Does My Company Need AI? A 10-Question Test (2026)
Does your company need AI? Instead of guessing, take a 10-question yes/no test in three blocks: process, numbers, organization. The score tells you plainly whether AI pays off now, which process is a candidate, and when the honest answer is not yet. You start with a free process scan.
Whether your company needs AI depends on one repeatable process, its numbers, and who will maintain it, not on whether AI is fashionable. Instead of guessing, take the test below: 10 yes/no questions in three blocks (process, numbers, organization). Count your yeses. The score tells you plainly whether it is worth doing now, which process is a candidate, and when the honest answer is not yet.
How this test works
This is not a personality quiz, it is a way to count one process before you spend anything. Pick one specific process (email handling, invoices, reporting, quoting) and answer all 10 questions only for that one. Each yes is one point. At the end you add them up and read the score from the table.
The three blocks are not arbitrary. Process checks whether there is anything to automate at all. Numbers check whether it pays off. Organization checks whether the implementation survives launch. The third block is the most overlooked and the most often decisive in failure.
Block 1: PROCESS (is there anything to automate)
1. Do you have a repeatable process that eats hours every week? Yes only if you can name it as a single unit of work: "triaging and first-replying to email," "entering invoices into a system," "assembling the weekly report." "We do lots of different things" is a no.
2. Can the process rules be written down? Yes if you can describe the steps and exceptions on one page, or they are already in a procedure. If decisions depend on one person's judgment and come out different every time, it is a no.
3. Does the volume add up? Yes if the process happens often: tens or hundreds of times a month, not once a quarter. Automating a rare event rarely pays back, because the build cost stays the same while the saving is tiny.
Block 2: NUMBERS (does it pay off)
4. Do you know the hours times rate of this process? Yes if you can state the annual cost of the manual work from this one formula:
Annual manual process cost =
hours per week on this process
x hourly rate of the people doing it
x 52If you cannot fill in the numbers, that itself is a signal: you are not measuring the process, so you will not measure the effect either.
5. Does the annual cost of manual work exceed the order of an implementation? Automating one process starts from €3,500 net (typically €3,500–9,000 net), and an agent that runs a whole process from €6,000 net. Yes if the annual cost from question 4 is clearly higher than that order of magnitude plus maintenance. If it is lower, the score is honest: manual handling is cheaper.
6. Does the data the process runs on live in systems, not in people's heads? Yes if the information you need sits in a CRM, ERP, inbox, or files you can reach. If the key knowledge lives in one person's memory and is written down nowhere, it is a no, and AI will not fix that.
Block 3: ORGANIZATION (does the implementation survive launch)
7. Is there a process owner? Yes if there is a specific person accountable for keeping the process working, and it is they who will talk to the vendor. "The process belongs to nobody" is a no, and the most common reason an implementation dies after three months.
8. Can you name a success metric? Yes if you can finish the sentence "this worked if \_\_\_" (handling time dropped from X to Y, the queue stays under Z, fewer than W errors). Without a metric you cannot tell success from failure, so every result will be "well, it sort of works."
9. Will the team use the output? Yes if the people whose work will change know about the project and see a benefit in it for themselves. A system the team routes around, because it does not trust or understand it, is a cost, not a saving.
10. Will someone maintain the system after launch? Yes if it is clear who handles changes, monitoring, and errors when the process shifts (and it will). AI is not a piece of furniture you buy once. It is a system that lives alongside the company.
What your score means
Count the yeses and read the row. The score is honest both ways: a low one does not mean "you lost," it means "for now you rightly stay as you are."
| Test score | What it means | The right next step |
|---|---|---|
| 0–3 yeses | You do not need AI yet. The process is too small, too irregular, or too disorderly for an implementation to pay back. | Tidy the process: write down the rules, name the exceptions, start measuring the hours. This is good news, not a failure: you spare yourself a costly project that would not have survived. |
| 4–7 yeses | One process genuinely qualifies. You have something to automate and it provisionally looks worthwhile, but it needs a precise count. | Start with a free process scan: we count that one process on your numbers and tell you whether it is worth it, before we build anything. |
| 8–10 yeses | There is more than one candidate and the company is organizationally ready. The risk is no longer "whether," only order and priorities. | Start with a scan, and with several processes consider an AI process audit: a process map with priorities, so you start with the one that pays back fastest. |
If you are stuck between two thresholds, look at the block with the most noes. None in PROCESS sends you back to tidying up. None in NUMBERS means measure first. None in ORGANIZATION is the most dangerous, and that is the next section.
Why organization predicts failure better than technology
Because the model is not what decides whether an implementation survives. Gartner (January 2026) reports that by the end of 2025 at least half of generative AI projects were abandoned after the proof-of-concept stage, above the same firm's earlier forecast (30%, from July 2024). The model itself is rarely to blame. Far more often the ORGANIZATION block breaks: no process owner, no success metric, a team that will not use the output, or no one to maintain the system after launch.
That is why the four questions in the third block weigh more in practice than the three in the first. You can have a perfect process and good numbers and still land in that abandoned half if the project belongs to nobody. Your no in the ORGANIZATION block is nothing to be ashamed of. It is an early warning that just saved you money: you fix it with a conversation, not another tool.
"Whether" is not yet "which"
This test answers whether to move into AI at all. It does not tell you which process should go first. That is a separate decision, made on different criteria: where the saving is largest, where the risk is lowest, and what can be shipped quickly. If the test came out at 4 points or more, the next step is choosing one specific process, not building everything at once.
How to set the order and spot the best candidate we break down in what to automate in your company and which processes fit first. And to understand why so many implementations fail, we gathered the causes in why AI projects fail.
When a no is a good answer
Honestly: there are situations where a low score is not a failed test but an accurate diagnosis. A no protects you from that abandoned half.
- The process happens rarely and irregularly. The build cost stays the same whether the process runs 5 or 500 times a month. At low volume, manual handling is simply cheaper.
- The rules change every week and live in people's heads. Write the process down first. That is 80% of the work before AI enters the picture, and often the problem disappears once the process is ordered.
- There is no owner and no one to maintain it. A project without a steward dies after launch, however technically perfect. That is an organizational decision, not a purchasing one.
- The annual cost of manual work is lower than the cost of building it. Then we will advise against building. Knowing what the process really costs is useful anyway, with any budget and any vendor.
If any of these fits your situation, we will say so plainly on the scan, before you spend anything.
FAQ
Does my company need AI?
It depends on one repeatable process, its numbers, and who will maintain it, not on whether AI is fashionable. Take a 10-question yes/no test in three blocks: process, numbers, organization. Scoring 0–3 yes means you do not need AI yet, only a tidier process. 4–7 means one candidate worth starting with. 8–10 means several candidates, and an audit decides the order.
How do I check whether AI will pay off?
Calculate the annual cost of the manual work: hours per week times hourly rate times 52. If it is clearly higher than the roughly €3,500 net order of an implementation plus maintenance, the process is a candidate. If it is lower, AI will not pay off and manual handling is cheaper. That is your substitution, not our promise.
Which processes are suitable for AI?
Repeatable processes with rules you can write down and volume that adds up. The data must live in systems, not in someone's head. There must be a process owner and someone to maintain the system after launch. Which of your processes goes first is settled by a separate what-to-automate test.
When should I NOT deploy AI?
When the process happens rarely and irregularly, when the rules change weekly and live in people's heads, when nobody owns the process or will maintain the system after launch, or when the annual cost of manual work is lower than the cost of building it. Then a no protects you from joining the half of projects that Gartner (January 2026) reports were abandoned after the pilot.
How to start
The cheapest sensible first step is to count one process, not to buy a tool.
- Book a free process scan and bring your test score to the call.
- Prepare for the chosen process: who does it, how many times a month, how long one case takes, which systems are in the path, who will own it, and how you will recognize success.
- After the call you get a recommendation: automation, an agent, an AI process audit if there are several candidates, or an honest "not worth it yet."
Book a free process scan | See pricing | AI process audit