Prex Studio

We build AI agents that reach production.

Most demo well. Few survive contact. The work is everything after the demo.

0People trained
0Systems shipped
0Sessions self-run

The problem

A demo has to be impressive.
A system has to be trusted.

Fewer than one in ten enterprises has scaled an agent in any single function. The models are not the reason.

Anything acting on your behalf must refuse when refusing is right, leave a record a human can read six months later, and pass a review by people paid to say no. Most teams learn this after the build.

9 in 100 enterprises have scaled an agent inside a single business function.

Selected work

What we have built.

Agentic system

Enquiry triage with a refusal boundary

Classifies inbound, drafts what it can, and withholds a draft entirely where a human must decide.

Long-running agent

A system that rewrites its own rules

Past a thousand sessions, revising how it operates under one condition: never silently.

Internal

A studio that runs on its own agents

Publishing, research and operations handled by the same systems we build for clients.

Personal AI

One system instead of a dozen apps

Tasks, health, projects and people in a single life OS that sees how they connect.

Enablement

Faculty AI programme, three days

A university's teaching staff taken from curiosity to tools they now use weekly.

Enablement

Multi-week GenAI bootcamps

Cohorts of working professionals building agentic systems end to end, not watching demos.

What our workshop participants say

Rated 4.7 out of 5.

This workshop not only made me familiar with interesting AI tools but has also helped me in learning the actual workflow and systems to get the best out of them.

Priyanka NegiDepartment of Chemistry

I gained practical knowledge of Gemini and NotebookLM and I plan to use them to create better learning resources, streamline research and improve my teaching workflow.

Saksham AgarwalAssistant Professor, Business Management

It will surely save me a lot of time and do so much of my work that I was doing manually.

Yuganshi GuptaAssistant Professor, Business Management

Websites and applications that are very useful for personal and professional work.

M Pency LimaBusiness Administration

The training was very informative and practical. I learned useful AI tools for teaching, research, writing and documentation.

Kavindra Singh DhamiAssistant Professor, Civil Engineering

This workshop not only made me familiar with interesting AI tools but has also helped me in learning the actual workflow and systems to get the best out of them.

Priyanka NegiDepartment of Chemistry

I gained practical knowledge of Gemini and NotebookLM and I plan to use them to create better learning resources, streamline research and improve my teaching workflow.

Saksham AgarwalAssistant Professor, Business Management

It will surely save me a lot of time and do so much of my work that I was doing manually.

Yuganshi GuptaAssistant Professor, Business Management

Websites and applications that are very useful for personal and professional work.

M Pency LimaBusiness Administration

The training was very informative and practical. I learned useful AI tools for teaching, research, writing and documentation.

Kavindra Singh DhamiAssistant Professor, Civil Engineering

All the live and hands-on sessions are excellent.

Dr. Tarun Kumar DasElectronics and Communications

The tools I have learned in this workshop will be very useful in my work. Anurag made it really easy to introduce these tools.

Divya VenugopalAssociate Professor, Microbiology

It was an excellent experience. He made every tough looking task so easy for us.

Dr. Anjali DimriAssistant Professor, GESoM

It has helped me prepare my teaching tools more effectively. My productivity is certainly going to be increased many folds.

Abhishek MisraAssistant Professor, GESoM

Effective and simple way to understand AI for professors.

Rajesh TiwariProfessor, GESoM

All the live and hands-on sessions are excellent.

Dr. Tarun Kumar DasElectronics and Communications

The tools I have learned in this workshop will be very useful in my work. Anurag made it really easy to introduce these tools.

Divya VenugopalAssociate Professor, Microbiology

It was an excellent experience. He made every tough looking task so easy for us.

Dr. Anjali DimriAssistant Professor, GESoM

It has helped me prepare my teaching tools more effectively. My productivity is certainly going to be increased many folds.

Abhishek MisraAssistant Professor, GESoM

Effective and simple way to understand AI for professors.

Rajesh TiwariProfessor, GESoM

What you can commission

Three ways in.

01

Review

What is blocking your pilot, and what shipping it would take.

02

Build

A system taken to production, with governance designed in rather than bolted on.

03

Enable

Your team trained on your own work, not a generic curriculum.

Questions

Before you write.

With a Review. A few days looking at what you have, ending in an honest read on what is blocking it and what shipping would take. Build or Enable follows only if it makes sense.

We do not quote before scoping. Every engagement starts with a scoping call, after which you get a written scope note and a fixed fee. Agentic work varies enormously by system count and data maturity, and a number given before that conversation would be fiction.

Automation agencies optimise for shipping a workflow quickly. We optimise for a system that survives a governance review: it refuses when refusing is right, logs every decision with its reasoning, and has a bounded blast radius. That is most of the work and it is the part that decides whether anything reaches production.

Either, and usually both. The default is building it with your team in the room, so the second system does not need us.

Yes. We build on what you already run rather than moving you onto something of ours. If a tool has to change we say why, and it is your call.

You do. The system, the specs, the decision logs and the reasoning behind them, all of it.

Who

Prex Studio was founded in 2021 by Anurag Kumar. IIT Bombay and IIM Bangalore alumnus, with a decade of AI product experience. He has trained over 5,000 professionals to build and use AI solutions, and built autonomous AI systems like Nautilus, which has run for over a thousand sessions and can rewrite its own operating rules.

Anurag Kumar, founder of Prex Studio

Anurag Kumar
Founder

Contact

If you are building with AI and it keeps stopping short of production, that is the conversation we want.

hello@prex.studio