ai-adoption · deliverable

Kit 6 — “AI at Your Desk” · department-level AI adoption program

One-line: a 60-day program that moves 300 desk workers (Finance, HR, Ops-planning) from AI-curious to governed weekly use — anchored in three personas with real Monday-morning tasks, taught through an interactive prompt-pattern trainer, and wrapped in manager enablement + GDPR-aware guardrails.

Why this kit exists (and how it differs from Kit 4)

Kit 4 teaches one team to verify AI output at the moment of send. Kit 6 is the wider question every HR/L&D generalist gets: how do general employees — not trainers, not developers — integrate AI into daily work? Different audience, different arc: adoption stages (Curious → Capable → Confident → Governed), audience segmentation with real task math, a manager enablement layer, and an AVG/GDPR-aware guardrail page.

Anti-generic mandate: the program is about three people, not about “AI adoption” — Marta (Finance, invoice-line anomaly spotting, 3.5h → 1h), Jeroen (HR, job ads + CV screening, 15 → 6 min/CV), Sanne (Ops, schedule-variance summary, 90 → 20 min). Every example prompt, huddle script and success story maps to one of their tasks; anything that doesn’t, gets cut.

Artifacts

File	What it is
`01-adoption-program.md`	Program design memo: persona-anchored segmentation with before/after time costs, 4-stage adoption arc with exit criteria, pilot → wave → champion rollout, proposed objectives (≥60% weekly active by day 60, zero breaches, ≥2h/week self-reported savings).
`02-prompt-patterns.html`	Flagship: interactive prompt-pattern trainer. 4 reusable patterns (Role-Task-Format · Few-shot sandwich · Step-by-step decomposer · Critic-loop), each with editable department-specific demo prompts (pick Finance/HR/Ops on entry — all 12 variants pre-fill), pattern anatomy labels, try-it sandbox, and a 3-question check gating the next pattern. Progress bar + printable cheat sheet. Pattern checks count first-try answers. Then a desk check (seven items: traffic light before the rule is shown, then a pattern) and a rewrite check (build a prompt from role, task, format and input). Reads `adoption-data.js`.
`3d.html` + `adoption-arc.js`	Helios office floor: the same seven items go to the external AI terminal, the redaction table and approved tool, or the locked cabinet.
`03-manager-pack.md`	Team-lead enablement: “first team AI habit” playbook (pick ONE task, not ten), four 5-minute huddle scripts matched to the four stages, red-flag list of what never enters external tools.
`04-governance-basics.md`	Employee-facing guardrails: data-classification traffic-light table, the disclosure norm (“client-facing AI text gets a human name on it, always”), escalation path + written-exception principle.

Design decisions worth naming

The trainer teaches by doing, not by listing. Each pattern has anatomy → live editable example for your department → sandbox → 3-question recognition check before the next pattern unlocks. Four patterns deeply practiced beat forty patterns skimmed.
Department switching is a feature, not a setting. All 12 demo prompts (3 departments × 4 patterns) are pre-filled with real task content, so the trainer demonstrates the segmentation claim instead of asserting it.
Managers get scripts, not advice. The four huddle scripts are word-for-word runnable, including the safety moves (“go first with your own mess”) — adoption dies in the gap between “we should” and “say this.”
Governance is one page on purpose. The traffic-light table, the disclosure norm and the escalation path fit on one page because guardrails nobody reads protect nothing.
Continuity: Helios is the Kit 3 client, now upskilling its office workforce; the measurement discipline (self-reported figures labeled as such, no inflated L4 claims) carries over from Kit 5.

Market mapping

Role / org type	What this kit demonstrates
Précon-type consultancies (Copilot/Claude training)	A deployable adoption program with persona segmentation and a working trainer artifact
Ricoh-type AI-literacy campaigns + skills-based learning	Pattern-based skills training with gated competence checks, not content consumption
Exact-type “AI leverage” roles	Knowledge-worker task math (before/after time costs) and habit-based rollout
FareHarbor-type behavior change	Stage-gated adoption arc, champion floor-support model, manager scripts
HR/L&D generalist roles	The full stack: program memo + training artifact + manager enablement + GDPR-aware guardrails

Honest limitations

Time-savings figures are self-reported estimates from the (fictional) pilot and labeled as such; no stopwatch study is claimed.
The trainer runs fully offline and therefore cannot track completion or scores across learners — LMS/xAPI packaging is an implementation-phase decision, noted in the memo.
“Zero policy-breach incidents” is an objective with an instrument, not a promised outcome; the escalation design exists precisely because breaches must be survivable to be reportable.
Built solo in one session; the 12 department variants are written but not learner-tested. The huddle scripts are written to be run verbatim by someone else, which is the first testable step.

Build notes

Single-session build. Quality gates: node --check on all script blocks, all 3 departments × 4 patterns verified to pre-fill correctly, Chromium headless render smoke-test, zero placeholder markers left in any file. AI-assisted drafting disclosed per artifact; pattern selection, examples, targets and final copy are human-owned.

by I Wayan Usada · runs fully offline · Helios Facilities Group is a fictional client (Kit 3 universe)

STATUS: DELIVERED 2026-08-29 — final artifact published as part of the six-kit portfolio suite (kits 1–6).

Process deliverables

storyboard.md — Text-based storyboard — the scene plan written before build
visual-mockups.md — Visual mockups & style guide — mood, type, key screens, component decisions