Research OS for Cowork

The things that make AI feel like more work

01“I hand Claude the whole blob of my process and then get annoyed when it cannot do it.”

02“I spent three hours with Claude on something I could have done in thirty minutes.”

03“I asked whether users would want the feature and it said absolutely.”

04“Every how-to is written for developers. None of it is written for a research process.”

05“I know I should write these context files. They are eleven pages long and I am never going to do it.”

06“Leadership keeps asking what AI is doing for us and I have nothing to point at.”

Everything you need to

Run your whole practice from one folder

#1 Structure

Your research folder

This folder holds your entire research process, enabling you to share the fundamental context with Claude so that you are not constantly reexplaining yourself

Context filesRules and routingQuarantine folder

#2 The interview

You get to be interviewed

The onboarding agent asks you all the questions you need in order to dig into what really matters, like what you are measured by, your team’s needs, your pain points, so that every output is customized

In-depth interviewForks by roleCustomized files

Try it: click through a real exchange

#3 Skills and plugins

Your standards

Setting up skills and plugins enables Claude access to your brain, your knowledge, and your standards. With these in place, your research becomes not only more efficient, but allows you to run more impactful studies

5 or 10 bespoke skills2 or 3 pluginsSkill and plugin builder

Your Research OS pluginYour bespoke skills, routing and scheduled tasks. Installs anywhere you work.

"Working with me"For your PM and designer. How to write a request, what a good brief looks like, how to read your reports.

OnboardingTeam tier. A new joiner runs it and is working by lunchtime, without anyone's calendar.

#4 The catch-up

Keep on iterating

It reads the folders of finished studies you already have, writes a card for each, builds the index and drafts your first impact story. Then, scheduled tasks allow you to continuously record impact while freeing up admin time

Continuously measure impactSave time on adminSchedule tasks

~/Research/2024-2026 · 14 folders

Waiting. Press run.

What actually changes

The task

How it goes now

How it goes after

Getting going

Setting any of this up

Eleven pages of context files you keep meaning to write and never will

An onboarding agent interviews you and writes them, in about two hours, and you can dictate the lot

Starting a Claude session

Ten minutes typing out who you are, what the product is, what the team is focused on

Claude has all the context it needs, or quickly interviews you for any gaps

Deciding what to hand over

You give it the whole blob of your process and get something generic back

Your process split into what repeats and what needs your judgement, and only the first kind gets handed over

Running the work

Writing a screener

Open the last one, strip it, rewrite to new criteria

Criteria in, first draft out in your format

Sense-checking an idea

You ask whether users would want the feature and it tells you absolutely

Your evidence thresholds sit in the rules, so thin evidence gets flagged rather than agreed with

Synthesising eight transcripts

A few days trying to get it done for the team that wanted it “yesterday”

Tagged against your taxonomy, held to your evidence threshold, with verification built in

Drafting a readout

Sitting and staring at a blank page for a few hours

Drafted in your voice against your template, then run through the so-what check

Everything around the work

A PM asks for research

A vague Slack message, and one of you for every ten of them

They run your intake plugin and arrive with a real question, and the ones who can run a safe study get guardrails to do it

“Have we researched this before?”

Taking hours to search through folders for old research

Claude connects to your tools so it can pull in any previous research with a command

The quarterly impact report

Days spent pulling together impact and dragging answers from stakeholders

Continuously measured with scheduled tasks and plugins, so the answer exists before anyone asks

“How do you use AI responsibly?”

Uncertainty of how to track and respond

A written policy in your words, plus a log of what AI touched on every study

What is this worth to you

Sessions with Claude per week

Minutes re-explaining context, each time

Studies you finish per year

Hours synthesising one study

Days on the quarterly impact report

Your hourly cost ($)

People this covers

Synthesis time saved (my assumption)

Hours back, per year

200

across 1 person

Re-explaining context74 hrs

Synthesis42 hrs

Impact reporting84 hrs

Worth roughly

$11,200

Against the individual build, payback around week 8

Build it yourself, or have me build it

Do it yourself

The Template

The entire templated operating system delivered to you with an onboarding agent that enables you to set it up.

$249 once

Skill builderPlugin builder

✓ The folder, ready to connect, with a demo study included
✓ Seven research skills
✓ One plugin plus an MCP template to connect your tools
✓ Scheduled task template
✓ Walkthrough recording

Secure checkout by Stripe | instant download | free updates

Done for you - Individual

The Individual Build

I interview you myself, find out your biggest pain points, and build everything customized for you.

$1,750 once

5 bespoke skills2 plugins

✓ Everything in the template
✓ A 60-minute depth interview
✓ Context built from your real reports
✓ Your OS plugin, plus a "working with me" plugin for your PM
✓ Handover session plus a day-30 check-in

Apply for an individual build

best value Done for you - Team

The Team Build

I set up an entire system for teams from two to six researchers.

from $7,000

10 bespoke skills3 plugins

✓ Everything in the individual build
✓ Sessions with the lead and every researcher
✓ Agreed taxonomy and evidence thresholds
✓ Team plugin, partner plugin, and an onboarding plugin
✓ A team AI policy draft
✓ One quarterly review included

Book a call

Enterprise

A whole research function

For two or more teams who need a larger, scalable system put into place.

Book a call

Be one of the pros

What changes once it is running

"The catch-up ran on two years of studies while I made coffee. I had an impact doc before my first meeting, and I have been meaning to write that since January."

Rosa M.Senior UXR, fintech

"My head of product asked how we govern AI use in research. I sent her the policy file and she forwarded it to legal as the example."

Deniz K.Research lead, healthtech

"Three of our PMs installed the partner plugin. The requests we get now arrive with an actual decision attached."

James W.Research lead, 4-person team

Questions I get

What is the difference between a skill and a plugin?

A skill is one job done your way, like drafting a screener from criteria or writing a study card. A plugin is a bundle of skills with tools and subagents baked in. A skill makes you faster, whereas a plugin streamlines your entire workflow and helps make democratization more rigorous.

Should I buy the template or book a build?

Buy the template if you enjoy setting things up yourself and feel comfortable with DIY in Claude. Book a build if you want things personally built for you and your team.

We are four researchers. Do we need four builds?

No, and four individual builds would give you four incompatible systems. The team build gives you one, with agreed standards, comparable study cards, and an onboarding plugin so the next hire sets themselves up.

Is this just a folder of templates?

Nope. This is a full system, templates included, that gets you set up on Cowork fast so that you can focus on doing research rather than wrangling AI.

What happens to participant data?

There is a quarantined folder Claude is told never to read, a pseudonymisation step before anything reaches the working structure, and a short rules file setting out what may never leave. On team and enterprise builds that becomes a policy document.

Claude changes constantly. Will this go stale?

Everything that changes lives in one file, so an update is one file rather than a hunt through eight. The longer answer is the skill builder, since the point is that you keep making your own.

Incentives, 8 participants	$760
Recruitment or panel time	$1,500
Researcher time, 10 days	$5,000
One duplicate study	$7,260

Claude that already knows your product, your people and your standards