The best way to introduce AI to your team is to start with what your people already do with it. Most staff are using AI tools before management has a plan. So the first job is to learn which tools, for which tasks and with what data. Then you set simple rules, talk openly about jobs, train people for the work they actually do, and review the results every quarter. A structured, anonymous staff survey is the fastest way to get that starting picture, and it is what our AI readiness assessment is built around.
This guide is for owners, managers, IT leads and planners at small and mid-sized Australian businesses. It covers the sequence that works, where each step tends to go wrong, and how to handle the question every team asks: “Will this cost me my job?”
CalibreforumSee How Your Team Uses AICalibre Forum’s AI readiness assessment gathers employee insights and turns them into a scored report with a clear 90-day action plan.VisitTL;DR
- Begin with visibility: find out which AI tools staff already use, for what, and with what data, before you buy or mandate anything.
- Pick one contained process, document it step by step, and introduce AI there first rather than across the whole business.
- Write a short AI policy that says what staff can and cannot paste into tools, who approves higher-risk uses and how to report problems.
- Address job fears directly and early. Consult staff, explain where AI will and will not be used, and keep humans accountable for decisions about people.
- Train by role, build AI into induction for new starters, and review results every quarter against specific, measurable use cases.
Table of Contents
- Why you are probably not starting from zero
- Step 1: Find out what your team already uses
- Step 2: Choose one process, not the whole business
- Step 3: Set the rules before you scale
- Step 4: Talk about jobs before staff do
- Step 5: Train by role, not by tool
- Step 6: Build AI into induction for new starters
- Step 7: Keep a human in the loop and watch the risks
- Step 8: Measure, then repeat every quarter
- Get the starting picture in one report
- FAQ
- Sources
Why you are probably not starting from zero
A common mistake is to treat AI adoption as a launch: pick a tool, announce it, run a workshop. In practice your team is likely ahead of you. A 2026 survey of 1,500 Australian digital workers by an AI software vendor found that 90% use AI at work and 72% say it makes them more productive. That is one vendor’s survey of digital workers, not a measure of your business, but it points at the same thing: staff try tools on their own, often on personal accounts, and management only sees the result when something goes wrong.

Board-level data points the same way. Research by HTI with 419 directors and senior executives in late 2025 found that 90% of surveyed organizations are using or planning to use AI, up from 64% in 2022.
So “introducing AI” really means three things at once: bringing existing, unofficial use into the open, choosing where AI should be used deliberately, and giving people the skills and rules to use it well. Doing it in that order protects you from the two usual failures. One is buying a tool nobody needs. The other is writing a policy that ignores how people actually work.
Step 1: Find out what your team already uses
The National AI Centre’s Get ready for AI guidance tells businesses to understand the likely impact on people, customers and stakeholders before introducing AI, and to engage workers to identify where AI could create value or risk. It also asks whether you have meaningfully consulted workers, are transparent about AI use and know what training your teams need.
To answer those questions you need honest information from staff. That is harder than it sounds:
- Asking in a meeting gets polite answers. If people think they might be in trouble for using an unapproved tool, they will say they do not use one.
- Asking only managers gets a partial picture. Managers often do not see how a receptionist, bookkeeper or sales coordinator uses a chatbot to draft messages or summarize documents.
- Checking systems alone shows logins, not workflows. It tells you a tool exists, not what staff paste into it or what they do with the output.
An anonymous survey of every staff member solves most of this. Our assessment works this way: each person answers a structured survey about the AI tools they use, their workflows and how those tools fit into their work. Nothing is installed and no individual data is collected. We combine the answers into one scored report that maps current capability, flags risks and sets out a 90-day action plan. You can see the contents on the what you get page and the setup on how it works. If you want to run your own survey, our guide on running an AI staff survey honestly and anonymously covers the question design and the trust issues.
The point of this step is not to catch anyone. Tell staff beforehand that the goal is to understand how work gets done and to decide what to support. People who use AI well are your best source of use cases.
What to look for in the results
- Which tools are in use, and which are on personal rather than business accounts.
- Which tasks they are used for: drafting, summarizing, research, data analysis, customer replies, meeting notes.
- What kind of information is going in: client names, financial data, health or employee details.
- Who is confident and who is avoiding AI altogether.
- Where staff say AI saves time, and where they say it creates rework.
Our resource on shadow AI in the workplace explains the account, browser extension and meeting recorder risks that usually surface at this stage.
Step 2: Choose one process, not the whole business
Once you can see current use, resist the urge to roll AI out everywhere. The National AI Centre’s Map your processes page advises starting with a single business process that is high volume, a pain point, data-rich and contained. It then asks you to document each step: the action, who is responsible, the inputs, the outputs and the time it takes. It also advises keeping an up-to-date record of every AI system in use.
A worked example: a 25-person accounting practice finds in its survey that several staff already use a chatbot to draft client emails and summarize ATO correspondence. Rather than rolling out a firm-wide tool, the partners pick one process: responding to routine client queries. They write down each step (who receives the query, what information is needed, who drafts the reply, who checks it, how long it takes). The map shows two steps where drafting help saves real time and one step where a human must always check the figures. Now AI has a defined job, an owner and a check, and the team can judge whether it worked.
Things that go wrong here:
- Choosing a process because it sounds impressive, rather than because it is repetitive and measurable.
- Skipping the step-by-step map, so nobody can say afterwards what changed.
- Letting one enthusiastic person run the pilot alone, so the knowledge never reaches the rest of the team.
If you want a longer-range sequence, our AI adoption roadmap lays out stages for a small business.
Step 3: Set the rules before you scale
Staff need to know what is allowed. The National AI Centre’s Create an AI policy page says a policy should state what AI can and cannot be used for, who approves higher-risk use cases, what data staff can put into tools, when staff need to oversee AI use, how to report issues or misuse, and when the policy will be reviewed. It should also set out core principles, roles and responsibilities, and approval points.
For a small business, a one-to-two page document is enough if it answers those questions in plain words. Three rules cover most of the risk:
- Data rules. Name the categories staff must never paste into a public tool, such as client identities, health information, financial records and employee details. Our guide on what staff can paste into AI tools under the Privacy Act gives the practical version.
- Approved tools. List which tools are approved for which tasks, and who to ask about a new one.
- Human checks. Anything that goes to a client or affects a person is reviewed by a human before it leaves the business.
For the drafting itself, see our AI policy outline and clauses for small business and the longer piece on what to include in an AI policy and how to enforce it. If you are unsure which obligations apply to you, our overview of AI regulations in Australia is a good place to check. Your security lead can work from the AI security checklist.
The government is also moving on workplace guidance. In November 2026 the Government released a draft Best Practice Guide on the Adoption of AI in the Workplace, built around three themes (Plan, Enable and Embed) and eight principles. These include identifying use cases and assessing workplace readiness, cooperative planning and worker engagement, building capability, and establishing AI governance. It is a draft, so check the department’s page for the final version. The sequence in this article already follows the same logic.
Step 4: Talk about jobs before staff do
If you say nothing about job security, people fill the gap with the worst version. The research on this is consistent in its advice: consult early, be specific and share the gains.
The Business Council of Australia’s report Adopting AI in the Workplace argues that good adoption needs executive leadership, deliberate deployment, human oversight and a workforce that is skilled, consulted and shares in the gains. It names wages, bonuses, training, job creation and reinvestment as ways productivity gains can flow back to people. You do not need to be a large employer to apply the principle.
In practice, a manager can do five things:
- Say what AI is for here. “We are using it to cut the time spent on first drafts and data entry, so you have more time for client work” is more credible than a general reassurance.
- Say what it is not for. If you are not using AI to measure individual performance or plan redundancies, say so in writing.
- Ask for input. Use the survey results and team conversations to find where staff think AI helps. People support what they helped choose.
- Share the gain. If a task that took five hours now takes two, decide in advance whether the time goes to better service, growth, training or reduced overtime, and tell the team.
- Keep humans accountable for people decisions. The BCA report says high-impact uses such as workplace surveillance, performance management, disciplinary processes or redundancy planning need more testing, human oversight and review than drafting or admin support. It also says human decision-makers must stay accountable for significant employment outcomes, including recruitment, performance management, disciplinary action, redundancy selection and termination.
Do not promise that no role will ever change. That is a promise you cannot keep, and staff know it. Promise a process: you will consult, you will explain, and you will retrain before you restructure.
Step 5: Train by role, not by tool
Generic “intro to AI” sessions fade within a week. Training sticks when it is tied to someone’s actual tasks. The National AI Centre’s implementation guidance says organizations should evaluate and document the training needed to build broad AI understanding and a culture of accountability, source or deliver training to close the gaps, and regularly check that skills stay up to date.
That gives you a workable structure:
- A shared baseline for everyone (about an hour): what the approved tools do, what they get wrong, what data stays out, and how to check output. Cover “hallucinations”, where a tool states something false with confidence, because that is the failure staff meet first.
- Role-specific sessions: the bookkeeper practices on invoice queries, the sales team on proposal drafts, the front desk on enquiry replies, using real (de-identified) examples from your own process map.
- A named go-to person per team who answers questions and collects tips.
- A refresh each quarter, because tools and your own rules change.
Cost is a real concern for small firms. The government’s gen AI transition case studies report that small businesses often lack time, resources and digital and AI capability, and one small business operator in the legal industry warned that the cost of training and upskilling would be a major barrier for many. The National AI Plan notes the government has committed $17 million to the AI Adopt Program to support SMEs, and that the National AI Centre released practical resources in October 2025, including editable AI policy templates simplified with business.gov.au. Using free official templates and in-house sessions keeps the cost of a first round low.
Our survey results make the training plan easier to build because they show where confidence is low and which tasks matter, so you train for gaps you can see. Our guide to AI adoption barriers covers the obstacles that most often hold small teams back.
Step 6: Build AI into induction for new starters
New starters are the easiest group to get right, because habits are not set yet. Add a short AI module to your induction:
- Day one: read the AI policy and sign that you have read it.
- Week one: a walk-through of approved tools and the data rules, with two examples of acceptable and unacceptable use from your own business.
- First month: a role-specific session with the team’s go-to person, and a check-in on what the new starter has tried.
- Ongoing: the same quarterly refresh as everyone else.
Ask new starters what tools they used in their previous jobs. They will often bring practices you have not seen, both good and risky, and it is a gentle way to find out before it becomes a problem.
Step 7: Keep a human in the loop and watch the risks
AI output goes wrong in ordinary ways: invented facts, outdated information, an inappropriate tone, a confidential detail in the wrong place. A simple routine prevents most of it:
- The person who sends it is responsible for it, whatever tool drafted it.
- Figures, names, legal or medical statements are checked against the source.
- Anything affecting a customer’s or employee’s rights or finances gets a second pair of eyes.
- Mistakes and near-misses are reported without blame, so the rules improve.
The Glean survey above found that 77% of Australian AI users report at least one form of “botshitting”, its term for passing along low-quality AI output without checking it. That was the highest rate of the three countries surveyed. It is a vendor survey, so treat it as a signal rather than a benchmark, but it supports the check-before-sending rule.
For deeper reading, see our guides on AI risk mitigation, AI model risk management and AI ethics policy.
Step 8: Measure, then repeat every quarter
Without measurement, AI adoption becomes a feeling. The AICD’s Director’s Guide to AI Governance notes that more value comes when AI transforms processes, systems or business models, yet relatively few organizations have moved beyond the pilot phase. It says return-on-investment assessments should focus on specific, measurable use cases and test whether the investment has improved efficiency and outputs.
For a small business that means picking two or three measures per use case before you start:
| Use case | Before | Measure after 90 days |
|---|---|---|
| Routine client replies | Time per reply, error rate | Time per reply, corrections needed |
| Meeting summaries | Time to write up | Time to write up, staff rating of accuracy |
| Proposal drafts | Hours to first draft | Hours to first draft, win rate |
| Policy compliance | Unknown tool use | Share of staff using approved tools only |
Then repeat the staff survey. Confidence, tool use and risk change as the team learns, and a repeat survey shows whether your training and rules are working. That is why we offer a recurring option: the same survey each quarter gives you a trend, not just a snapshot.
What we have learned about introducing AI
The businesses that struggle with AI are rarely short of tools. They are short of visibility. Owners tell us they do not know which tools staff use, and staff do not know what is allowed, so nobody is acting on a shared picture.
That is why we built our assessment around every staff member answering anonymously. When people can be honest without worrying about consequences, management gets a map of real workflows rather than a guess. We do not install software or collect individual data, because trust is what makes the answers useful. The scored report then turns that map into a 90-day plan, so the next step is concrete.
Our view is that introducing AI is a change-management task first and a technology task second. Start with the people, make the use visible, set plain rules, and revisit every quarter.
Anthony
Get the starting picture in one report
If you want the baseline from Step 1 without designing and running the survey yourself, our report does it. Every staff member completes an anonymous survey, and you receive a scored report covering current capability, risks and a tailored 90-day action plan. It costs A$1,990 as a one-off report. If you want the survey repeated each quarter to track progress, the membership is A$2,388 per year.

Industry-specific versions are also available, for example for managed service providers, recruitment agencies and medical clinics.
FAQ
How do I reassure staff who worry AI will replace their jobs?
Be specific rather than soothing. Explain which tasks AI will support, which decisions will stay with people, and how any productivity gain will be shared, for example through training or reinvested time. The Business Council of Australia recommends a workforce that is consulted and shares in the gains. Involving staff in choosing use cases also makes the change feel like something done with them.
How long does it take to introduce AI to a small team?
A sensible first cycle is about 90 days: one to two weeks to survey and map current use, a few weeks to set the policy and train, and the rest to run one contained pilot and measure it. Wider rollout comes after that, process by process. Rushing past the mapping step is the most common reason pilots stall.
Do I need an AI policy before staff start using AI?
Yes, and a short one is enough. The National AI Centre says a policy should cover what AI can be used for, what data staff can enter, who approves higher-risk uses and how to report problems. Because staff are probably already using tools, write the policy early and update it after your first review. Check which privacy obligations apply to your business with a lawyer if you handle sensitive personal information.
What should AI induction for new starters include?
The policy, the approved tools, the data rules and a role-specific session with examples from your own business. Add a check-in after the first month to see what they have tried and ask what tools they used before. Repeat a short refresher each quarter alongside the rest of the team.
Sources
- Get ready for AI | National AI Centre
- Map your processes | National AI Centre
- Create an AI policy | National AI Centre
- Guidance for AI adoption: implementation guidance | National AI Centre
- Adopting AI in the Workplace | Business Council of Australia
- A Director’s Guide to AI Governance | AICD
- Draft best practice guide for AI adoption in the workplace released
- National AI Plan
- Our gen AI transition case studies | Jobs and Skills Australia
- Work AI Index 2026: Australia
Recommended
- AI Policy in Australia: What to Include and How to Enforce It
- AI Governance Committee: How to Set One Up in Australia
- AI Security Awareness: What Your Staff Need to Know
- AI Adoption Roadmap: A Practical Plan for Australian Businesses
General information only. Articles on this blog are written with the help of AI writing tools and published by Calibre Forum. They are not legal, insurance, tax or financial advice, so check anything important with a qualified adviser.