carbon. TaxDome not Tax Dome.

Generic content that should be specific. If the extraction provided a concrete welcome experience and the roadmap says "you'll receive a welcome package" — rewrite it with the specifics.

Inconsistencies across the three artifacts. If the roadmap says 4 meetings per year but the action plan references quarterly estimates as a separate service — one of them is wrong. Check the extraction data and fix.

Missing maturity layers. If Level 2 or Level 3 is absent from an artifact, write it. If you don't have enough practice-specific data to make it concrete, calibrate honestly: "As we build these systems together, here's what becomes possible."

Audience-boundary violations. Pricing structure in client-facing output — remove. Internal constraint notes — remove. Advisor frustration or emotional processing — remove.

Template language that survived. Anything that reads like instructions to the AI rather than content for the client — rewrite.

What to Escalate to Kathryn

These are genuine judgment calls. Note them clearly in the output.

Ambiguous extraction data. The member said two contradictory things and the conversation didn't resolve it. Example: "Everything is included" AND "bookkeeping cleanup is separate." You can't fix this without knowing which one the member meant.

Practice model mismatch. The extraction says "lean" but the data looks institutional (5+ team members, multiple tiers). The roadmap structure depends on getting this right.

Special-sauce boundary. You're not sure whether something is structural (goes in the kit) or proprietary (stays with the practice). Err toward proprietary and flag it.

Maturity layer realism. The extraction suggests the practice is ready for Level 3 but nothing in the data supports it — no team, no tools, no automation. Writing a Level 3 section would be aspirational fiction.

Cross-member pattern that needs a decision. Two members have nearly identical completion gates. Is that because the universal spine is doing its job, or because the extraction wasn't specific enough? Kathryn decides.

Cross-Member Quality Intelligence

As the kit runs for more members, the quality framework learns:

After each run, note:

• What manual fixes were needed? (These become kit fixes via Mode 2)
• What did the practice owner change by hand? (These reveal where the kit's assumptions were wrong)
• What escalations came back from Kathryn? (These become new self-fix patterns)

Use prior runs to improve current ones:

• If Rob's roadmap needed a different structure for the maturity section and this member is similar — build it right the first time.
• If Bev's extraction missed the team-role detail and the output suffered — push harder on that during the current extraction.
• If a previous action plan was "too vague" per Kathryn — make this one more concrete.

Hard Stops

These are non-negotiable. If any is true, the output cannot ship. Period.

1. Fabricated data. Anything that doesn't trace to the extraction or reference data.
2. No extraction completed. The kit requires per-member offer data. No data = no output.
3. Special-sauce leak. A practice's proprietary approach baked into the universal spine.
4. Client-facing AI jargon. "Leverage," "synergy," "It's not X — it's Y" in the roadmap.
5. No completion gate. The roadmap has no defined endpoint.
6. No first win. The roadmap doesn't name a concrete, dollar-adjacent result in 30 days.

Everything else is fixable during production or escalatable to Kathryn.

The Client Perspective Test

Every section of every artifact must pass:

"If the client read this cold, with no advisor present, would it (a) make sense, (b) feel respectful of their time and intelligence, (c) tell them what to expect, what they're committing to, and what they'll get?"

If any section fails any of those three — fix it. If you can't fix it without more information — escalate with a specific question, not a vague flag.

Cross-references

• File 01 — universal process (design constraints live here)
• File 02 — terminology lock (words that mean specific things)
• File 03 — golden example (the quality benchmark)
• File 05 — output intelligence (where production happens)
• File 06 — extraction conversation (where the data comes from)