■ Services

We research, design, and build AI products that ship.

Three engagement shapes, run through the same arc, every time.

Engagements·················································3 engagement shapes

Duration·················································2 weeks to 6 months

Entry point·················································UX Research

Capacity·················································Studio-led delivery

Booking·················································Accepting inquiries

SECTION.01

Pick by goal

Match your intent to an engagement.

Want behavioral evidence before greenlighting a buildUX Research2 to 3 weeks Want a shippable AI product, evals from day oneAI Product Build6 to 10 weeks Want embedded research and build across a roadmapRetained Partnership6 months

SECTION.02

Engagements

UX ResearchResearch leadMinimum engagement

Moderated behavioral sessions and a decision-ready report, in two to three weeks.

██░░░░░░░░░░░░░░░░░░░░░░░░2 to 3 weeks

■5 deliverables■4 phases

Duration

2 to 3 weeks

View details

AI Product BuildBuild

A shippable AI product, with evals and monitoring wired in before launch.

██████░░░░░░░░░░░░░░░░░░░░6 to 10 weeks

■4 deliverables■4 phases

Duration

6 to 10 weeks

View details

Retained PartnershipEmbedded

Six months of embedded studio capacity across research, design, and build.

████████████████████████░░6 months

■5 deliverables■3 phases

Duration

6 months

View details

SECTION.03

Phases

Every engagement runs through the same arc: discover with real users, shape a decisive artifact, ship the system that implements it. When the engagement includes build, the studio that ran the research ships the code, so findings end up in production rather than a deck.

01

Twelve sessions across four segments. Rapid synthesis.

Real users in their real environments, not staged tests. We moderate, transcribe, and synthesize as the sessions run so the team gets findings in days, not three weeks.

01Twelve moderated sessions, four segments

02In-context (real boards, real data, real constraints)

03Rapid synthesis: patterns within the week, not after

Session logsynthesis live

Analyst

Support

Researcher

S.01Analyst"I do not trust the summary without citations"DONE

S.04PMAbandons if first answer feels wrongDONE

S.06SupportVerifies every claim against ticket historyDONE

S.07Support"If confident and wrong, I look bad"DONE

S.08ResearcherCorpus provenance > answer fluencyDONE

12 / 12 sessions complete72hr synthesis window

3 of 6 phases shown · Discovery, Evals & Handoff live on the methodology pageSee the full methodology

SECTION.04

What you can expect

The operating terms of an engagement.

Cadence·················································Weekly written readouts, async Loom updates

Attendance·················································Studio practitioners research, design, and ship

Scope·················································Locked at proposal

Access·················································Shared repo, eval dashboard, session log

Ownership·················································You own code, data, and monitoring on exit

Support·················································30-day post-engagement Slack window

Stop conditions·················································Studio recommends stopping if research does not support shipping

SECTION.05

Evals and gates

Production gate values the studio runs to before cutover. Same set on every AI build engagement. Final values are set in the shape doc against your workload.

Retrieval hit@kCorrect source in top-k for the research-corpus query setk=5 ≥ 0.92 before cutover

Answer faithfulnessClaim-level support check against retrieved sources≥ 0.95, zero tolerance on fabricated citations

Refusal calibrationSystem declines when confidence is below thresholdFalse-accept on known-unanswerable set < 5%

Latency p95End-to-end response under measured production loadSet in shape doc, usually < 3s interactive / < 12s async

Drift signalWeekly regression on frozen eval slice vs last stable runAny metric slip > 2% opens an incident

Edge-case coverageHand-curated adversarial set from session transcriptsGrows monotonically; no removals without sign-off

SECTION.06

Always included, never included

Every engagement ships with the same commitments. A few things fall outside our scope by design.

■ Always included

The studio runs the work end to end. No offshore layer, no relay
Scope and deliverables written into the SOW
Weekly written update, not a status meeting
All research data, raw footage, and code owned by you
NDA and data handling by default, SOC 2 / GDPR / EU AI Act aligned

■ Never included

Hourly billing or time and materials
Pitching work we have not actually done
Model training from scratch. We integrate, we do not pretrain
Ongoing feature factory once an engagement ends. That is a retainer

Research firms hand you a deck. Dev agencies hand you a codebase. The studio does both, in one engagement.

About the studio

Start an engagement

30-min discovery call. Proposal within the week.

Book a call

Loading services...

::::::::::: ::: ::: :::::::::: :::::::::: ::::::::::: :+: :+: :+: :+: :+: :+: +:+ +:+ +:+ +:+ +:+ +:+ +#+ +#++:++#++ +#++:++# :#::+::# +#+ +#+ +#+ +#+ +#+ +#+ +#+ #+# #+# #+# #+# #+# #+# ### ### ### ########## ### ###