After 25+ DACH enterprise engagements, we can predict with reasonable accuracy whether an AI initiative will reach production — before a single line of code is written. Not because we have a crystal ball, but because the same six factors determine success or failure with remarkable consistency.
These six dimensions form the diagnostic core of the AI Operating System methodology. Each one represents an area where we have seen initiatives stall, fail, or succeed. Score poorly on two or more and production deployment becomes statistically unlikely. Score well across all six and the question shifts from "will this work?" to "how fast can we deploy?"
1. Workflow Readiness
The question: Can you articulate, in measurable terms, which process you want AI to enhance?
This is not "we want to use AI in customer service." It is: "Our support team processes 800 tickets per week. 60% follow predictable patterns. We want to classify incoming tickets by urgency, route to the correct department, and draft initial responses for the predictable 60%."
The first statement is a wish. The second is a workflow.
What we assess:
- Is the target workflow documented with clear inputs and outputs?
- Can throughput be measured in units per period?
- Is there a definition of "good output" that a model can be evaluated against?
- What is the current error rate, cycle time, and cost per unit?
Red flags: no process documentation, no measurable throughput, no clear definition of success. These mean you need Discovery before you need a build. For how workflow readiness connects to production success, see From AI Pilot to Production.
2. Data Accessibility
The question: Can you get the data from where it lives to where a model needs it — in weeks, not months?
Not data quality. Data accessibility. Every organisation has messy data. That is not the question. The question is whether there is a viable path to make relevant data available to an AI workflow without a multi-month infrastructure project.
What we assess:
- Where does the relevant data live? (SAP, Dynamics, Salesforce, Excel, shared drives)
- Are there APIs, export functions, or database access available?
- Can a staging environment be provisioned without touching production systems?
- How long would IT estimate to deliver a functional data feed?
Red flags: data locked in legacy systems with no API access, IT estimates exceeding 8 weeks for data delivery, no sandbox or staging environment. This dimension kills more Mittelstand initiatives than any other — silently, after the project has started.
If legacy systems are the blocker, AI-native modernisation can run in parallel with AI workflow delivery — stabilising the stack without a big-bang cutover.
3. Decision Authority
The question: Who can say "yes, deploy to production" — and how long does it take?
In our experience, this is the single strongest predictor of project success. A named executive with budget authority and operational mandate — who can make go/no-go decisions in days — is worth more than any technical capability.
What we assess:
- Is there a named exec sponsor?
- Does this person have pre-approved budget (typically €30–150K for a first initiative)?
- Can they approve production deployment without committee escalation?
- Are they operationally involved or purely strategic?
Red flags: decision authority distributed across a committee, budget not pre-approved, multi-month approval cycles. In the Mittelstand, the Geschäftsführer as direct sponsor is often the strongest configuration — short decision chains beat sophisticated governance every time.
4. Compliance Posture
The question: Is your organisation's stance toward AI regulation enabling or blocking?
With the EU AI Act in force and DSGVO applying to any AI system processing personal data, compliance is non-negotiable. The question is not whether compliance matters — it does — but whether the organisation treats it as a guard rail or a roadblock.
What we assess:
- Has legal/compliance reviewed AI use cases and provided clear guidance?
- Is there a framework for classifying AI applications by risk level (per EU AI Act)?
- Can compliance review happen in parallel with development, or is it a sequential gate?
- Does the organisation understand what DSGVO requires for AI systems specifically?
Red flags: no compliance review framework, "we need to fully understand all regulations before starting," legal treating every AI use case as high-risk by default. For practical guidance, see our EU AI Act resource centre.
The productive posture: compliance provides boundaries, and the team builds within them. The unproductive posture: compliance must approve everything before anything starts.
5. Team Capacity
The question: Do you have people with time and mandate to work on this?
This is not about AI talent. For Level 1 deployments, you need domain experts who understand the workflow, a technical lead for integration management, and access to external engineering capacity for the build. You do not need data scientists on staff.
What we assess:
- Can the business team assign 1–2 domain experts for 20–30% of their time?
- Is there a technical counterpart (internal or external) for integration work?
- Does the IT team have bandwidth to support data access and infrastructure?
- Are the assigned people actually available, or are they allocated to three other priorities?
Red flags: all named resources are already at 100% capacity, no technical counterpart identified, IT backlog exceeds 6 months. A fully staffed team with no available bandwidth is, functionally, zero capacity.
6. Operating Model Clarity
The question: Do you know how AI will change who does what?
This is the dimension most organisations skip — and the primary reason Level 2 deployments fail. Deploying AI without redefining roles, responsibilities, and success metrics creates confusion, resistance, and shadow processes.
What we assess:
- Has the organisation defined which tasks will shift from human to AI?
- Are there new roles or modified role descriptions for the post-deployment state?
- Have success metrics been updated to reflect the new operating model?
- Is there a change management plan for the affected team?
Red flags: no discussion of role changes, assumption that "people will just use the tool," no updated KPIs. If you deploy AI into a team without telling them how their work changes, expect the AI to be unused within 60 days.
How the dimensions interact
The six dimensions are not independent. They create dependencies:
- Workflow Readiness enables everything. Without it, you cannot define data requirements, compliance scope, or success metrics.
- Data Accessibility and Decision Authority are the two most common blockers. Fix these first.
- Compliance Posture and Team Capacity determine execution speed.
- Operating Model Clarity determines whether deployment sticks.
An organisation with strong workflow readiness and decision authority but weak data accessibility needs a different intervention than one with accessible data but no exec sponsor. The dimensions do not just diagnose — they prescribe.
Scoring your organisation
In The AI Operating System, each dimension is scored on a four-point scale: blocking, weak, adequate, strong. The pattern across all six determines the recommended next step:
- Multiple blocking scores: Discovery engagement to establish fundamentals
- Mixed weak/adequate: Targeted intervention on blocking dimensions, then Accelerator
- Mostly adequate/strong: Ready for Accelerator — proceed to Level 1 deployment
Use the diagnostic for a structured self-assessment across all six dimensions.