How to use this

Work through the four sections below for one specific process you are considering automating — not for your business in general. General readiness is not a useful concept; a business can be perfectly ready to automate its expense processing and completely unready to automate its customer support.

Mark each item yes, no, or partly. At the end, count. The scoring guidance tells you what to do next.

Section 1 — Process

1.1 The process can be described in one page of plain English. What comes in, what decisions are made, what the rules are, what comes out. If every attempt collapses into "it depends", you have found undocumented judgment that must be surfaced first.

1.2 The volume justifies the effort. Ten occurrences a week is a rounding error. Two hundred is a business case. So is a low-volume process where each miss is expensive.

1.3 The process is currently working. Automating a broken process makes it broken faster and at scale. Fix it on paper first.

1.4 The exceptions are known and listed. You can name the situations where the normal rules do not apply. These become your escalation rules.

1.5 There is a defined start and end. You know what triggers the process and what "done" means. Processes without a clear end state are hard to automate and harder to measure.

Section 2 — Data and systems

2.1 The information exists in digital form. A spreadsheet counts. A whiteboard, a paper diary, or one person's memory does not.

2.2 There is source material to ground the system in. Policies, price lists, FAQs, past work, standard responses. The difference between accurate answers and plausible invention is whether the system is anchored to your actual documents.

2.3 The source material is current and singular. One version of each policy, dated, with an owner. Two versions produce two answers.

2.4 The destination system can be read. An API, a webhook, a monitored inbox, a shared folder.

*2.5 The destination system can be written to.* Verified with the vendor, in writing, including whether your current plan permits it. This is the item that most often changes the shape and cost of a project, and the one most often assumed rather than checked.

2.6 There is a test environment. Or an agreed safe way to test without touching live data.

Section 3 — People

3.1 There is one named internal owner. Not the vendor. A person in your business responsible for the source material, the exception queue, and the decision to continue or stop. This item is skipped more often than any other and is fatal more often than any other.

3.2 The owner has time allocated. One to two hours a week. Unfunded ownership is not ownership.

3.3 The people who do the work today have been consulted. They know the exceptions, the edge cases and the real rules. Building without them produces a system that handles the documented process rather than the actual one.

3.4 The framing is honest. Your team knows what this is for and what it means for them. Projects framed as headcount reduction acquire opponents; projects framed as removing the worst part of everyone's week acquire owners.

3.5 Someone will read the outputs. For the first fortnight, a human reads real transcripts or real outputs daily. This is where most of the early learning comes from.

Four or more of six readiness conditions means start. The one that most often decides the outcome is whether a named person owns it.

Section 4 — Security, data protection and control

4.1 Data boundaries are defined. What the system may access, and what is explicitly excluded — personal correspondence, HR files, board material, anything under legal privilege.

4.2 Your data does not train external models. Confirmed in writing with the provider. Vague answers here are the answer.

4.3 Access is scoped and auditable. The system reads what it needs, writes only where permitted, and every action is logged.

4.4 Personal data obligations are addressed. Under Singapore's Personal Data Protection Act, conversation logs, extracted documents and customer records processed by these systems fall within your obligations. Include them in your data inventory, set a retention period, and know where they are stored.

4.5 Human approval points are defined. By transaction type and by value. Which actions require a person, and which require two.

4.6 Uncertainty behaviour is configured and tested. When the system does not know, it says so and hands over. Test this deliberately by asking it things it should not be able to answer.

4.7 Disclosure is decided. Where customers interact with an automated system, you have decided how and where you tell them. Disclosed automation is far better tolerated than discovered automation.

4.8 There is a rollback. If it goes wrong, you know how to switch it off and revert to the manual process without losing anything.

Scoring

Count your yes answers across all four sections (out of 24).

20 or more. Scope the pilot now. Remaining gaps are detail and are best resolved by building something small.

15 to 19. Start, but close the gaps inside the pilot. Identify which section you are weakest in — it is usually Section 2 or the ownership items in Section 3 — and fix those first.

10 to 14. Not yet. The preparation is typically two weeks of writing things down and confirming integrations, and doing it first will roughly halve the cost of the eventual project.

Under 10. Choose a different process. There is almost always another candidate in the same business that scores far better.

What this checklist deliberately ignores

Your industry. Your headcount. Your technology stack. Whether anyone on your team has used AI before. Whether your competitors are doing it.

None of those predicts whether a specific automation will succeed. The twenty-four items above do.

The next step

Once you have a process scoring 15 or above, write the one-page brief: the job in operational terms, the escalation rules, the baseline, the target, the owner, and the stop date. That page is what you take to any provider, and it is what turns an ambition into something that can succeed or fail on evidence.

If you would like an outside read on your score and your shortlist, an AI Readiness Assessment produces a prioritised view for your specific operation — including an honest opinion on the processes you should leave alone.

Start with a free AI Readiness Assessment

Normally valued at SGD 1,500, currently free. You receive an opportunity report, a prioritised roadmap and honest ROI estimates for your own processes — with no obligation. Book yours at aigentify.tech/assessment.

AIgentify — Singapore-based AI implementation specialists. We design, build and support AI agents, workflow automations and intelligent business applications with measurable ROI. Live in weeks, not months. Your data stays yours.

This article is general information, not advice. Grant schemes, platform rules and regulatory requirements change; confirm current details with the relevant authority or provider before relying on them. Any figures shown are illustrative unless a source is stated.