A business is ready for AI when it can point to one specific, repetitive task, describe how that task runs today step by step, trust the data the task depends on, name the person who will check the output, and show that a simple rule or a connection between two systems wouldn't do the job just as well. If you can't tick those, the money is better spent fixing the process and the data first. The checklist below covers six areas: processes, data, systems, people and approvals, risk and privacy, and the "would a rule do?" test. Score each line honestly. The gaps tell you what to fix before you pay for an AI product or project.
Why readiness is mostly about process and data
Most AI readiness checks on the market are really a check of your Microsoft setup: licences, permissions and security settings. Those matter to your IT provider, but they don't answer the owner's question, which is "will this actually save us time, and can we trust it?"
That depends on things that sit outside any one product. Where do the hours go? Which steps involve information typed in again (double data entry)? Which systems disagree? Who checks the work? AI that reads a supplier invoice is only useful if the invoice can then go somewhere reliable, and if someone knows which version of the supplier record is right.
The National AI Centre takes the same view. Its free planning tools include a business process mapping template and a data quality checklist, and they're worth downloading alongside this list.
1. Processes
- ☐ You can name one task, not a theme. "Enter supplier invoices from three suppliers into the job system" rather than "use AI in finance".
- ☐ The task is frequent and repetitive. It happens daily or weekly, in volume.
- ☐ It's written down. Someone could follow the steps without asking. If not, start with a standard operating procedure or a quick process map.
- ☐ You know roughly how long it takes today, timed on a normal week, so you can tell later whether anything improved.
- ☐ The right answer is clear. A person can check the result quickly against the original.
- ☐ The process is stable. If the steps change every month, fix the process first. Automating a messy process just makes the mess faster.
2. Data
- ☐ You know where the data for this task lives, and which system holds the master copy.
- ☐ Customer, supplier and item records are reasonably clean. Duplicates, old codes and missing details confuse people and AI alike. Our guide to cleaning up duplicate records shows what to look for.
- ☐ There's one number everyone uses. If two reports give two different answers, AI will pick one of them. A single source of truth (one place everyone takes the number from) comes first.
- ☐ You have enough real examples of the documents or records involved to test with, including the awkward ones.
- ☐ You know what's sensitive. Personal, health, pay, bank or commercially sensitive information is identified, so you can keep it out or handle it properly.
3. Systems
- ☐ The output has somewhere to go. The system that receives the result can accept it, through an import, a connection or a screen a person uses.
- ☐ You've checked your existing software first. Many accounting, job and document systems now include features that read bills or suggest coding. Switch those on before buying something new.
- ☐ You know which steps cross systems. These hand-offs are where time goes and errors creep in. Our page on what happens between your systems explains why.
- ☐ You know whether the systems can connect. Ask each vendor whether there's a way to read and write data automatically. Our list of questions to ask a software vendor helps.
4. People and approvals
- ☐ A named person owns the task and will judge whether the AI is helping.
- ☐ A person approves every output before it reaches your accounts, customers or suppliers, at least at first.
- ☐ The check is quicker than doing the work. If reviewing takes as long as the original task, the gain isn't there.
- ☐ The team knows why. People are more likely to support a tool that removes the boring part of their job than one that appears without explanation.
- ☐ You know who can switch it off, and the manual process still works if you do.
5. Risk and privacy
- ☐ You know which AI tools staff already use. The National AI Centre calls unapproved use "shadow AI" and recommends clear guidance on which tools are acceptable (Manage unapproved AI use).
- ☐ You have a short AI use policy, or a plan to write one. The National AI Centre's free AI policy template is a good start, and our AI policy for staff template gives you a short version to adapt.
- ☐ The tool's data terms are checked. Business plans from major providers state whether your inputs are used to train their models. Check the terms for your exact plan and keep a copy.
- ☐ Privacy obligations are understood. The OAIC's guidance on commercially available AI products is the reference. If your business is covered by the Privacy Act, then from 10 December 2026 your privacy policy must also describe decisions that software makes or substantially helps make about people, where they could significantly affect them. That includes simple rules as well as AI. Our guide to the Privacy Act automated decision-making disclosure explains what to list.
- ☐ The task doesn't decide something important about a person (hiring, credit, pay) as a first project. Start with paperwork, not people.
6. Would a simple rule do the job?
This is the question most AI pitches skip. AI is good at reading messy input: PDFs, scans, emails and free text. It's the wrong tool for jobs a plain rule handles reliably. Tick the ones that apply:
- ☐ The input is already structured (a spreadsheet, a form, an export), so a rule or an import can read it.
- ☐ The decision follows fixed criteria ("approve if under the limit and the supplier is approved"). Our page on automating routine decisions covers this.
- ☐ The real problem is two systems that don't talk, so a connection would remove the typing without any AI.
If you ticked any of these, look at plain workflow automation or an integration first. It's usually cheaper, more predictable and easier to check. AI earns its place where the input is messy and a person still approves the result.
How to read your score
- Mostly ticked in all six areas: you're ready for a small, time-boxed pilot. Our guide to running an AI pilot safely sets out the steps.
- Gaps in processes or data: fix those first. Mapping the process and cleaning the data usually saves time on its own, before any AI is involved.
- Gaps in people or risk: write the one-page policy, name an owner and decide the approval step before switching anything on.
- Several ticks in section 6: a rule or a connection may beat AI for this task. Our article on where AI actually helps in operations gives examples of both.
Frequently asked questions
What is an AI readiness assessment?
It's a check of whether your business can use AI usefully and safely. Done well, it looks at how your processes run, the quality of your data, how your systems connect and who approves the output, then says which tasks suit AI, which suit a simple rule, and which should be left alone. A licence or security check is a different, narrower job.
Do we need perfect data before using AI?
No, but the data for the task you choose needs to be good enough that a person can trust and check the result. Start with one task whose data you can clean, rather than trying to fix everything.
How do we know if a task is suitable for AI?
Good candidates are frequent, repetitive, based on messy input such as PDFs or emails, easy for a person to check, and low harm if the first pass is wrong. If the input is already structured, a rule or integration is usually better.
Is there a government AI readiness checklist?
The National AI Centre publishes free planning tools on ai.gov.au, including a business process mapping template, a data quality checklist, questions to ask AI suppliers, an AI policy template and an AI register template. They're general and worth using. This checklist adds the cross-system view: where data moves between your systems and where a rule beats AI.
When you want an independent view
If the checklist shows gaps you can't easily judge yourself, our AI readiness assessment is the next step. It's carried out as part of our $5,000 Process & Systems Audit: we map how your key processes run across your systems and spreadsheets, check the quality of the data they depend on, and give you a written report that says which tasks suit AI, which suit a simple rule, and which to leave alone. It isn't a Microsoft licence check. Kenomont was founded by Aaron Chambers and brings 20+ years of building business systems, including a forestry export and port system that receives ticket data from the weighbridge system.
Not sure yet? Book a free initial chat and we'll tell you honestly whether an assessment is worth it for you.