Lesson 03 — A2 Delegation: Splitting Research And Admin Work
Standard: — · Bloom's: — · Structure: —.
Notice the Say-See-Do cycles running on Priya S.'s actual work — not a canned exercise — and that every capability claim is cited to the frozen doc-set. The exit ticket climbs Bloom's to —, and the lesson closes by writing to the ledger.
Lesson 03 — A2 Delegation: Splitting Research And Admin Work
Learning objective (Bloom's: Apply): Apply a deliberate delegation split — what goes to Claude, what stays with you, and how you'll check the handoff — to two real chunks of your business: interview-notes research legwork, and pitch-tracker admin.
Standard: AIHC.1.A2 (Delegation), Band 1. Touches X9 (ask/act boundary: don't sit on work Claude could safely run; don't silently hand off a judgment only you hold) and X7 (your attention is the scarcest resource in a one-person business — delegation is how you protect it).
Incoming mastery (from Lesson 01): AIHC.1.A2 = 0.35 — you hand work over reactively, without naming the split.
Structure: PBL — your real pitch tracker and a real background question from "The Last Mill" are today's cases. Band 1; timing-tolerance honored: 4 cycles, each a full pause point.
Trust boundary named up front (R13): two features this lesson uses have real limits worth knowing before you lean on them. File creation/editing is on by default on your plan but capped at 30MB per file, both uploads and downloads [S08]; Research, once toggled on, "searches both connected internal context… and the broader web" [S11] — you have no integrations connected, so today it's web-only, but that's a fact worth knowing before you ever connect one.
Cycle 1 — What a delegation split actually names
- SAY: A split names three things: Claude-side (research legwork, transcription, spreadsheet upkeep), Priya-side (which source to trust, which anecdote leads, final wording), and check-method (how you'll confirm the handoff before it ships). Skipping this either wastes your attention on work Claude could run, or silently hands off a call only you can make.
- SEE: A described split-table, generic newsroom version, three columns populated with examples of each.
- DO: Draft this exact three-column table for your real current workload: one real interview-notes task and one real pitch-tracker task, split honestly. (Evaluation: checks all three columns are filled per task, not just two.)
⏸ Pause point. Banked: one real three-part split table.
Cycle 2 — Delegating pitch-tracker admin, for real
- SAY: Claude can create and edit working spreadsheets — Excel files "with working formulas" [S08] — which is genuinely admin work: recomputing days-since-pitched, sorting by follow-up deadline. None of that touches which pitch you actually prioritize; that stays yours.
- SEE: A described before/after of a real pitch-tracker row: the same data, now with a
=TODAY()-[date pitched]formula Claude added, highlighted. - DO: Using Cycle 1's split, have Claude turn your real pitch tracker into a working
.xlsxwith at least one live formula (Settings > Capabilities must have "Code execution and file creation" enabled — on by default on your plan [S08]). Review the output against your split before accepting it. (Evaluation: checks a working formula exists and that she logged a review step, not just an accept-on-sight.)
⏸ Pause point. Banked: one real working pitch-tracker file.
Cycle 3 — Delegating research legwork, keeping the check-method explicit
- SAY: Research runs multiple interconnected searches and returns citations you can check back against [S11] — real legwork delegation, not answer-delegation. Enable it via the "+" button (bottom left of the chat box) → "Research" [S11]; note the hard prerequisite — "You must have web search turned on for research to function" [S11, S09]. Delegating the search is not the same as skipping the check — that's next lesson's job (B1); today you're only naming which category this output falls into.
- SEE: A described Research transcript backgrounding one real source mentioned in her interview notes, citations visible.
- DO: Run Research for real on one background question from "The Last Mill." Log which of Cycle 1's three categories the output falls into (Claude-side legwork? something you still need to weigh?). (Evaluation: checks the log names a category, not just "it worked.")
⏸ Pause point. Banked: one real Research run, categorized.
Cycle 4 — Making the split standing
- SAY: Like last lesson's fixed/open rules, a delegation split can live in your project instructions too — so you're not renegotiating who-does-what every single task.
- SEE: A described project-instructions panel with one new line added, naming the standing split.
- DO: Write one project instruction sentence naming your standing delegation split for this project, and save it. (Evaluation: checks it references at least the Claude-side/Priya-side distinction, not just "help me with admin.")
⏸ Pause point. Banked: one standing delegation instruction, saved. Four cycles complete.
Independent at-bat
Take one real admin task you haven't delegated yet — an overdue line in your pitch tracker. Delegate it cold using the three-part split, no table given this time.
Exit ticket (climbing to Apply — the objective's level)
- (Remember) Name the three parts of a delegation split.
- (Understand) Why is "whatever's fastest today" a worse delegation rule than a named split, specifically for someone running a one-person business? One sentence.
- (Apply) New task: "chase down whether the mill closed in 1987 or 1989." Split it into the three parts.
- (Apply — objective level) Take one task from your real pitch tracker you haven't delegated at all yet. Write its three-part split and run it.
Ledger write
ledger_append:
learner_id: L3-CONS-PRIYA
lesson_id: L3-cons-priya-03-A2-delegation
standards:
- {standard: AIHC.1.A2, mastery_before: 0.35, mastery_after_simulated: 0.60}
bloom_reached: apply
auto_score: ""
self_score: ""
structure_used: PBL
referents_used: []
journal_prompt: >
Which task did you catch yourself about to hand off blind before you'd named the split — and
which column (Claude-side, Priya-side, check-method) did you have to think hardest about?
next_lesson_seed: "A2 moving fast; keep heavy scaffolding on A3 (her lowest, 0.20) next"
RUBRIC SELF-AUDIT
| # | Indicator | Verdict | Evidence |
|---|---|---|---|
| R1 | One Bloom's objective, ≥1 named standard, learner-visible | PASS | Apply-level objective stated; AIHC.1.A2 named |
| R2 | Every product claim traces to the frozen doc-set | PASS | S08 (file creation, 30MB limit, capabilities toggle), S11/S09 (Research mechanics + web-search prerequisite) cited |
| R3 | 3–6 SSD cycles complete | PASS | 4 cycles, each SAY/SEE/DO |
| R4 | Each SEE ground-truth-verified | PASS | SEEs describe the actual working-spreadsheet formula and Research transcript mechanics named in-SAY |
| R5 | Every DO on real work; fictional only inside SEEs | PASS | All 4 DOs + at-bat on her real notes, tracker, project; generic split-table example is inside Cycle 1's SEE only |
| R6 | Media doctrine | PASS | Static/described SEEs only |
| R7 | Exit ticket 3–5 Qs, Bloom's-climbing | PASS | 4 questions, Remember→Apply(objective) |
| R8 | Ledger write complete | PASS | Standard, both scores , journal prompt present |
| R9 | Scaffolding + independent at-bat | PASS | Split-table given in cycles, absent in at-bat |
| R10 | Referents elected-only, flavor-only | PASS | No referent forced this lesson (none needed for the point being made) — vacuously clean, none misused |
| R11 | Timing-tolerance pause points | PASS | Pause point after each of 4 cycles |
| R12 | Non-replication | PASS | No DO or exit question overlaps Lessons 01, 02, or the exemplar |
| R13 | Consumer trust boundaries named | PASS | Named 30MB file-creation limit [S08] and Research's connected-context scope [S11] up front, before either feature is used |
Escalation: all load-bearing indicators PASS → auto-ship.