RESEARCH DESIGN
THREE-TRACK STRUCTURE

No single track could describe a field this size

Once AI was reframed as an area of knowledge rather than a product category, one benchmark was no longer enough. Three parallel tracks each answered a different question, and the insight lived in the intersections, not in any single track.

The deliverable format was decided by the tracks, not before them. A field moving weekly cannot be described by a document that stops moving.
Input
The brief
"Understand what is being developed with AI." No methodology. No deliverable format. Two months and a blank page.
Track 01
Technology landscape mapping
Structured glossary of AI subfields, scored by how often each appears in recent literature and product announcements.
Output · 12 knowledge domains with relative activity
Track 02
Competitive intelligence
Published papers, institutional sites, LinkedIn activity, client materials, hands-on trials of competitor products.
Output · competitors plotted onto the same domains
Track 03
Barriers and infrastructure
Hardware gaps, data quality, regulatory uncertainty (EU AI Act, no Brazilian framework), talent scarcity, IP ambiguity.
Output · constraints read as R&D direction
Synthesis
Knowledge domain map
Every finding filed by technology area instead of by company. Activity, presence and constraint became three readings of one structure.
Deliverable
The Technology Thermometer
A live Airtable dashboard with three interconnected views, queried together. Zero adoption friction: the team already used the tool.
View 01 activity · View 02 competition · View 03 barriers
Before: six weeks of research, depreciating from the day it is printed After: an instrument 45 people query and keep current