AI readiness checklist for SMEs: data, tools and owners

By KYN AI Advisory Team — AI implementation specialists, Singapore

You are ready to automate a workflow when four things are true: the data it needs sits in one place you can reach, the tools it touches have a way in, a named person owns the outcome, and you know which actions must wait for approval. This checklist has 24 yes or no checks across those four areas. Score it per workflow, not per company. A business can be ready for one workflow and not for another.

Use it before you talk to a vendor. It also shortens the review we run at the start of an engagement.

How do you use this checklist?

Pick one workflow. Examples: replying to enquiries, chasing overdue invoices, reconciling the bank, sending a daily operations report. Answer each check for that workflow only. Count the yeses.

| Score | Reading | |---|---| | 20 to 24 | Ready. Start the build. | | 14 to 19 | Close. Fix the gaps first, and some can be fixed inside the project. | | Under 14 | Not yet. Spend a month on data and ownership before buying anything. |

These bands are our rule of thumb, based on how projects go, not a standard. They tell you where to spend effort, nothing more.

Data: can the system see what it needs?

Six checks.

1. The data for this workflow lives in a small number of places (ideally one or two).
2. You can export or query it without asking a vendor for a special report.
3. Key fields are filled in most of the time. Test with 50 recent records, not a feeling.
4. Names, company names and identifiers are spelled consistently, or can be matched.
5. Duplicates are rare, or you know how to find and merge them.
6. You know which data contains personal information and who may see it.

Weak data is the most common blocker. An agent writes confidently from stale or inconsistent records, so output quality is capped by input quality. If check 3 fails, the first task is cleaning, and the AI waits.

For a spreadsheet-based operation, see moving warehouse stock tracking off Excel for how that cleaning step works in practice.

Tools: can the system get in and act?

Six checks.

1. Each tool in the workflow has an API, a scheduled export or a mailbox the system can read.
2. You hold admin rights to those tools, or know who does.
3. You can create a separate login for the system instead of lending a person's login.
4. The tool's terms allow automated access.
5. Your cloud accounts, domains and key subscriptions are in the company's name, not a former employee's or a vendor's.
6. You can test changes somewhere other than live data.

Check 11 is the one that surprises people. If your website host, domain or messaging account is under an individual's name or a vendor's, a build will stall on access. Settle ownership first.

People: who owns it?

Six checks.

1. One named person owns the outcome of this workflow.
2. That person has time in the next 90 days to answer questions and review output.
3. The people who do the work today have been told, and at least one is willing to describe how they do it.
4. You can state the current cost: hours per week, errors per month or delay in days.
5. There is a named approver and a backup for any outbound or financial action.
6. Someone, internal or external, will maintain the system after launch.

Check 14 predicts success better than any technical check. Projects stall when the owner has no time. We return to this in why most SME AI pilots stall after 90 days.

Check 16 matters because you cannot judge a result without a baseline. Write it down before the build.

Risk: do you know where it can hurt?

Six checks.

1. You have listed the actions that must wait for a person: outbound messages, money, filings, deletions.
2. You know what happens if the system is wrong once. You can say who is affected and how you would correct it.
3. You have a rule for unusual cases, such as complaints, legal mentions and large amounts, and it says "stop and hand over".
4. You know where customer data will be stored and which vendors will see it. For Singapore data, see PDPA for AI systems.
5. You have a way to see what the system did: a log with who or what acted, and when.
6. You can turn the system off quickly without losing data.

Checks 19 and 21 turn a risky system into a safe one. They cost little to define and a great deal to skip.

What if you score low?

Look at where the misses cluster.

• Misses in data. Spend four to six weeks on a clean-up and a single source of truth. A tidy spreadsheet is a fine start.
• Misses in tools. Move accounts into company names, request API access, add test environments. Some of this takes a call to a vendor.
• Misses in people. Name the owner and free their time. No technical fix substitutes.
• Misses in risk. Write the approval list. It takes an afternoon.

Low scores are not a verdict on AI. They are a to-do list.

What if you score high?

Choose the workflow where the cost is visible and the scope is narrow. Good first candidates:

• sorting inbound enquiries and logging them;
• chasing overdue invoices on a fixed schedule;
• matching bank lines to ledger entries, with exceptions to a person;
• a daily operations report by message to each department.

Each has a right answer you can check, and each can run with a few gates. Broader scopes can follow once the first one runs.

How long does the review take?

A review of one workflow against this checklist takes a few hours of conversation if the owner is present. A review of a whole business takes longer, because each department has its own data and tools. Our first 14 days with a client are devoted to this: an inventory of the stack and interviews with the team.

Where KYN fits

KYN builds custom systems and agents for SMEs in Singapore, Hong Kong, Indonesia and Malaysia. We use the first 14 days as a review of your stack, which covers the data, tools, people and risk questions above. A full build ships in under 90 days.

Systems run in a cloud account in your name, with named accounts, row-level security and a log of every action. Outbound messages, money, filings and deletions wait for a named person. See operations and infrastructure, or run a free check of how AI search sees your business.

Frequently asked questions

Do I need clean data before I start?

You need data that is good enough for one workflow, not a perfect company-wide dataset. Test 50 recent records. If most fields are filled in and consistent, start. If not, clean first.

Do I need an in-house technical person?

No, but you need an owner. The owner understands the business process and has time to review. A builder handles the technical side.

How often should I redo this checklist?

For each new workflow, and once a year for the ones already running. Tools, staff and data change, and the answers change with them.

Checked 30 October 2026.

Curious whether Marketing agents fits your business? Talk to KYN on WhatsApp — no forms, just a conversation.

Start Building →