AI transformation · Agentic AI · Owned and provable

A key, not a subscription.

The world is about to rent its intelligence: metered, hosted elsewhere, on someone else's terms, and closed to inspection by the person relying on it. We make the kind you can hold.

Let intelligence grow with trust, and all of us rise with it.

Find your door See it proved
  • Narrow
  • Cost-effective
  • Owned
  • Provable

What do you need to hold?

Three doors. Pick the one you came for.

ONE

For you

You have a decision that cannot leave the building, cannot be wrong quietly, or cannot wait for a cloud. We decide, build, prove and run it on your hardware, and you hold the result.

Walk through door one
TWO

With you

You have an idea and want a company built on it. The Studio takes it from ideation to the world, with you.

Walk through door two
THREE

By you

You want the craft itself. Workshops, training and open material in post-training, verification and on-premise deployment.

Walk through door three

Dependence, then partnership, then independence. Each door leaves you holding more than the last.

What we mean by AI transformation

Three things you are left holding.

Models you own.

Deep learning for the specific problem, or a small language model made expert at one job. Built for you, deployed on your hardware, and yours to keep, retrain and inspect.

Agents you can hold.

Agents that act inside your walls: orchestrated, given tools, and put to work on your systems. Every action they take is graded by a verifier that does not care whose agent it is, because an agent that is wrong takes a bad action, not just a bad answer.

Proof you can show.

A record of what the model or the agent did, what it got wrong, and what happened then. Evidence an auditor accepts, a board can read, and a regulator has already started to ask for.

Transformation is what everyone sells. What you hold at the end is the difference.

What the doors have produced

The Studio has already produced something.

The Key 4 Learning

The first company to come through the With you door, and the home of the By you door. It exists, it has a product, and the product is built with the same craft this site describes.

From The Key 4 Learning

Clarify

  • Builds a profile of how each learner thinks, rather than sorting them into a type.
  • Teaches by questioning rather than answering.
  • Lets a student learn by teaching a virtual pupil who has misconceptions of its own.
  • Finds the gap between what a learner believes they know and what they can do.
  • Gives teachers a view of the whole class in which no student is invisible.

Its tutor runs on a small model made expert in school mathematics and science, trained against a verifier: the same Prove described below, applied to a classroom.

First market: school curricula and entrance examinations in India. Built to travel.

Product page The Key 4 Learning Brochure v0.1 (PDF)

What we believe

Let intelligence grow with trust, and all of us rise with it.

We make thinking machines light enough to live on what people already own, and honest enough that their answers can be tested by the person relying on them. We build them so that whoever holds one can do more, own more, and answer for more than they could before. Everything else we leave to the people it belongs to.

  • TrustWith it, never ahead of it.
  • TrustHonest enough to test.
  • The riseThe last hand first.
  • The riseNo one smaller, everyone larger.
  • The riseTheirs to command.
  • UsSmall, and leave the rest.
Company

The craft

Decide. Build. Prove. Run.

1Decide

Most problems are not language-model problems. Choosing correctly is the capability; building afterward is the easier part.

2Build

Real deep learning for the specific problem in front of you: vision, language, structured data. Or a small model made expert at one job, trained against a verifier.

3Prove

The machine that decides what is correct. Pointed at our own model it is training. Pointed at anyone else's it is assurance. One role per model.

4Run

Deployed on hardware you already have, then operated, not handed over and forgotten. Agents included: orchestrated, given tools, and graded on what they did.

Decide · Build · Prove · RunThe craft in depth

Prove, made visitable

A verifier, running in your browser.

A grader that decides what is correct has to exist before a model can be trained against it. Once it exists, it can grade anyone's model.

The Bench

The items below are records of the kind institutions rely on: a payment, a maintenance log, a shipment, a lab result, a meter reading. The answers are hand-written illustrations, not model outputs, and the grader is genuinely executing on this page. View source.

  1. The value must match the ground truth to two decimal places.
  2. Date must be ISO format, yyyy-mm-dd.
  3. The name field is compared after trimming whitespace and case-folding.
  4. Reference must match exactly.

Lenient also accepts dd/mm/yyyy dates and ignores currency symbols on amounts.

Answer set A-
Answer set B-
Answer set C-

1. Payment: paid 4,250.00 to Meridian Tools on 3 March for invoice 88-A.

Answer set A

Amount
4250.00
Payee
Meridian Tools
Date
2026-03-03
Reference
88-A

Answer set B

Amount
4250.00
Payee
Meridian Tools
Date
03/03/2026
Reference
88-A

Answer set C

Amount
4250.00
Payee
Meridian Tools
Date
2026-03-03
Reference
88-A

Method notes and the full bench

Your numbers, not ours

Rent, or hold.

Yearly rental cost at these numbers:$43,800

Against that, a model built once for the job and run on hardware you already own. We do not put our number here; the point is that you can put yours.

Not included: engineering, evaluation, hardware, operation. Those are the conversation.

The Ledger

Every claim we make, with its status.

Dated, append-only, losses included. New claims enter here before they appear anywhere else.

  1. Held
    Every regulator we quote is quoted from its primary text, with its status stated.

    Thirteen entries; two are drafts, one a consultation, one a proposed issue, two publisher's abstracts.

  2. Claimed
    AI transformation and agentic AI are what we do; owned and provable is how.

    Stated on the front page. Every agent we build or assure is graded by a verifier.

  3. Claimed
    The Key 4 Learning is the first Studio company. Its product is Clarify.

    Product page published; brochure v0.1.

Full ledger

Different regulators, one sentence

Federal Reserve and OCC · SR 11-7 / OCC 2011-12, model risk management guidance

"Also, organizations should maintain an inventory of models implemented for use, under development for implementation, or recently retired."

FDA, Health Canada, MHRA · Good Machine Learning Practice, guiding principles (GMLP)

"Deployed models have the capability to be monitored in 'real world' use with a focus on maintained or improved safety and performance."

Central Bank of the UAE · Model Management Standards, Notice 5052/2022

"Institutions must maintain a comprehensive inventory of all their models employed in production to support decision-making."

NIST · AI Risk Management Framework 1.0

"Mechanisms are in place to inventory AI systems and are resourced according to organizational risk priorities."

Bank of England, PRA · SS1/23, model risk management principles

"A comprehensive model inventory should be maintained to enable firms to: identify the sources of model risk; provide the management information needed for reporting model risk; and help to identify model inter-dependencies."

NAIC · Model Bulletin, use of AI systems by insurers

"Regardless of the existence or scope of a written AIS Program, in the context of an investigation or market conduct action, an Insurer can expect to be asked about its development, deployment, and use of AI Systems."

ISO/IEC · 42001:2023, AI management systems

"ISO/IEC 42001 is an international standard that specifies requirements for establishing, implementing, maintaining, and continually improving an Artificial Intelligence Management System (AIMS) within organizations."

European Union · Regulation (EU) 2024/1689, the AI Act

"High-risk AI systems shall be designed and developed in such a way as to ensure that their operation is sufficiently transparent to enable deployers to interpret a system's output and use it appropriately."

ISO/PAS · 8800:2024, AI safety in road vehicles

"This document describes safety-related properties of AI systems that can be used to construct a convincing safety assurance claim for the absence of unreasonable risk."

Monetary Authority of Singapore · Consultation paper, AI risk management guidelines

"MAS proposes that FIs establish and maintain an accurate and up-to-date inventory of AI use cases, systems or models to support governance and oversight, as well as risk management throughout the AI lifecycle."

Reserve Bank of India · Draft guidance, model risk management principles

"It should ensure that no model is used, relied upon, or deployed unless it is part of inventory."

EASA · Concept paper, AI applications guidance

"Explainability is a key property that any safety-related AI-based system should possess."

Central Electricity Authority · Cyber Security in Power Sector Regulations, 2026

"maintain asset register- (a) for all cyber assets along with the requisite details including ownership, hardware, firmware, software, and patch as per the procedure defined in Cyber Security Policy".

Thirteen documents, five continents, one sentence. Sourced from the primary text; statuses stated as they are.

Regulators

One address

One sentence about what you need to hold.

info@thekey4.ai