The board wants AI. The CTO has a proposal. And you, as CFO, are being asked to sign off on a budget that ranges from "let's start small" to "this will transform the company." The problem is that you have seen both claims before — for ERP migrations, for cloud transformations, for digitisation programmes — and the pattern is familiar. Ambitious scope, unclear ROI, cost overruns by month six.
AI does not have to follow that pattern. But it will, unless someone in the room is asking the right financial questions before the first euro is spent.
This checklist is that set of questions.
Why the CFO perspective matters more than you think
Most AI readiness frameworks are built for CTOs. They evaluate technical infrastructure, data maturity, and model capabilities. What they do not evaluate is whether the organisation can absorb the financial and operational cost of an AI initiative without overcommitting resources or triggering compliance risks that nobody budgeted for.
In our experience across 25+ DACH engagements, the initiatives that stall rarely fail on technology. They fail on budget authority — the initiative starts in IT, grows beyond the CTO's signing authority, and then sits in a queue for board approval while momentum dies.
The CFO who gets involved early — not as a gatekeeper but as a financial architect — is the single biggest predictor of whether an initiative reaches production on time.
The realistic budget range for a first AI initiative
Let us be specific. For a Mittelstand company deploying its first production AI workflow — not a proof of concept, not a demo, but a workflow that runs daily and produces measurable business impact — the realistic budget range is €30,000 to €150,000.
That range depends on three variables:
- Workflow complexity. A classification model on structured data (e.g., claims triage, lead scoring) sits at the lower end. A multi-step workflow with unstructured data, human-in-the-loop review, and integration into legacy systems sits at the upper end.
- Integration depth. A standalone tool costs less than one that writes back to your ERP, CRM, or core business system.
- Compliance overhead. Regulated industries add 20–40% in compliance cost — not because the regulation is excessive, but because most companies discover their compliance posture late and retrofit it.
If someone is quoting you less than €30K for a production deployment, they are either cutting corners on integration or selling you a pilot that will never reach production. If someone is quoting more than €150K for a first initiative, they are building infrastructure you do not need yet.
The ROI timeline: what to expect and what to demand
The honest ROI timeline for a well-scoped first AI initiative is 3 to 6 months to measurable impact. Not "we project savings of X over three years" — but "this workflow now runs 40% faster and we can measure it."
Three months is achievable when the workflow is well-defined, data is accessible, and the team has capacity. Six months is realistic when integration complexity or compliance requirements add cycles.
What you should not accept: a 12-month timeline to first measurable results. If the team cannot articulate where value appears within two quarters, the scope is wrong.
What "measurable impact" actually means
Demand specificity. Not "improved efficiency" but:
- Hours saved per week in a named team
- Reduction in error rate for a specific process
- Throughput increase for a defined workflow
- Cost avoidance in a quantifiable category (e.g., external review hours)
If the proposal cannot name the metric, the workflow, and the team — it is not ready for approval. It is ready for Discovery.
The hidden costs nobody puts in the first proposal
Every AI initiative has costs that do not appear in the initial budget request. As CFO, you should ask for them explicitly:
1. Compliance and legal review
A Data Protection Impact Assessment (DPIA) for any workflow touching personal data. Legal review of the EU AI Act classification for your use case. For regulated industries, sector-specific supervisory alignment. Budget €5,000–€20,000 depending on complexity and whether you need external counsel.
2. Integration and data engineering
The model is 20% of the cost. Getting data out of your existing systems, cleaning it, and piping results back is 60%. Most first proposals underestimate this by half.
3. Ongoing operations
Models need monitoring, retraining, and maintenance. A production AI workflow is not a one-time build — it is an ongoing operational cost. Budget 15–25% of initial build cost per year for operations.
4. Change management
The team that currently runs the manual version of this workflow needs training, communication, and transition time. Ignore this and you get passive resistance that silently kills adoption.
The 6 questions every CFO should ask
Before signing off on any AI initiative, run these six questions. If the answers are vague, the initiative is not ready for budget — it is ready for scoping.
Question 1: What is the named workflow?
Not "customer service" but "inbound email classification and routing for the claims team." If they cannot name it precisely, they have not scoped it.
Question 2: What is the total cost including integration, compliance, and 12-month operations?
Reject any budget that only covers the model build. Demand the full picture.
Question 3: Where does measurable ROI appear and when?
Accept months, not years. Accept specific metrics, not directional improvements.
Question 4: Who is the executive sponsor with operational authority?
An AI initiative without a named sponsor who controls the workflow and the team is an orphan. Orphans do not reach production.
Question 5: What is the compliance posture for this workflow?
Does it touch personal data? What is the EU AI Act risk classification? Has a DPIA been scoped? If the answer is "we will figure that out later," add 30% to the timeline and budget.
Question 6: What happens if we stop after phase one?
Every initiative should be scoped so that phase one delivers standalone value. If the business case only works after three phases, the risk profile is wrong for a first initiative.
How the AI Operating System framework handles the CFO perspective
In The AI Operating System, we built the financial dimension into the readiness framework from the start — because we learned that the CFO's sign-off is where most Mittelstand AI initiatives either accelerate or die.
The framework evaluates budget authority alongside technical readiness, scoping every initiative so that phase one fits within the CFO's direct approval authority. No board escalation required for the first workflow. No multi-year commitment before the first result.
This is deliberate. The fastest path to AI value in a Mittelstand company is not a transformation programme. It is a single workflow, scoped to deliver results within one quarter, budgeted within existing authority, and measured on metrics the CFO already tracks.
What to do next
If your organisation is evaluating AI initiatives and the financial questions remain unanswered, start there. Not with technology selection. Not with vendor evaluation. With financial scoping.
Our Fit Call is a 20-minute conversation to determine whether your organisation's first AI initiative is financially and operationally scoped for success. No pitch. No proposal. Just the six questions above, applied to your specific situation.