Case Study

Collins Aerospace

How MAIK transformed early-stage design research at a global aviation company. From weeks of manual effort to 15 minutes of structured AI collaboration.

80,000
Collins employees worldwide
25
Industrial design engineers in the CXD team
2 weeks
Typical First Diamond window for a project
0
Shared AI frameworks before MAIK

The company

Collins Aerospace is one of three business units of the RTX Group, employing approximately 80,000 people across more than 300 locations globally. The Cabin Experience & Design (CXD) team designs passenger-facing cabin solutions: galley systems, lighting concepts, seating configurations, and full interior experience visions.

The team of approximately 25 industrial design engineers is split between Seattle and Ireland. Collaboration happens remotely. Every piece of design work must simultaneously satisfy airlines, OEMs, engineering teams, certification bodies, and internal business leadership. Each stakeholder speaks a different language and evaluates the work through a different lens.

"We spend as much time formatting the deck as we spend thinking about the design. And by the time the deck is done, there is no time left to question whether we are going in the right direction."
Collins CXD Designer

The problem: sensemaking under pressure

The early stage of every project (the "First Diamond") is where the most ambiguous, cognitively demanding, and time-pressured work occurs. Designers must gather research, synthesise findings, build arguments, and produce stakeholder-ready presentations. Often all of this in two weeks.

Interviews and workflow analysis revealed eight structural challenges that shape every project:

01

Limited exploration time

RFP cycles can require a full concept vision in as little as one week.

02

Scattered information

Research inputs across email, Teams, SharePoint, and personal files. Context rebuilt manually for every project.

03

Siloed stakeholder inputs

Engineering, regulatory, market, and client inputs arrive at different times in different formats.

04

High communication load

The same content re-presented in different registers for different audiences.

05

Hidden design knowledge

Tacit intuition and spatial reasoning are central to decisions but difficult to articulate to stakeholders.

06

Pressure toward early lock-in

Phase gate deadlines push designers to commit before they have adequately explored alternatives.

07

Lack of structured evidence

Early trade-offs are difficult to quantify in the Cost, Weight, Comfort hierarchy stakeholders require.

08

Weak transfer of design intent

When engineering industrialises a concept without oversight, the original design rationale erodes.

"I know there's probably a better direction. I just don't have the time to find it."
Collins CXD Designer

The real barrier: not tools, but structure

Designers at Collins already used AI tools like ChatGPT. The problem was not lack of access. Each designer used AI differently, producing results that could not be compared, synthesised, or built upon at team level. There was no shared vocabulary for describing what a good AI configuration looks like. No institutional memory of what had worked. No progression path from basic use to sophisticated workflows.

The absence of a framework was itself the problem.

The solution: MAIK deployed in three layers

MAIK was designed to match AI complexity to designer readiness. Each layer introduces one new concept and one new level of control.

Layer 1

MAIK Cards & Canvas

56 physical cards, each built around an established design method. Two-sided: the front presents a recognisable scenario, the back contains a pre-written prompt. The MAIK Canvas structures team sessions with eight configuration sections plus reflection fields. Individual prompting becomes collective learning.

Layer 2

Claude Skills

Canvas configurations become persistent instruction sets. Two Skills were built: a YouTube Researcher (qualitative video research agent) and a Collins PPTX Maker (branded presentation agent). Triggered by natural language, executed reliably without re-explaining the method.

Layer 3

Orchestrated Workflows

Skills connect in sequence with human review checkpoints. Research question enters. Structured research document comes out. Designer adds context. Branded presentation is generated. The designer judges. The AI gathers.

The result

3-5 hours manually.
15-25 minutes with MAIK.

The full research-to-presentation workflow. From a natural language research question to a structured Word dossier with timestamped evidence to a branded Collins slide deck. The majority of the 15-25 minutes is the human review step where the designer adds context that only they can provide.

What the client actually asked for

Five success criteria were agreed with the Collins client after the midterm review. They describe a practice, not a product:

1. Relevance over visual perfection. Accurate, decision-relevant data. Not polished slides containing hallucinated content.

2. Two interaction modes. A Data Bank for continuous background monitoring. A Specific Ask for on-demand queries.

3. Speed as the primary benchmark. Weeks of human effort vs. minutes of agent effort.

4. User ownership. The team must be able to build, adapt, and maintain agents without external support.

5. Structure reduces risk. Formalised AI use reduces organisational risk. Unstructured informal use is the higher-risk scenario.

"The MAIK Canvas teaches structure. The Skill makes that structure persistent and shareable. Together they transform AI from an individual productivity shortcut into a team-level, organisationally auditable capability."
Thesis synthesis

See MAIK in action

Explore the toolkit that was built for Collins Aerospace

Try the toolkit →