AI governance for organisations

Capable AI. Governed authority.

ISAC is a governance layer for AI use in organisations. It is designed so that every AI request is checked against the organisation’s policy and gets an allow, refuse or escalate decision before any model runs, with each decision recorded in an audit trail. Models are interchangeable; permissions, memory and policy stay with the organisation.

For regulated UK SMEs · In development (technical preview) · Built on GCF, our Governed Cognitive Framework

Our technology

ISAC is designed to check every AI request. GCF is the framework underneath.

GCF Frameworks Ltd is building ISAC (Intelligent Systems Architecture Core), a governance layer for AI use in organisations. It is built on GCF (Governed Cognitive Framework), our underlying system architecture. ISAC is designed to check each AI request against the organisation’s policy before any model runs, decide whether to allow, refuse or escalate it, and record the decision. ISAC is in development and not yet available.

Learn more about our technology

Our first product

In development (technical preview) · not yet available

ISAC Intelligent Systems Architecture Core

A governance layer for AI use in organisations, built on GCF.

What ISAC is designed to do

Policy check first
Check each AI request against the organisation’s policy before any model runs.
Allow, refuse or escalate
Give every request a clear decision, and bring in a person when the policy says so.
Audit trail
Record each decision, so people can see who allowed what, and when.

Models are interchangeable: permissions, memory and policy stay with the organisation.

The system

GCF 2.0 baseline frozen

GCF Governed Cognitive Framework

Capability does not equal authority. GCF is designed to manage how intelligence operates, and to set and enforce its boundaries outside the model itself.

What the architecture is designed to do

Authority
Decide which actions are permitted, blocked or need a person’s approval.
Written record
Keep an audit record of actions, for people to review.
People in control
Keep meaningful human control: people can review, step in and override.

Some governance components exist today. End-to-end integration is still being developed.

01 · Why governed AI

Capability isn’t authority.

A model can be remarkably capable. That doesn’t make it the one who decides what it’s allowed to do.

Think of a new starter

Someone joins your team. They’re sharp, well read and ready to help on day one.

You still don’t hand them the company bank card and every password on their first morning. That isn’t doubt. It’s how trust works at work: access is given for a job, someone signs it off, and there’s a record of what was done.

We think AI deserves the same common sense.

GCF is designed so that every action answers three questions: is it permitted, who authorised it, and what actually happened?

02 · How it works

Six rules the system is designed to keep. Whatever the model.

GCF isn’t a model or a chatbot. It’s the architecture the whole system is built on, with models as replaceable parts. It is designed so that requests are checked on the way in, answers on the way out, and people can step in at any point.

  1. 01

    Authority stays with the system

    Models are not meant to grant themselves access or credentials. Authority is designed to be defined and held by the system.

  2. 02

    Tools through controlled routes

    Actions are designed to be checked against what has been permitted before they happen. Full governed tool execution is not yet confirmed as implemented.

  3. 03

    Models are replaceable parts

    Routing between models is designed so that an organisation isn’t tied to one provider.

  4. 04

    Memory belongs to the system

    Persistent memory is designed to sit with the system, not inside one model, so it can survive a change of model.

  5. 05

    Budgets and a written record

    Use is designed to be bounded by budgets, with actions recorded for people to review.

  6. 06

    People can always step in

    The system is designed so that people keep the ability to review, intervene and override.

Diagram of how ISAC is designed to work: an AI request from a person or an app goes to a policy check against the organisation’s policy before any model runs. The check gives one of three decisions: allow (the request goes to a model), refuse (the request is stopped) or escalate (a person decides). Every decision is kept in an audit record. Models are interchangeable; permissions, memory and policy stay with the organisation. Design intent; ISAC is in development.
How ISAC is designed to handle an AI request. Design intent: ISAC is in development (technical preview) and not yet available.

03 · Architecture

The architecture, and the system built on it.

People often ask which is which. GCF is the foundation. ISAC is the system we’re building on it.

The architecture01

GCF

Governed Cognitive Framework

Defines how cognition, memory, tools, resources, authority, budgets and the written record work together, whichever models are in use.

GCF 2.0 baseline frozen

The system02

ISAC

Intelligent Systems Architecture Core

A governance layer for AI use in organisations, being built on GCF. ISAC is also designed to work as a persistent AI system with memory across interactions, rather than a conventional chatbot.

In development (technical preview) · not yet available

The model is replaceable. The intelligence system is not.

04 · Who it’s for

For teams putting AI to real work. Within clear lines.

  • 01

    Regulated UK SMEs

    Small and medium-sized firms in regulated sectors that use AI and need to show who allowed each AI action, and why.

  • 02

    AI and software teams

    Teams that want to change or upgrade models without rebuilding permissions, memory and tooling each time.

  • 03

    Security-sensitive organisations

    Organisations where it matters who authorised an action, and where a clear record of what happened is expected.

  • —

    Robotics and physical intelligence Longer-term research

    We’re researching how the same principle could apply to physical systems. A research direction only: not a product, and not available.

05 · Progress

Where we are, honestly.

A plain record of the work so far. We’ll add dates only once they’re confirmed.

  1. GCF 2.0 baselineFrozen
  2. ISAC foundation workComplete
  3. Foundation StableReachedAn internal milestone, not an external certification.
  4. Support for external and custom AI providersIn progress
  5. ISAC next technical milestoneIn developmentNo date yet.
  6. Robotics and physical intelligenceResearchLonger-term research only.
Billy Payne, founder of GCF Frameworks

06 · Company

Why we’re building this.

AI systems are getting more capable, and more of them are being given real access: to inboxes, to files, to the tools a business runs on. Too often, the question of who authorised what is left to the model itself.

That felt backwards to us. In a well-run organisation, authority is given on purpose, written down, and can be taken back. We think AI should work the same way, so we’re building the governance first, one step at a time.

Billy Payne is the founder and director of GCF Frameworks Ltd. He has built GCF and ISAC on his own, and is based in Blackpool. Billy on LinkedIn (opens in a new tab)

Independent UK companyFounded in Blackpool in 2026 by Billy PayneCompany no. 17468924

More about the company

07 · Get in touch

Build it with us.

Register interest, or work with us. Tell us what you’re working on and what you’d want AI to be allowed to do.

  • Private-sector organisations first
  • No commitment on either side
  • A reply from the people building it

Email us

There’s no form on this site. Email opens in your own mail app, so you choose what to send.

hello@gcf-framework.com

LinkedIn

Email to register interest

We use what you send only to reply. See our privacy notice.

GCF Frameworks

GCF Frameworks Ltd is a member of the NVIDIA Inception program.

NVIDIA Inception program member badge