Organisation & AI · AI Readiness Assessment
I build the position that is missing on the org chart. The entry point is the AI Readiness Assessment.
The entry product
Where does your company stand before the first agent enters a core process? The assessment measures seven dimensions, each on its own: Organization, Data, Technology, Processes, Know-how, Governance and Culture. The backbone is the organization, because that is where it is decided whether someone owns the decision.
The four questions, per core process. Is the decision named. Is there a role that owns the process as a whole. Does its effect show in the reporting. Are limit and escalation machine-readable. A company can be far along on data, technology and governance and at zero on organization. That is the normal case, not an outlier.
Data: availability, quality, access. Technology: systems, interfaces, traceability. Processes: core processes, standardization, automation. Know-how: knowledge, training, AI literacy under Art. 4. Governance: rules, oversight, reporting path, plus AI inventory and pre-sorting by risk class as the basis for the legal review. Culture: openness, change, tolerance for error.
After the assessment
The assessment says where you stand. The modules build the position that is missing. Each is bookable on its own, the order follows the questions.
01
Who carries the damage today, and do their KPIs reflect the decision?
There is no new organisational form for agents. There are known forms that create result responsibility across department boundaries, and each has its price. For companies between €20 and 200 million revenue, a process owner with their own P&L is usually the most viable solution. I choose the model with you and build the KPIs for it, on contribution margin, delivery reliability and capital tied up.
→ A named person whose KPIs reflect the total
02
How does the agent recognise its limit, and what happens then?
Approval limits, escalation path and deviation detection are captured so a machine can read them. A permissions concept governs what the agent may do. It does not govern who carries the consequence if the limit was set wrong. Both belong together. IT gets its clear part in this: traceable, able to be switched off, logged.
→ Limits that exist for the agent, because they are stated in numbers
03
What is left for management when machines decide the exceptions?
What cannot be delegated: setting limits, detecting deviations, carrying the financial damage. I work out the tasks with CEO, COO and CIO by responsibility instead of by department. Resources follow responsibility. Whoever carries the damage gets the budget for the limit.
→ Roles that fit the new decision landscape
04
Is data and compliance ready?
The obligations under the EU AI Act are a tool. Based on the assessment, I build the organisation for them: an AI policy on one to two pages (organisational part), an AI literacy training concept under Art. 4 with records, a requirements catalogue for contracts with AI vendors, a reporting process for serious incidents, documentation of human oversight. The legal review and the contract texts are handled by your legal department or by a law firm I work with.
→ An organisation able to implement the obligations of the regulation, without you reading 100 pages of legal text
I am not a lawyer and do not provide legal services. Legal assessments and contract drafts are prepared by licensed lawyers.
AI Readiness Check
Sixteen questions, about five minutes, seven dimensions, the same as in the assessment. At the end you see where your company stands today and where it is most open.
Dimension 1 of 7
Leadership, ownership, decision rights
Dimension 2 of 7
Availability, quality, access
Dimension 3 of 7
Systems, interfaces, traceability
Dimension 4 of 7
Core processes, standardization, automation
Dimension 5 of 7
Knowledge, training, AI literacy under Art. 4
Dimension 6 of 7
Rules, oversight, reporting path
Dimension 7 of 7
Openness, change, tolerance for error
Your profile
0 / 16 questions answered
Answer all questions for your profile.
I will send you your profile as a short email, no attachment and no newsletter.
Thank you, your profile is on its way.
Sending did not work. Write to me directly at info@vonkoenemann-consulting.de.
That was too fast for the spam protection. Please click send once more and your message will go out.
Request the assessment → Write to me →
A first orientation, not an assessment. The check shows a profile, not a grade. How deep each dimension really holds is what the assessment clarifies.
When the topic comes up
Technically, the agent works. Every department set its own limit, and set it tight, because it alone answers for mistakes. So the agent may only do what the tightest of the five limits allows. That is how the result gap everyone complains about is created.
Whoever does not carry the financial damage cannot be responsible for the limit. IT has a different job, and it is important enough: the agent must be traceable, able to be switched off, and logged.
Sounds clean. An experiment by Wiles and colleagues (Boston University and BCG, 2026) shows what happens: among managers in organisations that already list AI agents on their org chart, the share of errors detected falls by around 17 percent when the same work comes from an "AI employee" instead of a tool. They also feel less responsible. The organisation becomes less certain and slower at the same time.