Home·Platform

One platform to put AI into production, and keep control of it.

Ajutant is a packaged AI runtime your own IT team can deploy and run, inside your own Azure tenant, without a specialist AI engineering team. Ready-built assistants, governance from day one, and your data never leaving your control. One governed platform for every use of AI in your organisation: the assistants people use, the AI inside your applications, and any third-party tool that needs a model.

· AI you stay in control of.

Ship AI use cases fast. Without breaking what depends on them.

The platform exists so a small build team inside your organisation can put new AI use cases into production at the pace the business needs, with the discipline that protects every commitment your IT team has already made: uptime, accuracy, audit, the lot.

Speed and safety aren't opposites here. The reason you can ship a new assistant in days is that the platform enforces the discipline around it. Test, version, draft, publish, roll back. The same loop that makes shipping fast is the loop that makes it reversible.

01 / Author

Prompt development interface

A proper authoring environment, not a settings page. Write the assistant's instructions, attach the grounding sources, configure tool calls and guardrails, and see what changes as you go.

02 / Test

Built-in test bed

Run candidate prompts against representative inputs before anyone outside the build team sees them. Compare responses side-by-side. Catch regressions before they ship.

03 / Publish

Draft, publish, retire

Every assistant has a status. Draft assistants are visible only to the build team. Published ones are live to their assigned teams. Retiring an assistant takes one click and preserves the history.

04 / Track

Version history

Every change to an assistant is versioned. You can see what changed, when, and by whom, and roll back to a known-good version if a new release doesn't behave the way you expected.

The platform is designed to be operated by a small build team inside your organisation, not by a vendor on your behalf. We can help; you don't need us in the room.

Inside your tenant. Not ours.

Ajutant doesn't run on our servers with your data passing through. It deploys into your own Azure subscription and stays there. We give you the platform; you keep the keys.

Your data never leaves

Everything runs within your Azure environment. No telemetry pipes your content back to us. No copies sit in a shared database we control. The boundary is your tenant boundary.

Your IT team can run it

Deployment is packaged and repeatable, designed for an IT team that knows Azure, not for a specialist AI engineering function you'd have to hire. We deploy it; your team operates it.

Your instance, your rules

Multi-tenant platforms route every customer through shared infrastructure. Ajutant doesn't. Your instance is yours alone, your access controls, your data residency, your rules. No shared backend to worry about.

Built for the conversation with your CISO.

Most AI tools treat governance as something you add later. In a regulated organisation, "later" is how projects die in review. Ajutant is governed from the first deployment, every control below is part of the platform, not a roadmap promise.

Full audit log

Every interaction recorded and reviewable. You can always answer "who asked what, when."

Acceptable-use gate

Your AUP is enforced at the point of use, not just written in a policy nobody reads.

Kill switch

Turn any assistant (or the whole platform) off instantly. Control doesn't wait for a support ticket.

PII guardrails

Sensitive data is detected and handled before it reaches a model, reducing exposure by design.

Role-scoped access

Teams see only their own assistants and data. Separation is enforced, not assumed.

Per-team cost controls

See and cap spend by team. Every use case has a visible cost, so nothing runs away unnoticed.

Feedback that stays yours

User reactions on whether an answer was useful route to the team that owns the assistant. The model provider sees none of it.

Issues straight to IT Ops

Ethical, explainability, or correctness concerns flagged by a user reach the operations team in real time.

The questions a security review asks (where does the data go, who can see what, how do we switch it off) all have answers before the review begins.

A starter kit, not the destination.

Ajutant ships with 76+ working assistants across the departments most organisations already run. They are not the reason to buy the platform; they are how a new build team gets going on day one without staring at a blank canvas. Use them, adapt them, ignore them. The value is what your team builds next.

The closest analogy: buying Salesforce and finding it ships with a Sales Pipeline template. Useful starting point, not the reason you bought it.

Legal

Practice & knowledge

Contract review, matter summaries, precedent search across your own document base.

HR

People operations

Policy answers, drafting support, onboarding guidance grounded in your handbook.

Finance

Finance support

Report drafting, policy lookup, and analysis assistants scoped to the finance team.

Recruitment

Hiring

Candidate validation and job-description authoring, built on your own criteria.

IT & Service

Internal support

First-line answers, triage, and knowledge retrieval for stretched service desks.

Sales & Marketing

Commercial

Drafting, research, and content assistants that keep your tone and your facts.

Organised into modules. Assistants come grouped by department, so a team gets a working set on day one. Publish what's useful, adapt what isn't, build whatever else you need from the same platform.

The right model for the right assistant. Opaque to the user.

Different work needs different models. A summarising assistant might run perfectly well on something cheap and fast. A complex legal analysis assistant might warrant the most capable model on the market. A customer service assistant might need both: fast triage, smart escalation.

Ajutant lets you choose the model per assistant. Cheap and fast where that's enough, smart and expensive where it matters. The end user never sees the choice; they just ask the assistant and get the answer.

And when something better comes along, swap it in. The governance, the assistants, and the controls don't change underneath you.

"Do what you want" isn't help.

Cheap and fast For volume work
Balanced Default
Most capable For complex work
Open-weight Run your own

The governed AI control plane for your whole organisation.

Most AI tools handle one part of the problem, usually the assistants people use. Ajutant handles all of it. Every AI request your organisation makes (whoever or whatever is making it) runs through one governed runtime, in your tenant, under your control.

01 / Assistants

For your people

Ready-built assistants your teams use directly, governed and grounded in your own documents. Priced per named user.

Used by: business teams, support, KM, internal services
02 / Headless

For your applications

AI baked into your own internal systems, document processing, classification, automated workflows. No user licence; consumption is governed and metered.

Used by: internal apps, batch jobs, integrations
03 / Gateway

For everything else

Any other tool that talks to a model routed through an OpenAI-compatible endpoint. Plus full REST APIs documented in OpenAPI. Your governance applies, even when it isn't your code.

Used by: third-party SaaS, dev teams, agents
One audit trail. One kill switch. One bill from Microsoft.
Whether the request came from a person, a process, or a third-party tool, it ran through the same governed runtime, in your tenant, with the same controls applied.

From one assistant to one hundred, without a new platform.

The second use case shouldn't cost what the first one did. Ajutant is built so that proving value and scaling it are the same motion, not two separate projects.

01

Start where the value is

Put a handful of governed assistants in front of a real team. Watch what they use daily and what they drop. Learn where the return actually is, cheaply.

02

Adapt and publish

Shape assistants to your own documents and rules with a draft-and-publish workflow. What works goes live; what doesn't stays in draft. Hours, not weeks.

03

Package and reuse

Export a proven assistant or module and reuse it across teams or sites. Your hundredth use case builds on the first, instead of starting over.

An AI project has a lot of stakeholders. There's something here for each of them.

Project leads
A proper build experience for new AI use cases, with a starter kit of working assistants. Ship in weeks, not quarters, and your team owns what comes next.
How you build →
AI developers
Prompt development interface, test bed, draft/publish workflow, version history, and a model gateway you control.
Architecture →
IT operations
Runs in your Azure tenant, audit and tracing into Log Analytics, kill switch, backup under your tenant policy, no specialist AI team required.
Architecture →
Security and risk
Your data stays in your tenant. Identity from your Entra ID. AUP enforcement, PII guardrails, ethical-issue escalation to IT Ops, never to the model provider.
Security →
Architects
Documented REST APIs (OpenAPI), slotting into your existing Enterprise Integration Platform or any other governed REST service. No special connector, no new integration tier.
Architecture →
Finance
Per-team cost caps, predictable platform fee, model spend metered in your own Azure bill, cheaper models for the work that doesn't need an expensive one.
Pricing →

See it running in a regulated context. In 30 minutes.

A 30-minute discovery call with someone who knows the platform end to end, not a sales intermediary reading from a script. We'll talk through your environment, your blockers, and how Ajutant fits, with straight answers to whatever your security team needs to know.