AskNiels

Turning a strategic design methodology into an operational AI-powered experience platform.

Role

Co-founder · Product · Design · Dev

Challenge

Methodology trapped in documents.

Scope

Knowledge · AI · UX · Workflows

Impact

57 activities. −54% delivery. 1 system.

01 · Situation

A strong methodology is useless if teams cannot activate it.

Niels had a rich methodology. Its value depended on people being able to use it without an expert in the room. The challenge was not documentation. It was productization.

✕ Trapped knowledge

57 activities, multiple phases and use cases, valuable but inaccessible without expert guidance

✕ Consistency gap

Method interpretation varied across coaches, clients and internal teams

✕ AI risk

Generic AI would dilute the doctrine and produce noise instead of methodological value

✕ Activation gap

Knowing the method existed and being able to use it were two entirely different things

The core decision

02 · Approach

Three tracks. One operating system.

Structure the corpus before building the interface. AI behavior model before UX. Platform logic around usage scenarios, not features.

Knowledge architecture

Tension

The method existed. It was not queryable.

57 activities lived in documents and human memory. Before AI could be useful, the corpus had to be restructured and made queryable.

Call

Structure the corpus before building the interface.

Every activity deconstructed into structured knowledge units. Chunking, embedding and retrieval designed to surface the right activity for the right context.

Result

57 activities. One queryable, doctrine-faithful corpus.

57 activities. One doctrine-faithful corpus. Guardrails embedded at corpus level, no generic AI noise.

AI experience design

Tension

AI without doctrine produces noise.

Default AI behavior is to be helpful in the broadest sense. For AskNiels, that was a failure mode.

Call

Define the behavior model before the interface.

Explicit behavior model: context before output, operational answers over inspiration, method-faithful over generic best practice.

Result

An assistant that activates the method. Not one that replaces it.

An assistant that activates the method. Answers that don't help a user decide, prepare or produce were treated as bugs.

Platform & workflows

Tension

Different users. Different maturity levels. One system.

Coaches, clients and learners had fundamentally different needs. One interface had to serve all three without degrading for any.

Call

Design for usage scenarios, not for features.

Five core usage scenarios designed explicitly: understand, choose, prepare, adapt, produce. Platform built around scenarios, not features.

Result

Reusable workflows across coaching, learning and consulting.

Reusable workflows across coaching, learning and consulting. Method accessible without an expert in the room.

Take away

03 · Outcomes

What became operational.

No artificial performance metrics. The proof is system performance: structured corpus, governed AI behavior and measurable delivery acceleration.

Before→After

Methodology trapped in documents

→

57 activities in a queryable corpus

Generic AI answers

→

Doctrine-faithful guardrails

Expert-dependent knowledge

→

Self-serve AI-assisted workflows

Delivery bottlenecks

→

−54% on operational tasks

delivery time

-54%

on selected operational tasks using AI-assisted workflows

activities

structured into a queryable, doctrine-faithful RAG corpus

doctrinal

Guardrails

AI behavior governed to stay faithful to the Niels methodology

usage contexts

coaching, learning and consulting served by one platform

04 · Takeaways

Three things this confirmed.

AI amplifies what is structured. The quality of output is directly proportional to the quality of knowledge architecture upstream.

Behavior design is product design. Defining how an AI thinks and prioritises is a design task, not a technical one.

Productizing expertise requires moving from knowledge transfer to system design. The method didn't change. Access did.

Closing

From method to product. From expertise to system.

Not a chatbot. A productization effort: expertise, methodology and operational knowledge turned into a system that scales.

← Previous case

Longchamp

A brand book with no system logic. 13 weeks later: 80+ components, 24 markets, zero localisation debt.

Next case →

Boucheron

Immersive phygital exhibition for Boucheron's 160th anniversary. 20k visitors over 16 days.