Answers from your own documents, with the source attached.
For organisations where expertise sits with a few people and everyone else waits for answers. We build AI assistants for specific roles that draw on your policies, procedures, price lists and past work, show where each answer came from, and only see what that person is allowed to see. Experienced staff are interrupted less, and everyone else produces better work, faster.
Signs you need this
If two or more of these sound familiar, there's likely time and money to recover.
The same questions, every day
Senior staff answer "what's our policy on X?" or "where's the latest template?" many times a week. Each interruption costs focus in the roles you can least afford to slow down.
Knowledge lives in a few heads
When a long-serving coordinator or estimator is away, work stalls or quality drops. If they leave, years of judgement leave with them.
Documents exist, nobody can find them
SharePoint, Google Drive and shared inboxes hold thousands of files with inconsistent names and versions. Staff recreate work that already exists, or use an outdated version.
New starters take months to be useful
Onboarding depends on shadowing busy people. Every extra month to competence is salary paid for partial output.
What changes for the business
Fewer interruptions for your experts
Routine questions are answered from the handbook, procedures and past work, with a link to the source. Your most experienced people spend more of the week on work only they can do.
First drafts in your own style
Assistants draft proposals, client emails, job notes and reports from your templates and previous examples. Staff edit instead of starting from a blank page, which shortens turnaround on revenue-producing work.
Answers you can check in seconds
Every answer cites the document and section it came from, and the assistant says when it can't find a source. Staff verify before relying on it, which keeps errors away from clients.
Access that matches your structure
The assistant inherits your existing permissions, so HR, finance and client files stay with the people already entitled to them. That reduces the chance of a well-meant question exposing information to the wrong team.
Faster onboarding
New starters ask how things are done here and get your answer, not a generic one from the internet. They become productive sooner and lean less on busy colleagues.
Know-how that survives resignations
Procedures and decisions are captured in maintained documents the assistant draws on. The business keeps what it knows when someone retires or moves on.
What we build
Knowledge source audit
We find, deduplicate and classify the documents that matter, and flag those that are out of date, contradictory or too sensitive to include.
Role-specific assistants
Assistants configured for defined roles, such as estimator, account manager, client services or practice manager, each with its own instructions, sources and boundaries.
Retrieval and citation layer
Search across your documents that returns the relevant passages with links, so answers are grounded in your material rather than the model's general knowledge.
Permission and data controls
Access mapped to your existing groups in Microsoft 365, Google Workspace or your platform, with sensitive categories excluded or restricted.
Prompt and template library
Tested instructions for common tasks, such as drafting a quote cover letter from an RFQ email or summarising a client meeting, so output is consistent across the team.
Accuracy testing and feedback loop
A test set of real questions with expected answers, run before launch and after every change, plus a one-click way for staff to flag a wrong or missing answer.
Technology we work with
- Microsoft 365 Copilot
- Copilot Studio
- SharePoint and OneDrive
- Microsoft Teams
- Google Workspace and Gemini
- OpenAI models
- Anthropic Claude models
- Postgres with vector search
- Cloudflare
How it runs
Timings are indicative and depend on your systems, data and approvals.
Scope the role
We pick one or two roles, collect the questions they ask and the documents they use, and agree how success will be measured.
1 to 2 weeks (indicative)
Prepare the knowledge
Documents are cleaned up, classified and connected, and gaps are listed for your team to fill.
2 to 4 weeks (indicative)
Build and test
We configure the assistant, run it against the test set, and tune sources and instructions until answers are accurate and cited.
2 to 4 weeks (indicative)
Pilot on real work
A small group uses it daily with feedback captured, and we fix gaps before wider release.
2 to 4 weeks (indicative)
Roll out and maintain
Training for each role, a named content owner, and a monthly review of usage, flagged answers and stale documents.
Ongoing
Worked example
Accounting firm: a knowledge assistant staff can trust, and an end to the document chase
An illustrative 60-person accounting and advisory firm loses senior time to routine questions, has no rules for AI, and chases client documents by email. The system starts with an AI use policy, then gives staff an assistant grounded in the firm's own procedures, with citations and permissions, and moves document collection into a client portal.
- Job created in the practice management system
- Portal sends the client a document checklist
- Uploads filed against the job
- Staff prepare the work
- Knowledge assistant answers from approved sources, with citations
- Manager and partner review
- Client sees status in the portal
Guardrails, fit and measurement
Guardrails we apply
- Assistants respect existing permissions. Microsoft states that Copilot only surfaces organisational data a user can already view, so we check SharePoint and Teams for overshared content before launch.
- No personal or sensitive information goes into publicly available AI tools, as the OAIC recommends. We use business services whose published policies say your inputs aren't used to train their models by default.
- Every answer cites its source, and the assistant is instructed to say when it doesn't know rather than guess.
- A person approves anything sent to a client, patient, member or regulator.
- You own the configuration, prompts and knowledge base, documented for handover.
Not the right fit if
- Your key documents are missing, contradictory or years out of date. An assistant will repeat the confusion faster, so we would start with a documentation sprint.
- You want an assistant giving clinical, legal or financial advice straight to customers without professional review. We build for staff use with a qualified person in the loop.
- A small team shares all its knowledge in one room. A well-organised shared drive and a business AI subscription may be all you need.
What we measure
- Weekly active users among the intended roles
- Share of answers carrying a valid source citation
- Accuracy on the test question set, re-run after each change
- Questions escalated to senior staff, before and after launch
- Time to first draft for target documents, such as proposals or reports
- Time for new starters to reach agreed competence
Pricing
For Marketing Department clients, an AI team assistant starts from $1,500 setup plus $300 a month (ex GST); larger knowledge systems are scoped individually, usually after an Altitude Diagnostic or a fixed-scope discovery.
Questions, answered.
Does this work with Microsoft 365?
Yes. If you already pay for Microsoft 365 Copilot, we often start there, using your SharePoint permissions and Copilot Studio for role-specific agents. Microsoft's documentation says Copilot only surfaces data a user already has permission to view, so we find overshared sites and folders before switching it on. Where Copilot isn't the right fit, we build on Google Workspace or a custom retrieval system.
Which AI models do you use, and will our data train them?
We work with OpenAI and Anthropic models and with the assistants built into Microsoft 365 and Google Workspace, choosing per task. We use business and API services whose published policies say your inputs and outputs aren't used for model training by default. Where data location matters to you, we confirm where processing happens before anything is connected.
What does an AI assistant cost?
For Marketing Department clients, an AI team assistant is available from $1,500 setup plus $300 a month, ex GST. Larger knowledge systems with several roles, integrations or custom retrieval are scoped individually, usually after an Altitude Diagnostic or a fixed-scope discovery.
What happens when it gets an answer wrong?
It will, occasionally, which is why every answer cites its source and staff are trained to check before relying on it. Wrong or missing answers are flagged in one click, reviewed monthly, and fixed by correcting the source document or the instructions. Anything leaving the business is approved by a person.
What if staff don't use it?
Low use usually means the assistant was built for a generic user, not a real role. We start with one role's actual questions, train people on their own work, and track weekly use so we can fix what isn't working rather than guess.
Related services
- AI-augmented workforceHelp your existing team do more valuable work with AI: role-based training, approved tools, prompt libraries and adoption tracking. Capacity, not cuts.
- AI governance & implementationAI governance for Australian businesses: a use policy, approved tools, data rules and human oversight, informed by OAIC and National AI Centre guidance.
- Client & customer portalsSecure portals where customers check status, share documents, approve work, book and pay. Fewer 'where's my…' emails, with permissions and audit built in.
- Business process automationAutomate intake, quoting, follow-up, onboarding and reporting across the systems you already use. Measured in hours released, with people on exceptions.
Where this matters most
Pick one role to start
Tell us which team loses the most time hunting for answers, and we'll show what an assistant for that role would need.
Sources
Third-party facts on this page, checked 14 October 2026.
- Data, Privacy, and Security for Microsoft Copilot, Microsoft Learn. learn.microsoft.com/en-us/microsoft-365/copilot/microsoft-365-copilot-privacy
- Guidance on privacy and the use of commercially available AI products, Office of the Australian Information Commissioner (OAIC). www.oaic.gov.au/privacy/privacy-guidance-for-organisations-and-government-agencies/guidance-on-privacy-and-the-use-of-commercially-available-ai-products
- Business data privacy, security, and compliance, OpenAI. openai.com/business-data
- Is my data used for model training?, Anthropic Privacy Center. privacy.claude.com/en/articles/7996868-is-my-data-used-for-model-training