It scores each process
Volume, standardisation (rules), data and error tolerance each score from 0 to 3. The total (maximum 12) is converted into a score from 0 to 100.
Tool · applied AI
Test: which process should I automate first?
Describe up to 3 of your company processes by choosing from menus (volume, people, rules, data, error tolerance and systems involved) and get an automatability ranking with an indicative solution type for each. No savings percentages, no ROI: it is a methodological prioritisation, not a promise.
In short: You describe up to 3 of your company processes —A is required, B and C optional— choosing for each one the monthly volume, the people involved, whether there are clear rules or expert judgement is needed, whether the data is structured or lives in documents and conversations, your error tolerance with human review, and how many systems it touches. The tool scores each process from 0 to 100 with a matrix (volume × standardisation × data × tolerance), ranks them from most to least automatable, and proposes an indicative solution type for each: RPA/workflow, document extraction, RAG agent or human copilot. There are no savings percentages and no ROI: it is a methodological prioritisation.
This tool is indicative and does not constitute professional advice. The result depends entirely on how you describe each process: it does not audit your real systems or their exceptions, and it does not replace a diagnosis. Check your situation with AI strategy advisory or with a professional before deciding what to automate.
How the calculation works.
Methodology · formulas in plain sightIt ranks by score
Processes are ranked from highest to lowest score: the first on the list is, by these criteria, the candidate to automate first.
It assigns a solution type
Clear rules + structured data → RPA/workflow. Data in documents → document extraction. Expert judgement with low tolerance → human copilot; in the remaining cases with expert judgement, conversational data or mixed rules → RAG agent.
Coordination and low-automatability warnings
More than 10 people or more than 3 disparate systems do not subtract points, but they are flagged as a coordination warning. A score below 25 is flagged as a candidate for individual analysis before automating anything.
Sources
- N. Yadav and S. P. Panda — "Developing standard criteria for robotic process automation candidate process selection", IAES International Journal of Artificial Intelligence (IJ-AI), vol. 13, no. 4, pp. 4291-4300 (Dec. 2024), DOI 10.11591/ijai.v13.i4.pp4291-4300: identifies repetitiveness (REP), rules (RULE) and volume (VOL) —together with human intervention (HI) and duration (DUR)— as the highest-scoring criteria out of a set of 20 attributes selected by PCA, which also include structured data (STRUC) and error probability (ErrorP) — ijai.iaescore.com (accessed 16-07-2026). It is the basis for the volume/rules/data/tolerance factors in the scoring matrix.
- Sevde K. (IBTech) — "How to Choose the Right Processes for Robotic Process Automation (RPA)" (21-05-2024): "RPA works best with processes that follow clear, rule-based procedures"; "Processes that involve handling and processing large amounts of structured data are ideal for RPA"; "Processes with a low exception rate are easier to automate" — medium.com/ibtech (accessed 16-07-2026). It is the basis of the "clear rules + structured data → RPA/workflow" criterion.
- McKinsey & Company — "One year of agentic AI: Six lessons from the people doing the work": "If the task is rule based and repetitive, with structured input... use rule-based automation. If the input is unstructured... but the task is still extractive or generative, use gen AI... If the task involves multistep decision-making and has a long tail of highly variable inputs and contexts, use AI agents" — mckinsey.com (accessed 16-07-2026 via live web search; the direct fetch of the article was blocked by timeout, and the quotation was verified independently in two separate queries). It is the basis of the "data in documents → extraction" and "expert judgement/conversational data → RAG agent" criteria.
- McKinsey & Company — "When can AI make good decisions? The rise of AI corporate citizens": "Low-risk, low-complexity decisions... are prime for full automation... high-risk, high-judgment scenarios... may still require human oversight, supported by AI copilots" — mckinsey.com (accessed 16-07-2026 via live web search; the direct fetch of the article was blocked by timeout, and the quotation was verified independently in two separate queries). It is the basis of the "expert judgement + low error tolerance → human copilot" criterion.
Frequently asked questions about this test.
How is each process scored?
With a 4-factor matrix that you describe from menus: monthly volume, standardisation (clear rules, mixed or expert judgement), type of data (structured, documents or free conversation) and error tolerance with human review. Each factor scores from 0 to 3 and the total is converted into a score from 0 to 100. Nothing is preloaded: the score depends entirely on what you select.
What does the recommended solution type mean?
It is methodological guidance, not a quote: RPA/workflow when the rules are clear and the data is already structured; document extraction when the data lives in invoices, contracts or PDFs; RAG agent when you need to search scattered knowledge or there is expert judgement with room for review; human copilot when the risk of error is high and the decision must remain with a person.
Why do the number of people and systems involved not change the score?
Because the literature reviewed gives no standard weighting for those two factors within the automatability score: they are shown as a coordination warning (more people and more disparate systems mean more implementation effort) but they neither add nor subtract points, so as not to feign a precision the source does not support.
What if a process gets a very low score?
The tool flags it as low automatability and recommends treating it as a candidate for individual analysis before automating anything: it may need redesign, more standardisation, or simply not be a priority yet.
Can I compare fewer than 3 processes?
Yes. Only Process A is required; processes B and C are optional. Once you start filling in one of them, you must complete all 6 fields of that process for the ranking to be correct.
Does it replace an automation diagnosis done by a consultant?
No. It is indicative methodological prioritisation to put several candidate processes in order. A real diagnosis reviews the whole process, its exceptions, the specific systems and technical feasibility — things a 6-question test per process cannot see.