AI Potential Assessment: A Four-Week Approach

An AI potential assessment can be completed in four weeks and produces a prioritized list of use cases with a business case, effort estimate and implementation sequence. The time frame forces selection: the team scores a small number of candidates against real workflows and puts a decision in front of the executive team at the end. This article describes the four weeks in detail, sets out the decision criteria for each step and shows how to recognize a robust assessment.
The assessment touches on questions of data protection and co-determination. This article provides an overview and does not constitute legal advice.
What does an AI potential assessment deliver?
An AI potential assessment delivers a decision paper with which the executive team decides on the first AI use case and sets its budget. To get there, it answers three questions that are often blurred in practice:
- Which use cases create the greatest measurable benefit for the company?
- Which of them can be implemented with existing data and systems?
- In what order should they be introduced?
The decision paper contains a long list of every idea examined and a scored shortlist. Each use case on the shortlist receives a business case and a named owner in the business unit, and the paper as a whole sets out an implementation sequence with dependencies. A list of ideas alone is a collection of possibilities and gives the executive team nothing to decide on. The scoring logic follows the approach in Prioritizing AI use cases: value, feasibility, risk, and the economics follow The AI business case: calculating benefits and costs robustly. How the assessment fits into the overall steering of AI in the company is set out in the article AI management: what executive leadership must steer.
The potential assessment differs from an AI readiness check in what it examines. The readiness check reviews whether the company as a whole meets the prerequisites for AI. The potential assessment reviews individual use cases for benefit and feasibility. Where a readiness check already exists, it shortens the feasibility review in week 2. Where it is missing, the potential assessment establishes the prerequisites only for the workflows on the shortlist. The readiness check is covered in detail in AI readiness: technology, data, organization.
Week 1: Understand
The first week is devoted to operations. The team talks to every business unit head involved and to the employees who carry out a workflow every day, because they hold the information that appears on no organization chart. For each unit the team uses value stream mapping from lean management to produce a map of the main value streams: where waiting times arise, where the same information is captured more than once, and where people decide on the basis of rules that can be written down.
In parallel the team collects baseline figures: case volumes and handling times, as well as error rates and lead times for each workflow. This baseline determines how robust the later results are. Without a starting value no benefit can be claimed, and without a benefit there is no business case. By the end of the week there is a shared view of the workflows that can be presented to management in an hour.
Limiting the scope to two to four business units is the precondition for the schedule. With more units, one week is not enough for interviews, workflow maps and a complete baseline. The company then runs the assessment in waves of two to four units each, starting with the units whose workflows have the highest case volumes.
Week 2: Score
In the second week the long list grows out of the conversations. Each use case is described in a standard profile that records the workflow concerned, the data required, the expected benefit and the owner in the business unit. The team then scores each use case against four criteria:
- Value: which saving, revenue effect or reduction in errors and risk follows from the baseline?
- Feasibility: is the data available in usable quality, and can the use case be connected to existing systems?
- Risk: which data is processed, and which data protection and regulatory requirements need to be clarified?
- Acceptance: does the business unit want to change the workflow, and is there an owner who is permitted to change it?
The fourth criterion is left out of many assessments. It explains why technically successful applications remain unused in operations. Legal questions are only identified and flagged in this week; assessing them case by case is a task for the qualified specialists. At the end of the week a shortlist of three to five use cases remains, and these go into depth. More candidates cannot be validated with a business case and a trial in week 3.
Weeks 3 and 4: Validate and decide
The third week tests the shortlist against reality. For each use case the team builds a business case that derives the benefit from the baseline and sets against it the one-off cost of development and the running cost of operation. To this is added the effort of introduction in the business unit, which many calculations overlook. The team also runs a trial with real cases from the workflow. Within a few days it shows whether the quality of the output is sufficient in principle. The proof for live operation then comes from the prototype with a complete test catalog.
The fourth week places the validated use cases in a sequence. Dependencies decide: a use case that provides data which others need moves forward, even if its own benefit is smaller. The sequence is condensed into a roadmap of the kind described in AI roadmap: the first twelve months. The team presents the decision paper to the executive team with a recommendation on budget and owners, and on the start date of the first use case.
Can a company run the assessment itself?
A company can run the potential assessment itself if it releases a team with experience in workflow analysis and business case modeling for four weeks. The bottleneck is rarely the method; it lies in time and in the team's independence. An internal team tends to score proposals from its own units with more regard for their sponsors, and it lacks the comparison with use cases already running in other companies.
The following checklist serves as a template. If the team can answer every question with yes, the assessment is robust, whether it was produced internally or with external support.
A survey by the German Federal Statistical Office shows that in 2024 only 20 percent of companies with 10 or more employees used AI, so for most companies the first application is still ahead. As a rule of thumb for choosing the route: if the company has already brought at least one AI use case into operation on its own, it runs the assessment with its own team. If the first use case is still ahead, external support saves the learning curve and provides the comparison with use cases in operation at other companies.
Niehues Consulting as a partner for the assessment
Niehues Consulting works in AI management from a delivery perspective. Marin Niehues builds organizations that deliver and brings AI into operations, in projects from VC-funded scale-ups to DAX 40 groups and in assignments in 20+ countries on six continents. The assessment is therefore geared towards implementation from the first day. AI management covers business analysis, requirements and the business case, as well as the selection and sequencing of use cases. It also includes steering the AI portfolio with key figures.
The approach follows the Niehues Delivery Model, whose first phase is called Understand. In this phase its review questions are answered for every use case: which business outcome should change measurably? What is the baseline, and how will success be measured? This is followed by a prototype that proves the benefit in the real workflow with real data. Niehues Consulting then supports the transfer into operations with internal or external teams, and the rollout with adapted workflows and enabled users. The entry point is a free initial consultation of 30 minutes, in which the starting position and the next steps are discussed.
Frequently asked questions
What is an AI potential assessment?
An AI potential assessment is a structured study that establishes in which workflows of a company AI creates a measurable benefit. It ends with a scored list of use cases, a business case for each and a recommended implementation sequence.
How long does an AI potential assessment take?
Four weeks are enough for a focused assessment of two to four business units with a shortlist of three to five use cases. With more units the assessment runs in waves, each on its own four-week cycle. Fixing the scope early keeps the time frame.
How much does an AI potential assessment cost?
The cost follows from the number of consultant days and the day rate. In the German market, external AI consulting is typically billed at around 1,000 to 2,500 euros per consultant day, depending on seniority. On top of this comes the time of the company's own staff for interviews, data provision and scoring, which should appear as a separate line in the plan.
Is there public funding for an AI potential assessment?
Federal and state programs in Germany subsidise consulting for small and medium-sized enterprises, at federal level for instance through the Federal Office for Economic Affairs and Export Control (BAFA). The federal funding database shows which programs are open and which conditions apply. The application usually has to be filed before the consulting starts, so the check belongs at the beginning of planning.
What is the difference between an AI potential assessment and an AI readiness check?
The readiness check reviews the company's prerequisites in technology, data and organization, while the potential assessment reviews individual use cases for benefit and feasibility. The two complement each other: an existing readiness check shortens the feasibility review within the potential assessment.
Can we carry out an AI potential assessment ourselves?
Yes, with a dedicated team that is skilled in workflow analysis and business case modeling. The checklist in this article serves as a template. If the company has not yet brought an AI use case into operation, external support shortens the learning curve.
Who needs to take part in the assessment?
The participants are the executive team as sponsor, the heads of the business units concerned as owners of the use cases, and a contact from IT for data and systems. Employees who carry out the workflows under review every day also take part. For questions of data protection or co-determination, the qualified specialists should be involved early.
What follows the potential assessment?
After the decision comes a prototype for the first use case in the sequence, which tests the benefit in the real workflow. If the test is passed, the use case is transferred into operations and introduced in the business unit. The baseline figures serve as the yardstick.
Let’s talk about your initiative.
In 30 minutes we clarify where your organization stands and which step will pay off most now.
Request a consultation30 minutes · free · no obligation



