AI Knowledge Assistants · Built on Microsoft Azure

Your organisation already
has the answers.
The problem is finding them.

Make your policies, procedures, and institutional knowledge instantly accessible — in the tools where your people already work.

Every organisation accumulates years of knowledge in policies, guides, SOPs, and the heads of experienced staff. The problem isn’t a lack of knowledge — it’s that the wrong person can’t find the right answer fast enough. AI knowledge assistants change that equation, permanently.

~1wk → 10min
Report generation time at NRMA after deploying NRM8, their AI knowledge assistant
3
Distinct AI personas at NRMA — NRM8, Chatm8, SIXTm8 — each grounded in different knowledge domains
100%
Data sovereignty — every assistant runs inside your Azure tenancy, no data leaves your environment
Weeks
From Discovery Workshop to first production deployment — not quarters

The knowledge is there.
The friction is in accessing it.

Most organisations don’t have a knowledge problem — they have a knowledge access problem. Decades of policy, process, and institutional expertise are locked in document libraries, intranet pages, shared drives, and the minds of long-serving staff. The result is wasted time, inconsistent answers, and decisions made on incomplete information.

The fifteen-minute search

Staff know the answer exists somewhere — in a policy document, an old email chain, a SharePoint folder someone else manages. Finding it reliably takes time they don’t have, so they ask a colleague, who asks someone else. The answer eventually arrives. Fifteen minutes after the moment it was needed.

Inconsistent answers

When the same question gets asked to three different people, three slightly different answers go back out. In regulated environments — insurance, education, aged care, NFP services — that inconsistency isn’t just inefficient. It creates compliance risk and erodes confidence in the institution.

Knowledge trapped in people

In organisations where a handful of senior staff carry most of the institutional knowledge, every departure is a knowledge loss event. New starters spend their first months asking questions that experienced staff find repetitive. The same knowledge gets re-explained, over and over, instead of being deployed once and made accessible permanently.

Outdated documentation nobody trusts

Document libraries fill up. Policies get updated but the old version stays on the intranet. Staff learn quickly that they can’t fully trust the documentation, so they verify with colleagues anyway. The documents exist to be useful; they stop being useful because nobody can trust they’re current.

Staff time misallocated

Subject matter experts — compliance officers, experienced nurses, senior educators, HR managers — spend disproportionate time answering the same questions from colleagues. This isn’t their highest-value work. AI knowledge assistants free them to do the work that actually requires their judgement and expertise.

The new starter disadvantage

New staff start at a significant knowledge deficit. Without fast, reliable access to the organisation’s accumulated knowledge, it takes months to become truly effective. The first person they ask shapes what they believe is true — which may or may not match the documented policy. An AI knowledge assistant removes that lottery from onboarding.

How it works

Your knowledge. Instantly accessible. Always sourced.

An AI knowledge assistant connects to your actual organisational content — not the internet — and answers in plain language, from the source, in the tools where your people already work.

Built on Microsoft Azure AI Foundry and deployed inside your own Azure tenancy. Every answer is grounded in your indexed content and cites the source, so staff can verify. Every interaction stays within your environment.

Connected to your documents

Your SharePoint libraries, policy documents, SOPs, training materials, and knowledge bases are indexed and made queryable. The assistant answers from your actual content — not generic training data — and cites the source document so staff can dig deeper if needed.

Role-based access controls

Different staff see different content. Microsoft Entra ID governs who can access what — a frontline worker gets answers relevant to their role; a manager gets access to additional content. Different personas can be deployed for distinct audiences: staff, students, volunteers, customers.

Where your people already are

Deployed natively inside Microsoft Teams, SharePoint, and web browsers — no new platform to adopt. Staff ask questions in the same interface they use for everything else. That familiarity is why adoption rates are high and why the knowledge actually gets used.

Fully inside your Azure tenancy

Every query, every response, every interaction log stays within your environment. Data sovereignty applies to the knowledge assistant exactly as it applies to everything else in Azure. No shared cloud platform, no data used to train external models.

Active, not static

As your documents are updated, the indexed knowledge updates. Usage analytics show which questions are being asked most frequently — surfacing gaps in documentation and opportunities to improve your knowledge base over time. The assistant gets more useful as you maintain it.

In production

Three organisations.
One consistent pattern: knowledge made accessible.

Across insurance, education, and the not-for-profit sector, the underlying challenge is the same — staff spending time looking for answers that should be instant. These are the deployments we’ve delivered.

Insurance · Enterprise AI · Azure

NRMA Insurance — NRM8, the workplace mate

~10min
Report generation (was ~1 week)
3+
Distinct AI personas deployed

NRMA is one of Australia’s best-known membership organisations — a large, operationally complex workforce spanning motoring, travel, tourism, and frontline services. As the organisation grew, staff spent significant time navigating fragmented systems to find information.

Antares deployed QBot inside NRMA’s Azure tenancy — branded through a staff naming competition as “NRM8” (NRMA Mate). The platform now includes the m8 series: NRM8 for general staff knowledge, Chatm8 and SIXTm8 for specific operational functions. Contact centres gained rapid access to accurate customer information. Talent teams had job ad creation automated. Technical teams found document search dramatically faster. Report generation fell from roughly a week to around ten minutes.

Security was non-negotiable: Azure AD SSO, role-based access controls, and private network endpoints ensure data sovereignty throughout.

“Since its introduction, QBot—known internally as the m8 series—has emerged as a vital enabler in NRMA’s journey toward AI & Automation adoption. By fostering greater awareness, championing ethical AI practices, and unlocking enhanced productivity across our business, it’s paving the way for a smarter, more innovative workplace.” Andy McCarthy, GM Technology Engineering, NRMA

Independent School · Multi-persona AI · K–12

Haileybury College — one platform, three audiences

3
Distinct personas — staff, students, parents
Azure
Fully inside Haileybury’s tenancy

Haileybury College presents a challenge common across large independent schools: a complex knowledge landscape spanning curriculum, policy, administration, and pastoral care — with three distinct audiences who each need different information delivered in different ways.

Antares deployed a multi-persona AI knowledge platform on QBot inside Haileybury’s Azure environment. Staff have an assistant grounded in school policies, procedures, and internal knowledge. Students can ask curriculum and admin questions within boundaries appropriate to them. Each persona operates from a separate knowledge scope, with tone and content access calibrated to the audience.

The result is a single platform that serves the whole school community — with governance that ensures each audience only accesses what they should, and that sensitive institutional knowledge stays where it belongs.

Deployed on QBot inside Haileybury’s Azure tenancy — distinct personas for staff, students, and parents, each grounded in the content and access controls appropriate to their role in the school. Production deployment · QBot by Antares · Azure AI Foundry

Not-for-Profit · Knowledge Access · National

Mission Australia — 450 programmes, one place to ask

2,800+
Staff across national operations
450
Programmes delivering frontline services

Mission Australia is a national NFP delivering around 450 programmes across Australia — homelessness, mental health, family services, and substance dependency support. With 2,800 employees and thousands of volunteers spread across metropolitan and regional sites, keeping a lean IT team and a geographically dispersed workforce connected to current, accurate information is a genuine operational challenge.

Antares has been Mission Australia’s strategic Microsoft technology partner across their digital transformation journey. Building on that foundation, the focus has turned to deploying an AI staff assistant — grounded in Mission Australia’s internal knowledge, accessible to staff wherever they work, and designed to answer the high-frequency questions that currently consume disproportionate time from central teams.

The Microsoft-based infrastructure already in place — Office 365, Azure, SharePoint — provides the platform the knowledge assistant needs to be deployed securely and quickly.

“If we provide people with the most efficient tools and the most efficient ways to do things, that allows them to spend more time with clients. In the future we want to be able to create smarter applications that help us serve clients better.” Peter Smith, CIO, Mission Australia

Where it gets deployed

The same capability.
Different audiences, different knowledge, different outcomes.

An AI knowledge assistant can serve multiple different audiences from a single platform — each with their own knowledge scope, tone, and access controls. Here are the most common deployment patterns we see across sectors.

Staff knowledge assistant

Your staff generate more ad hoc knowledge queries than any other audience — HR policy, IT procedures, compliance requirements, financial processes, procurement rules. Most of those queries have a documented answer. The assistant surfaces it in seconds, cites the source, and frees central teams from fielding the same questions repeatedly.

Most useful for organisations with distributed workforces, complex policy environments, or high rates of staff turnover — where consistent, reliable knowledge access is a daily operational need rather than an occasional requirement.

HR policy and entitlements queries
IT and system support, first-line resolution
Compliance and regulatory requirements
Procurement, finance and approvals processes
New starter onboarding — consistent, always-on answers

In production

NRMA Insurance (NRM8) Staff across motoring, travel, and frontline services access policies and process documentation via Teams. Report generation time dropped from ~1 week to ~10 minutes.
Haileybury College (staff persona) Teaching and operational staff can query school policies, procedures, and institutional knowledge instantly — without waiting on administration.
Mission Australia Knowledge assistant planned for 2,800+ staff across 450 national programmes — grounded in Mission Australia’s service delivery documentation and internal policies.

How to get started

From first conversation to
production deployment — weeks, not quarters.

The most common barrier to deploying an AI knowledge assistant isn’t the technology — it’s the absence of a structured starting point. Our engagement model is designed to give you clarity before you commit and something working in production as fast as possible.

Discovery Workshop

Fixed fee · 2 weeks

We map your knowledge landscape — what content exists, where it lives, who needs access to what, and which assistant use cases will deliver the most immediate value. Deliverable: a concrete deployment plan with sequenced priorities and effort estimates.

Knowledge Foundation

Typically 2–4 weeks

We prepare your source content — indexing SharePoint libraries and document repositories in Azure AI Search, configuring retrieval pipelines, and standing up your Azure AI Foundry environment inside your tenancy. The assistant has something real to draw from before we deploy it.

First Deployment

Typically 4–6 weeks total

First production assistant goes live inside Teams and SharePoint. Role-based access controls configured. Usage analytics activated. The assistant is in the hands of real users, answering real questions from your real content — not a demo, a live production deployment.

Expand and Manage

Ongoing sprint delivery

Additional personas, broader content indexing, deeper integrations, and ongoing tuning — delivered in fortnightly sprints by a named team that knows your platform. Usage data drives prioritisation. The assistant evolves as your knowledge base evolves.

Why Antares

We’ve built this for
organisations like yours — in production.

The case studies above aren’t representative deployments or controlled pilots — they’re live systems, used daily, by real staff. That production experience is what we bring to every new engagement.

Microsoft-only practice

Over 20 years of Microsoft partnership and a practice built entirely on the Microsoft stack. We use Azure AI Foundry, Azure AI Search, Semantic Kernel, and Microsoft Entra ID on every engagement — because that’s all we do. That depth matters when the architecture decisions get complex and the governance requirements get specific.

Data sovereignty by design

Every assistant we build runs inside your Azure tenancy. Your documents, your queries, your interaction logs — none of it leaves your environment. We don’t offer this as an option; it’s the only architecture we use. For organisations in regulated sectors, that’s not a preference — it’s a requirement.

Sector depth, not just platform depth

We understand what knowledge management looks like inside an insurance contact centre, an independent school, a national NFP delivering community services, and a health organisation operating across distributed sites. That sector knowledge shapes the architecture and adoption approach from the start — not as an afterthought.

Production in weeks

A two-week Discovery Workshop to scope the work. Four to six weeks to a first live deployment. We’ve designed the engagement model specifically to reduce the time between first conversation and something your organisation actually uses — because the knowledge value is in the deployment, not the planning.

Active after go-live

Deploying an assistant and walking away doesn’t work. Knowledge changes, usage patterns surface gaps, new content needs to be indexed, and access controls need to evolve as your organisation does. Our managed service provides ongoing monitoring, content updates, and capability development in fortnightly sprints — with a named Technical Lead who knows your deployment.

Australian, local, direct

Based in Sydney and Melbourne. Every engagement is led and delivered by an Australian team with no offshore handoff. The team that runs the Discovery Workshop is the team that builds the deployment. There’s no account management layer between you and the people doing the work.

Common questions

Questions we hear at
the first conversation.

An AI knowledge assistant is an AI agent that’s connected to your organisation’s existing content — your policies, procedures, SOPs, product guides, training materials, and knowledge bases — and allows staff to get instant, accurate answers in plain language without having to search through documents or wait for a colleague. Unlike general-purpose AI tools, it’s grounded specifically in your organisation’s content, enforces your access controls, and keeps all data inside your Azure tenancy.
Yes. Microsoft Copilot improves individual productivity within Microsoft 365 — helping staff draft emails, summarise meetings, and manage tasks. An AI knowledge assistant is designed to give staff authoritative answers from your organisation’s own content — policy documents, SOPs, internal knowledge bases, operational documentation — with role-based access and full audit logging. The two are complementary and many organisations run both.
Accuracy comes from retrieval-augmented generation (RAG) — the assistant retrieves the most relevant content from your indexed source documents before generating an answer, and cites the document so staff can verify. This means it answers from your actual policies and documentation, not from general training data. Maintaining accuracy over time requires active content management — keeping source documents current and the knowledge base in good shape. That’s part of what our managed service covers.
No. Every AI knowledge assistant we build is deployed inside your own Microsoft Azure tenancy. Your documents, your staff’s queries, and the assistant’s responses never touch a shared platform or leave your environment. The same data governance controls that apply to your Azure environment apply to the knowledge assistant. This is an architectural principle — not a feature you configure.
A first production knowledge assistant can typically go live within four to six weeks. The main variable is the state of your source content — whether your documents are accessible, reasonably current, and organised in a way that can be indexed effectively. Our Discovery Workshop (two weeks, fixed fee) assesses this and produces a deployment plan before you commit to the build. Most clients see their first production deployment within six weeks of completing the workshop.
Yes. A single platform can support multiple AI personas — each with their own knowledge scope, tone, and access controls. Haileybury College deploys separate personas for staff, students, and parents from a single QBot instance. NRMA’s m8 series includes NRM8, Chatm8, and SIXTm8, each serving a different function with a different knowledge base. Microsoft Entra ID governs what each persona can access, so there’s no risk of one audience accessing content intended for another.
A knowledge assistant needs active management to stay useful — your documents change, new content needs to be added, usage patterns surface gaps, and access controls need to evolve. Our managed AI service provides ongoing content updates, performance monitoring, capability development in fortnightly sprints, and monthly roadmap reviews — delivered by a named Technical Lead who knows your deployment. Many clients transition from a project engagement into managed service once the foundation is live.

Ready to make your organisation’s knowledge actually accessible?

Start with a fixed-fee Discovery Workshop — two weeks to map your knowledge landscape, identify the highest-value use cases, and produce a concrete deployment plan. Or book a 30-minute conversation first to talk through your situation.

We work with organisations across insurance, education, not-for-profit, health, and financial services. Sydney and Melbourne-based team — available for in-person sessions.
Scroll to Top