Digital Intelligence Resilience Index

Access to AI models is becoming a commodity. What lasts is the layer your organisation controls: evidence, policies, measurable targets, limits, documentation.

Access to AI models is becoming cheap and common. Your competitor rents the same tools. The question that matters is not which model to pick; it is what remains yours when the model, the provider, the team or the price changes.

18 questions, about 4 minutes

No sign-up, no email. The result appears on your screen and is stored nowhere: every calculation happens in your browser, and no answer is sent to a server.

Ownership and portability
Evidence and provenance
Measurability of outcomes
Independent verification
Limits of authority and shutdown
Data sovereignty
Resilience to changes of people

0 of 18 answered

Ownership and portability

Question 01

If your provider doubled its price tomorrow, how long would it take you to move elsewhere?

Days — our data and settings are in an open format that we hold Weeks: awkward, but feasible Months — much of it is tied to the provider The question has never been raised

Question 02

Can you export data and settings in a format that is readable without that particular provider?

Yes, and we have confirmed it in practice In theory yes; we have not tried it Only part of it No, or we do not know

Question 03

Have you ever tried to rebuild the system in a different environment?

Yes, as a regular exercise Once, and it worked We only discussed it Never

Evidence and provenance

Question 04

For a piece of data that feeds an AI tool, do you know its source, its date and who owns it?

Yes, recorded systematically For the important ones, yes Only patchily No

Question 05

Is there a record of who changed what, and when, in the data or the settings?

Yes, a full history with the ability to roll back Partial logging Only in some systems No

Question 06

Do you know which of your data may be used with AI (licences, consent, confidentiality)?

Yes, there is an explicit classification In broad terms Each person decides case by case It has not been examined

Measurability

Question 07

Before an AI project starts, do you set a measurable target and a baseline?

Always, with figures and a time horizon Usually, in qualitative terms Rarely No

Question 08

Can you say in numbers what the organisation gained from using AI over the past year?

Yes, with specific figures Approximately We assume it helped No

Independent verification

Question 09

Who decides whether an AI project succeeded?

A criterion set before it started, judged by someone independent The internal team, after the fact The supplier itself Nobody: it seems to work

Question 10

Has any AI solution of yours ever been reviewed by an independent third party?

Yes, with a written report Informally Only by the supplier Never

Limits and shutdown

Question 11

Is there an explicit, written limit on what an AI system may do automatically without a human?

Yes, documented and observed There is an informal understanding We judge it case by case No

Question 12

Is there a procedure for stopping it immediately if something goes wrong?

Yes, and it has been tested Yes, but untested We would handle it ad hoc There is none

Question 13

If an automated system caused damage, is there a recovery plan?

Yes, with backups and a tested restore Backups exist; the restore has never been tested Partly No

Data sovereignty

Question 14

The confidential data you hand to AI tools: where is it processed?

We know exactly; it is recorded and deliberately chosen We roughly know No, but we assume it is fine We had not considered that it leaves the organisation

Question 15

Have you identified which data is not allowed to leave for an external service?

Yes, with a clear classification that is observed There is a general instruction We leave it to individual judgement No

Question 16

Is there a data processing agreement with every AI provider you use?

Yes, with all of them With the main ones With some No, or we do not know

Resilience to change

Question 17

If the person who knows your systems best left tomorrow, what would remain?

Documentation that lets someone else carry on Partial documentation; it would take time Very little in writing The knowledge is in their head

Question 18

Are the rules and decisions about how you work with AI written down?

Yes, in a living document that is kept current In some old document Verbally They have never been articulated

Please answer all 18 questions.

Resilient

Your organisation keeps real control. The tools can change without the accumulated knowledge being lost. From here on the task is maintenance: regular export and restore drills, and keeping the limits current as automation grows.

Controlled

The foundations are solid, but control rests on practices that have not all been tested. The difference between “we have backups” and “we have tested a restore” is usually where the risk hides.

Dependent

Operations deliver, but much of the value sits in systems and people you do not fully control. A change of pricing, of terms of use or of one person would cost you disproportionately.

Rented

What you use today your competitor can rent tomorrow, and what you have built on top of it does not travel easily. This is not a tooling problem; these are questions that have not been asked yet.

Ownership and portability. Set yourself an exercise: export all data and settings, then try to stand the system up in a clean environment. Whatever does not come out is the part that is not yours in practice.

Evidence and provenance. Start from the data that feeds decisions. For each one: where it came from, when, who owns it, what may be done with it. Without that, no AI output is defensible.

Measurability. Before the next project, write down two numbers: where you are today and where you want to get to. Without a baseline, success becomes a matter of impression.

Independent verification. The supplier cannot also be the judge. Appoint someone, internal or external, who has no stake in the outcome.

Limits and shutdown. Write down what may happen automatically and what may not, and test the stop. A recovery plan that has never been tested is not a plan.

Data sovereignty. Classify: public, internal, confidential. The public can go anywhere; the confidential needs a conscious decision about where it is processed.

Resilience to change. Knowledge that lives only in people leaves with them. A living record of decisions, why we did what we did, is worth more than a manual of steps.

Ownership and portability Evidence and provenance Measurability Independent verification Limits and shutdown Data sovereignty Resilience to change

out of 54 points

By axis

Where it is worth looking first

If you would like to see how these translate into concrete steps for your own organisation, a short conversation is enough to show what deserves to come first, and what does not need doing at all.

Discuss your result

Context and attribution. The structure of this assessment is inspired by the concepts set out in the white paper “Neural-Symbolic SUCCESSOR Ω” by MONTREAL.AI (August 2026), in particular the distinction between rented access to models and durable, client-controlled intelligence.

Statement of independence. IWH has no partnership, representation or endorsement relationship with MONTREAL.AI. The questions, the scoring, the levels and the recommendations are entirely our own work, adapted to the scale of Greek small and medium-sized organisations. No text from the white paper is reproduced.

What this is not. It is not a certification, legal advice or a compliance audit. It is a tool for thinking. A low score does not mean something is wrong; it means some questions have not been asked yet.

No answer and no result leaves your browser: nothing is sent to a server. As on every other page of this site, only an anonymous visit is counted (Umami, without cookies).

Related Resources

Explore Further

Dive deeper with related case studies, white papers, assessments, and articles.

Loading resources...