Questions and answers

TrustGenAI Limited is a UK company, registered in England (company number 17468508). We make Agentry, governance and assurance software for AI agents, and we help organisations adopt agents through four lines of work: strategy, delivery, assurance and capability.

The company was incorporated in September 2026; the experience behind it is its director's. Mohammad Ali Akeel has worked for over twenty-five years as a data and AI architect and consultant in local and central government, public bodies, infrastructure and regulated industries. He is completing doctoral research at the University of Portsmouth on how organisations evidence responsible use of AI, which is where the RAID-T assurance framework comes from. We will tell you plainly which of that experience is relevant to your work, and you can weigh it accordingly.

The director leads every engagement and is accountable for it. Where the work needs more hands or a specialism, named associates join under written contracts with TrustGenAI that carry the same confidentiality and data protection terms as your contract with us. You know who they are, and what each one does, before work starts. There are no hidden subcontractors.

Yes, and we tell you which ones before we start. Interviews and workshops are transcribed only with participants' consent, inside our own Microsoft 365 tenancy. Analysis and drafting support run on an open-weight language model on our own server in the UK, so nothing is sent to an external AI provider. No personal data goes into public AI services, and a person checks every output. Personal data is handled under the UK GDPR and the Data Protection Act 2018, kept to the minimum, and returned or deleted at the end as you direct.

You own the new material we create for you. Our own methods, frameworks and software stay ours; where they are part of what we deliver, you can use them as delivered. We list anything pre-existing in our proposal, so there are no surprises later.

Yes. We set our cover to what each contract requires, including professional indemnity, public liability and employer's liability, before the work starts.

We are based in London and work across the UK. Most of the work runs online; we come on site where it helps, by agreement.

Defined pieces of work at a fixed price, and open-ended support by the day. We quote after a short conversation about what you need, and the quote states what is included and what is not.

Software that gives every AI agent an owner, limits it cannot exceed, and a record of everything it did. It registers agents, classifies them by the data they touch, governs their lifecycle, runs them within limits, samples their work for human review, and reports on all of it. What it does walks through each part.

Components run today; the integrated platform is in development. The console marks which parts are demonstrations rather than production software, and in a demonstration we show you which is which.

It is not tied to one cloud or one model provider. It works with services that offer the common chat-completions interface, including open-weight models you host yourself, so an agent that has to stay inside your own network can run against a local model.

That is in development. Read-only connectors look in your identity directory and cloud accounts for agent identities and what each one can reach, and list them for an owner to claim. They run today against test accounts, not yet against a customer's estate.

From what the agent can reach, how sensitive that data is, what it can do, and whether its actions can be undone, including whether it can send data outside your organisation. Each control you apply is scored for how much it reduces that risk, so you can see the risk before and after, and what remains. A first version of the scoring model runs today and is still being calibrated.

Yes. Each agent can be given an allow-list of the destinations and methods it may use, and anything outside it is refused before it leaves, with the decision recorded.

It runs in a region you choose, and in your own tenancy where that is required. You decide whether the record of an agent's work keeps the full text, a redacted version, or fingerprints only. Security and data has the detail.

No software can do that, and we do not certify anything. Agentry produces the records those frameworks ask an organisation to keep. Frameworks sets out what it produces against each, and nothing on this site is legal advice.

Book a thirty-minute demonstration on screen. We show what runs today, what is still being built, and what we cannot do yet.