AI and automation

Your team has AI. Nobody's sure what they're allowed to put in it.

We set up Microsoft 365 Copilot and the automations that take filing, drafting and data entry off your staff. Before any licence is bought we check who can see what in your files, write a plain AI use policy, then train each team on their own work. When something stops working or a new tool turns up, you call the same engineers who set it up.

Rules first, licences second. Australia wide.

  • Plain AI use policy
  • Files tidied before Copilot
  • Training by role

Reviewed by Blake Bath, General Manager, on 31 October 2026.

An engineer running a workshop with a small team around a table, a whiteboard with simple boxes and arrows behind
Rules before licencesPolicy, permissions, then rollout.

What changes

AI that saves time without leaking data.

The tools are the easy part. These are the parts that decide whether it works.

Everyone knows the rules

A plain policy on which tools, which data and who checks the output.

  • Approved tools
  • Data rules
  • Owned output

AI use policy starter

Draft a plain AI use policy in a minute.

Pick the tools you'll allow and the data you hold. Copy the draft, then we help you finish it and back it with settings.

Pick the tools you allow and the data you hold. You get plain policy lines to start from.

Which AI tools will you allow? (pick all)
What sensitive data do you hold? (pick all)
Could AI help make decisions about people? (hiring, credit, pricing, eligibility)

Your AI use policy, first draft

6 plain rules

  1. Use only these AI tools for work, signed in with your work account: Microsoft 365 Copilot.
  2. Don't use personal AI accounts or free tools for anything to do with the business.
  3. Don't paste customer names, contact details or identifiers into AI unless the tool is on the approved list.
  4. Client work stays in approved tools, and client contracts are checked for AI restrictions first.
  5. A person reads and owns everything AI drafts before it leaves the business.
  6. Report anything odd an AI tool does, or anything sent by mistake, to IT straight away.
  • Before Copilot goes live, check who can see what in SharePoint and OneDrive. Copilot shows people what they already have access to.
  • A starting point, not legal advice. We help you finish it and back it with settings, not just words.
Finish it with us

What goes wrong

Six ways AI projects stall.

  • Licences before groundworkCopilot bought before files and permissions are ready.
  • No policyStaff guess what's allowed, or stop using it.
  • Generic trainingA demo that doesn't connect to anyone's actual job.
  • Permissions too broadCopilot surfaces files people shouldn't have had access to.
  • No baselineNobody timed the task before, so nobody can prove the gain.
  • IT-only rolloutThe teams who'd use it weren't in the room.

The data

Most AI use is happening on personal accounts.

Verizon's 2026 Data Breach Investigations Report found regular generative AI use at work tripled in a year, from 15% to 45% of employees. Of that use on work devices, 67% went through personal, non-company accounts.

That's data leaving the business with no record of where it went. The fix isn't banning AI. It's giving people an approved tool and a clear rule.

Verizon 2026 DBIR figures as reported from the report, read 31 October 2026.

The engagement

What we do, step by step.

  • Discovery

    Which tools are in use now, which jobs AI could help with, and a baseline of how long they take.

  • Microsoft 365 readiness

    SharePoint structure, permissions and sensitivity labels, so Copilot finds the right things.

  • Copilot rollout

    Copilot in Word, Excel, Outlook and Teams, with settings documented.

  • Copilot Studio agents

    Single-purpose agents for IT, HR or customer questions, with limited access.

  • Automation

    Power Automate and AI Builder for invoices, forms and document routing.

  • Policy and controls

    Acceptable use policy, Purview and Defender settings, and an audit trail.

  • Training and coaching

    By role, with follow-up sessions and champions in each team.

  • Support

    The same engineers look after it once it's live.

Platform choice

Copilot, ChatGPT, Claude or Gemini?

The tool matters less than the data and the rules around it. Here's how we usually think about it.

ToolUsually fitsWatch for
Microsoft 365 CopilotBusinesses that live in Outlook, Teams and OfficeOnly as good as your SharePoint permissions
ChatGPT business plansA flexible standalone tool across teamsAnother place for data to live, outside Microsoft 365
Claude business plansLong documents, policy and careful writingAnother place for data to live, outside Microsoft 365
Gemini in Google WorkspaceBusinesses on Google WorkspaceGenerally not the pick for a Microsoft business

Our view. Check each vendor's current data terms for the plan you buy.

Where it helps

Jobs AI is good at today.

  • Document reviewPull out key terms and summarise long documents. A person approves the result.
  • Meeting notesSummaries and action items from Teams meetings.
  • DraftingProposals and follow-ups from your own examples. Your team edits, not writes.
  • Invoices and formsData pulled out and posted to your system, with review steps.
  • Staff questionsAgents that answer IT and HR questions and hand off to a person.

The Australian picture

Two things worth knowing in 2026.

  • Privacy policies and ADM

    From 10 December 2026, if a computer program makes or substantially helps make decisions that could significantly affect people using their personal information, your privacy policy must say so.

  • Agentic AI guidance

    ASD's ACSC and Five Eyes partners published joint guidance on the careful adoption of agentic AI services, stressing layered defences and strict access controls.

From the OAIC (published 30 September 2026) and the joint agentic AI guidance listing, read 31 October 2026. Not legal advice.

What we design around

Five AI risks for small businesses.

  • Sensitive data in public toolsPasted in, with no record and no way back.
  • Better phishingAI writes convincing emails and clones voices.
  • Shadow AITools spreading across teams on personal accounts.
  • Prompt injectionAgents redirected by instructions hidden in emails or documents.
  • Privacy obligationsPersonal information used in decisions without the right disclosure.

Free resources

Download and share with your team.

  • AI in your business right now

    An eight-page guide to the risks and opportunities, with practical first steps.

    Download PDF
  • 10 ways to stay secure using AI

    A four-page checklist on permissions, labels, MFA and private versus public AI.

    Download PDF
  • Securing tomorrow: AI and data protection

    A 12-page ebook on backup, data lifecycle and Copilot readiness.

    Download PDF
  • Microsoft's Copilot readiness questionnaire

    Microsoft's own workbook to self-score before you buy licences.

    Download workbook

How we start

Four steps before anyone buys a licence.

  1. Step 1Discover

    Tools in use, jobs to target, a baseline.

  2. Step 2Set the rules

    Policy, permissions and labels.

  3. Step 3Roll out

    Copilot or agents for the first jobs.

  4. Step 4Train and support

    By role, then ongoing.

Questions

What people ask before an AI rollout.

Where to start with AI in a small business?

With the rules and the data, not the licences. Work out which tools staff already use, write a plain AI use policy, tidy who can see what in your files, then pick one or two jobs where AI saves real time. The policy tool on this page gives you a first draft.

Does our data get used to train public models?

Microsoft says prompts, responses and the files Copilot reads aren't used to train its foundation models. Other vendors' business plans make similar commitments, but terms differ by plan, so we check the plan you're buying before we recommend it.

Where is Copilot data stored?

Australian tenants are covered by Microsoft's data residency commitments for Copilot in its Product Terms. We confirm your tenant settings during rollout and write them down, so you can answer client and procurement questions.

What's the difference between Copilot and a custom agent?

Copilot is a general assistant inside Word, Outlook, Excel and Teams. A Copilot Studio agent is built for one job, like answering HR policy questions or triaging IT requests. We build both, depending on what saves the most time.

What changes with the Privacy Act in December 2026?

From 10 December 2026, if a computer program makes, or substantially helps make, decisions that could significantly affect people, and uses their personal information, your privacy policy has to say so. If AI touches hiring, credit, pricing or eligibility, check with your lawyer now.

How do you secure AI agents?

In 2026 ASD's ACSC and its Five Eyes partners published joint guidance on the careful adoption of agentic AI services. It stresses layered defences and strict access controls. We scope each agent to one process, give it only the access it needs, and require approval before it takes any action beyond looking things up.

How do we measure whether it's working?

Where the job allows, we time it before rollout and again after. Useful measures include time on the task, turnaround, and requests an agent handles without a person. Results depend on adoption, not just licences.

Next step

Get clarity before you commit to licences.

A discovery session to map the jobs worth doing, check your Microsoft 365 readiness, and recommend the right tools and order. No obligation.

Chip, the Real Bytes support assistant

Cookie Preferences

We use cookies to improve your experience, analyse site traffic, and personalise content. By clicking "Accept All", you consent to our use of cookies. Privacy Policy

Privacy Act 1988 compliant. Your data is never sold.