How I work

A five-phase methodology, tuned for AI-assisted L&D.

Every project in the case studies uses the same underlying framework. It borrows from ADDIE and Kirkpatrick, leans on Mager for objectives, and updates the development phase for what AI tooling now makes practical — hand-built interactions and client-grade data pipelines on small-team budgets.

01 Discover 02 Design 03 Build 04 Deliver 05 Measure

Why the framework matters — especially now

AI tooling has made it easier than ever to produce something that looks like training. The constraint has moved. The hard part isn't generating content anymore — it's deciding what to build, how to measure whether it worked, and whether training is even the right intervention. This framework is where those decisions get made on purpose, not by accident.

Phase 01

Discovery & analysis

Learn what the business actually needs before promising what training can deliver.

What happens

→ Stakeholder interviews with sponsors, managers, and target learners
→ Performance analysis — where exactly is the gap between what people do and what the business needs?
→ Root-cause check — is this a knowledge problem, a skill problem, or a system/environment problem?
→ Success metrics agreed before any content decisions happen

What comes out

• A one-page problem statement signed off by the sponsor
• Baseline metrics (the numbers we're moving)
• Learner profile and context map
• Go/no-go recommendation — sometimes the answer is "this isn't a training problem"

""No training" is a legitimate recommendation. If it's a process problem, say so on page one — the sponsor will respect you for it later."

Heuristic

Phase 02

Design

Make the hard structural decisions on paper before a single screen gets built.

What happens

→ Write measurable terminal objectives (Mager-style) tied to the business metrics
→ Pick a modality against a checklist — classroom, blended, code-built, Rise, microlearning — using evidence, not taste
→ Storyboard scenarios and branching with realistic detours, not just happy paths
→ Design the assessment before the content — what would prove the objective was met?

What comes out

• Design document covering objectives, modality rationale, assessment strategy
• Scenario map with branching logic
• Rough wireframes or click-through sketches
• Risk list — what's most likely to break in build?

"The modality decision is where most projects fail. A "just make a Rise module" instinct is usually a design problem dressed as a tooling problem."

Heuristic

Phase 03

Development

Prototype early, iterate in small loops, and pick tools based on what the learner needs — not what's in the department license.

What happens

→ Prototype the highest-risk interaction first (the scenario engine, the simulator, the branching module) — not the intro screen
→ Rise 360 for linear rapid build; code-built HTML/CSS/JS when the interaction matters or data fidelity is required
→ AI-assisted scaffolding in Cursor + Claude for code-built work — compresses weeks into days
→ Two rounds of learner testing minimum before handoff — cheaper than fixing it after launch

What comes out

• Working prototype of the riskiest piece, shown to stakeholders in week 2
• Production build against the agreed modality
• Accessibility audit (target: WCAG 2.1 AA)
• Facilitator and IT deployment notes

"If the first demo is a polished intro screen, you built the wrong thing first. Show the hardest part working, or don't show anything."

Heuristic

Phase 04

Delivery & change

Treat the launch as a rollout project, not a file transfer.

What happens

→ LMS or host deployment and smoke test
→ Train-the-trainer where people facilitate, SME briefings where they don't
→ Communications plan with managers — people need to know why this exists and when they're expected to complete it
→ Support channel for the first 30 days so early friction gets fixed fast

What comes out

• Deployed experience with access pathways verified
• Manager toolkit (email copy, talking points, completion tracking)
• Clear escalation path for issues in the first 30 days

"Managers are the single biggest variable in whether training sticks. Build for them, too."

Heuristic

Phase 05

Evaluation

Prove the work landed by measuring against the metrics agreed in Discovery — not what's easy to count.

What happens

→ Level 1 — reaction surveys, kept short and pointed
→ Level 2 — learning, via xAPI statements from interactions, scenario outcomes, and assessments
→ Level 3 — behavior change, measured in the operational system (help desk tickets, error rates, cycle times)
→ Level 4 — results, tied back to the sponsor's business metric

What comes out

• xAPI statement schema and LRS pipeline (SCORM Cloud or client LRS)
• Dashboards for L&D + business partners
• 30/60/90-day review of behavior-change data
• Lessons-learned doc to feed the next project

"Kirkpatrick 4 is where the story lives. If you can't connect the training to a business outcome, the next project will be a harder sell."

Heuristic

See the framework in action

Each case study shows how these phases played out on a specific project — the decisions, the detours, the numbers.

Browse case studies