BonoficiGrade your business
Menu
Guide9 min read

AI Readiness Assessment: 12 Questions to Score One Process

A 12-question AI readiness assessment scored out of 24, with score bands, a copyable self-assessment checklist, and how to pick the first process to automate.

An AI readiness assessment checks whether a specific piece of work is ready to be handed, in part, to AI: whether the process is understood, the data is reachable, someone owns the result and a mistake would be affordable. Score the 12 questions below from 0 to 2 for one process, add them up out of 24, and the score band tells you what to do next. Assess a process, not the whole company; a company is ready in some places and not in others.

Why assess a process, not the company

Most AI readiness questionnaires ask about the whole organisation: strategy, culture, data maturity, budget. The answers are usually “somewhat” across the board, and they don’t tell you what to do on Monday. Your accounts payable team might be ready to automate invoice matching tomorrow while your quoting process lives in one person’s head and isn’t ready at all.

So pick one job that moves through the company, such as a customer order from the first email to the invoice, a supplier invoice from inbox to payment, or a quote request from enquiry to sent proposal, and assess that. If you have two or three candidates, score each and compare.

The 12 questions

Each question scores 0, 1 or 2. Be honest: a generous score now becomes an expensive surprise later. Where you are unsure, score low.

Area 1: the work

  1. Is the process written down, and does the written version match what people do? 0: nothing is written, or what is written is out of date. 1: there is a document, but people work around it. 2: there is a current map or procedure, checked against a real recent case.
  2. Is there enough volume? 0: it happens a few times a month. 1: a few times a day. 2: dozens of times a day or more, or a few times a day with each case taking real time.
  3. Are the rules clear and the exceptions known? 0: “it depends” and only one person knows on what. 1: the main rules are clear, but exceptions are handled case by case. 2: the rules are explicit and the common exceptions are listed, with what to do for each.

Area 2: the data

  1. Where does the information live, and can software reach it? 0: in inboxes, paper, or screenshots in a chat. 1: in a system, but only reachable by manual export. 2: in systems with an API or a scheduled export, such as NetSuite, Xero, QuickBooks, SAP or HubSpot.
  2. Is the data good enough? 0: duplicates, missing fields and free text everywhere. 1: usable, with known gaps people fix by hand. 2: consistent enough that a new starter could trust it.
  3. Is personal data involved, and is it handled properly? 0: personal data is involved and nobody has checked what may be processed or by whom. 1: personal data is involved and there is a policy, but no review of the tools. 2: no personal data, or it is covered by a data processing agreement with each vendor and a clear lawful basis.

Area 3: the people

  1. Is there one owner? 0: several teams touch it and nobody owns the whole thing. 1: there is an owner on paper, but hand-offs between teams are disputed. 2: one named person is accountable for the outcome end to end.
  2. Is it clear who approves? 0: approvals happen informally, by whoever is around. 1: there is an approver, but no rule for what needs approval. 2: the approval rules are written down, including limits and who covers absences.
  3. Will the people doing the work have time to teach it? 0: the team is underwater and nobody can spare an hour. 1: a lead can spare time, but the people doing the work can’t. 2: both the lead and two or three people doing the work can give short interviews and review the result.

Area 4: systems and risk

  1. Is there a system of record? 0: the “truth” is a spreadsheet someone maintains by hand. 1: there is a system, but key steps happen outside it. 2: one system holds the record, and any automation can read from it and write back to it.
  2. What does a mistake cost? 0: a mistake is costly or hard to reverse (money leaves the company, a customer is harmed, a legal deadline is missed). 1: a mistake is annoying and takes time to fix. 2: a mistake is caught by a person before it matters, or is cheap to reverse.
  3. Are there regulatory constraints? 0: the work falls under specific rules (for example credit, employment or health decisions) that nobody has checked against automation. 1: there are rules and they are understood, but untested with AI. 2: no specific constraints, or they have been checked and the design respects them.

What your score means

ScoreWhat it meansWhat to do next
0–8Not ready. The work isn’t understood well enough to automate.Map the process first. Interview the people who do it, write down the real steps and exceptions, and name an owner. Re-score in a month.
9–14Partly ready. Some steps could work, but gaps would cause failures.Fix the lowest-scoring area. Usually that means documenting exceptions or getting data out of inboxes and into a system. Use AI as an assistant to people, not to run steps.
15–19Ready for a first step.Pick one step, automate it with a person approving every result, and measure before and after.
20–24Ready.Automate the process step by step, keeping approvals where a mistake would be costly. Move approvals to sampling only once the error rate has been measured.
AI readiness self-assessment

Process: [name, from trigger to outcome, e.g. supplier invoice from inbox to payment]
Assessed by: [role]. Date: [date].
Interviewed: [lead of each team involved] and [two or three people doing the work].

#QuestionScore (0–2)Evidence (who said it, or where it is written)
1Process written down and matches real work[ ][ ]
2Enough volume[ ][cases per day or week]
3Clear rules, known exceptions[ ][ ]
4Data reachable by API or export[ ][systems involved]
5Data quality[ ][ ]
6Personal data handled properly[ ][ ]
7One owner (veto if 0)[ ][name the role]
8Clear approvals[ ][ ]
9Time to teach the process[ ][ ]
10System of record[ ][ ]
11Cost of a mistake (veto if 0)[ ][worst realistic mistake]
12Regulatory constraints checked[ ][ ]

Total: [ ] / 24. Vetoes (question 7 or 11 at 0): [yes / no].
Contradictions found between teams: [hand-off, what each team said].
Lowest-scoring area: [ ]. Next action: [ ]. Owner: [role]. Re-score on: [date].

Score with evidence, not opinions

The template has an evidence column for a reason. A readiness score filled in by one manager in a meeting tells you what that manager believes. Fill it in from short interviews instead.

  • Leads first, for breadth. Ask each team lead what their team does in this process, who they hand work to and which systems they use.
  • Then the people doing the work, for depth. Ask how a step really happens, how often, how long it takes, and what they do when it goes wrong: the version on a Tuesday when a supplier is late, not the one in the procedure.
  • No quote, no score. Only give a 2 where someone said it, or you can point to where it is written. “Exceptions are rare” is a belief until the person handling them says so.
  • Look hardest where teams disagree. If sales says orders are confirmed on the day and the warehouse says they hear about them two days later, the hand-off between them has no owner. Question 7 scores 0 until that is settled, however good the rest looks.

Our course on mapping your business processes covers these interviews in detail, and the lesson on documenting the exceptions will raise your score on question 3.

How to pick the first process

If several processes score well, choose using four filters, in this order:

  1. Revenue side first. Time back is good; more revenue is better. A faster quote, a follow-up that no longer gets forgotten, or an order confirmed the same day often matters more than an hour saved in the back office. Look at sales and customer-facing work before finance and admin.
  2. High volume. A step that happens hundreds of times a month pays back the effort of building and checking it. One that happens weekly rarely does.
  3. Clear rules. Steps where a new starter could follow a written rule: reading fields from a document, classifying an email, matching two records, drafting a reply from known facts. Leave judgement calls with people for now.
  4. Low cost of error. Start where a mistake is caught before it reaches a customer or a bank account, and keep a person approving every result until you have measured how often it is wrong.

Then pick one step inside that process, not the whole process. Measure it before you change anything: cases per week, time per case, error or rework rate. After a few weeks, measure again. If you are already looking at accounts payable, the three-way match guide is a worked example of a step with high volume, clear rules and a person kept on the exceptions. Before anyone starts using AI tools on the work, make sure you have an AI acceptable use policy in place.

This assessment is the inside view. For the outside view, the free Bonofici grader reads your company’s public pages and shows where to start.

Start with what your team already knows.

Book a 30‑minute call.We come back with the first three tools worth building, in writing.

Not ready for a call?

See where your company stands and which jobs are worth the most, from your public pages.