Many companies are already using artificial intelligence in their services. AI answers inquiries, supports employees, analyzes information, or performs individual tasks independently. What is often more difficult to answer is: To what extent is the organization actually able to manage these AI-powered services reliably and responsibly?
That's exactly what HUCAISM—short for Human-Centered AI Service Management, a dedicated maturity model. It helps organizations assess their current status and identify where AI service management is already working and where gaps still exist.
The point is explicitly not to reach the highest level as quickly as possible. Rather, the HUCAISM Maturity Model is intended as a tool for honestly assessing one’s current status. What matters is what is actually practiced in day-to-day operations, not what is outlined in guidelines or strategy documents.
Why Organizations Need an AI Service Management Maturity Model
A single AI service can be examined in detail. However, companies also need to know how well-positioned they are overall.
Perhaps there are already clearly defined responsibilities, but no standardized oversight of AI vendors yet. Perhaps AI is well monitored from a technical standpoint, while its impact on people is scarcely measured. Or perhaps there is already an AI policy in place, but no one is checking whether it is actually being followed in day-to-day operations.
A maturity model highlights precisely these kinds of differences.
HUCAISM therefore does not view maturity as a single metric for the entire organization. The assessment is conducted on a domain-by-domain basis. The result is a maturity profile that may well be uneven. This is precisely where its value lies: organizations can identify the areas in which they are already well-developed and those where action is needed.
Thus, the HUCAISM maturity level answers not only the question:
"How far have we come?"
But above all:
"Where do we need to improve next?"
The HUCAISM Maturity Model consists of five levels
HUCAISM describes a development path consisting of five sequential stages of maturity:
- Ad hoc
- Consciously
- Controlled
- Integrated
- Optimizing
The levels are cumulative. This means that an organization first develops the fundamentals before moving on to the next level. What matters here are not statements of intent, but tangible evidence and actual behavior within the organization.
The five levels thus do more than just show which structures are in place; they reveal how naturally responsible AI service management is already embedded in the organization.
Level 1: Ad hoc
At the first stage, AI is already being used, but it is not yet managed systematically.
Individual teams are experimenting with AI applications, employees are using new tools, and the first services are emerging. At the same time, there is often no complete overview of where AI is actually being used.
Typical examples of this stage include:
- No comprehensive overview of AI-powered services
- no clearly designated persons in charge
- Inconsistent or missing rules
- The Use of AI as an Initiative by Individual Teams
- Decisions and checks arise based on the situation
According to HUCAISM, a particularly clear sign of this stage is when no one within the organization can fully explain where AI is being used.
Therefore, the first step in the development process is not to immediately establish extensive governance structures. What is needed first is transparency.
Stage 2: Conscious
In the second stage, the organization recognized that AI must be actively managed.
Initial guidelines are being developed, points of contact are being designated, and the AI services in use are increasingly being documented. The topic of AI governance has thus taken hold.
However, there is often still a gap between what is expected and how the business actually operates.
For example, an organization may already have an AI policy. However, if no one checks to see whether it is actually being followed, it remains, for the time being, nothing more than a statement of intent.
This is precisely what characterizes the " Aware" level: The necessary knowledge and initial structures are in place, but have not yet been fully integrated into day-to-day operations.
The transition to the next stage therefore begins when rules become actual routines.
Level 3: Controlled
In the third stage, AI service management becomes a reality.
Responsibilities are not only defined but also put into practice. For AI-powered services, there are defined procedures; appropriate oversight models are carefully selected; and relevant events are integrated into existing service management practices.
An important example from HUCAISM: A technically incorrect AI result is treated as an incident. A person is assigned responsibility, an analysis is conducted, and corrective action is taken.
This highlights a key difference from the previous stages.
AI is no longer viewed as a separate area of innovation. It is becoming an integral part of regular service operations.
At this stage, for example, the focus is on:
- AI Service Owners are responsible for AI-powered services
- human supervision actually works
- Supervisory models should be selected deliberately
- AI-specific incidents and changes are taken into account
- There is a possibility of relapse
- Quality and human impact are observed
This turns governance on paper into governance in everyday life.
Level 4: Integrated
At the fourth level, HUCAISM is no longer limited to individual services. AI service management is integrated with the organization's existing structures.
Governance, service management, operations, and supplier management are all interlinked. Documentation is increasingly becoming a natural part of the processes and no longer needs to be compiled at the last minute for an audit.
The HUCAISM book illustrates this with a particularly vivid example: An audit is a meeting, not a project.
This means that responsible AI service management is now part of normal operations.
This integration is particularly important for organizations. HUCAISM is not intended to create a separate world alongside existing service management. Rather, existing structures are expanded to meet the requirements of AI-powered services.
This avoids the creation of an additional parallel bureaucracy and instead establishes a shared operating model.
Level 5: Optimizing
In the fifth stage, the organization uses its experience and measurement data to continuously improve its AI-powered services.
It’s not just about making processes faster or more cost-effective. HUCAISM also explicitly focuses on the human value added.
Organizations therefore look beyond technical performance and efficiency. They also examine whether the AI service actually produces better results for people.
Lessons learned from operations are incorporated back into standards and design. This allows new AI capabilities to be adopted more quickly and in a more controlled manner, because the necessary responsibilities, practices, and governance structures are already in place.
Optimization, therefore, does not simply mean more automation.
It means further developing the entire AI service management system based on actual experience.
The HUCAISM maturity score is not an overall score
An important feature of the model is that organizations are not simply assigned a single maturity score.
HUCAISM evaluates each domain separately. For example, a company may already be well-positioned in the areas of technology and data, while its approach to people and roles—or the measurement of actual value added—may be less developed.
This is not a flaw in the model; it is intentional.
An uneven maturity profile highlights where development is particularly needed. HUCAISM thus also warns against a common misconception: Technical maturity does not automatically equate to organizational or human maturity.
To assess their current situation, companies should therefore consider various perspectives:
- Have roles and responsibilities been clarified?
- Do the adapted service management practices work?
- Is governance embedded in the organization?
- Is actual human value added being measured?
- Are technology and data managed reliably?
- Are fairness and bias taken into account?
- Does the collaboration between humans and AI work?
- In the case of agent-based systems: Is their scope for action also controlled?
This provides a much more meaningful picture than a single maturity rating.
The highest level of maturity is not necessarily the right goal
Maturity models can quickly give the impression that every organization must reach Level 5.
HUCAISM explicitly does not follow this approach.
The appropriate level of maturity also depends on the risk. An internal AI service with minor, easily reversible impacts does not necessarily require the same level of control as a service whose decisions have significant consequences for people.
The requirements are based, among other things, on three criteria:
- Far-reaching
- Reversibility
- Regulatory Framework
The maturity model is therefore not a competition to reach the highest level. Rather , the right question is: What level of maturity do we need to implement this AI application responsibly?
This makes the model relevant even for smaller organizations. Not every company needs the same structures or the same level of formalization.
From Maturity Level to Concrete Further Development
Knowing where an organization stands is only the first step.
HUCAISM therefore combines an assessment of the current state with a roadmap for further development. The framework explicitly avoids an approach in which all processes, roles, and guidelines are immediately redesigned as part of a large-scale transformation program.
Instead, the approach is to develop the service step by step in a real-world setting.
HUCAISM outlines four guiding principles for this:
Risk First
Organizations do not necessarily start with the simplest AI service, but rather with a service that is particularly relevant or high-risk. That is where the need for action is evident and the benefits of improved management are especially clear.
Show value
Early, visible improvements help build support for further development.
Build on top of existing structures instead of building in parallel
HUCAISM expands existing service management. The goal is to avoid creating a second layer of bureaucracy alongside existing processes.
Suction Instead of Pressure
Where regulatory requirements, trust, or a specific added value already call for change, this need should be leveraged to drive development forward.
This ensures that the maturity level is not merely a theoretical assessment, but rather a starting point for concrete improvements.
Five phases guide you forward step by step
Based on these principles, HUCAISM outlines a five-phase development path. This path is not strictly linear. Different areas of an organization may develop at different rates.
1. Determine the location
The first step is to get an honest picture of the current situation.
The organization tracks its use of AI, brings previously less visible applications to light, and determines the level of maturity in the relevant domains.
Next, the first services are selected for development to begin.
A comprehensive overview of the available AI services is therefore already a significant first success.
2. Laying the Foundation
This results in a minimal but viable system of governance.
AI services are linked to responsible parties and risk classifications; initial guidelines are established, and fundamental decisions regarding human oversight are made.
The point here is explicitly not to create the most comprehensive set of rules possible right away.
The foundation should be sturdy and actually work in everyday life.
3. Establish a system of oversight during operations
Now the operating model will be put into practice using the selected services.
Supervision becomes effective, service management practices take AI-specific situations into account, potential relapses are identified, and the impact of the service is monitored.
At this point, documented intent becomes operational reality.
It is precisely this phase that leads toward the " Controlled" maturity level.
4. Integrate into existing structures
What works for individual services is then applied to other areas.
HUCAISM will be integrated with the existing service management system and, where applicable, the AI management system. Supplier management and documentation will also become part of the existing processes.
The goal is not to create new parallel structures.
Rather, AI service management is becoming a standard part of the organization.
5. Continuously optimize
In the final phase, the improvement becomes systematic.
Experiences, key metrics, and insights from operations are incorporated back into the design of the services. In this process, the human value added remains a key benchmark.
Organizations thus evolve from using individual, controlled AI applications to an operational model that allows them to incorporate new AI capabilities in a structured manner.
HUCAISM explicitly describes this process as iterative. Domains evolve differently, and individual phases may overlap.
Why Going Big Is Often the Wrong Approach
When it comes to AI governance in particular, there is a strong temptation to start by developing extensive guidelines, committees, and processes.
HUCAISM warns against doing exactly that.
The book contrasts two different approaches. One organization first attempts to establish a comprehensive set of rules for every conceivable situation. Meanwhile, little changes in day-to-day operations. Another organization starts with a specific service, improves how it is managed, gathers experience, and then applies those insights to other services.
The key takeaway is this: HUCAISM will not be rolled out all at once.
Typical stumbling blocks therefore include:
- trying to solve everything at once
- Implementing tools before the operational model
- to start with the most convenient service rather than the most relevant one
- to create additional parallel bureaucracy
- to measure efficiency exclusively
A step-by-step approach, on the other hand, provides the opportunity to learn from real-world experiences and continuously refine the operating model.
Maturity looks different in every organization
Even the size of a company does not change the fundamental logic of HUCAISM.
A smaller organization usually requires fewer formal structures than an international corporation. Responsibilities can sometimes be handled by the same individuals, and processes can be streamlined.
In larger organizations, this may lead to the creation of specific roles, committees, and more formalized governance structures.
However, the fundamental requirements remain the same. Responsibilities must be clear, AI services must remain controllable, and human oversight must be in place where necessary.
This means that the HUCAISM maturity model is not limited to large companies. Its specific implementation evolves in line with the size, risk, and complexity of the organization.
Questions Companies Should Be Answering Right Now
An initial assessment doesn't have to be complicated. Just a few basic questions can reveal how far AI service management has actually progressed within an organization:
- Can we fully identify where AI is used in our services?
- Does every AI-powered service have a clearly designated person in charge?
- Are our AI guidelines actually being followed in everyday life?
- Does human oversight work in practice, or only on paper?
- Should we treat technically incorrect AI results as a service issue?
- Is it necessary to provide documentation during normal operations?
- Are AI governance and existing service management linked?
- In addition to efficiency, do we also measure the quality of AI performance and its impact on people?
- Which domain is currently less developed than the others?
- Do we know what the next step actually needs to be for our organization?
The answers don't have to be positive across the board. That is precisely the point of an honest maturity assessment.
Conclusion: AI maturity develops step by step
Responsible AI service management does not result from a single policy, a new tool, or a completed AI project.
It evolves through day-to-day operations.
The HUCAISM Maturity Model maps out this path. From “Ad Hoc” through “Conscious,” “Controlled,” and “Integrated” to “Optimizing,” it shows how individual AI applications gradually evolve into a resilient operational model. It examines not only which structures are in place, but also whether they actually function in day-to-day operations.
Therefore, the key is not to reach the highest maturity level as quickly as possible. The key is to have a realistic understanding of where you stand and to take the next responsible step.
In this way, the use of AI evolves into systematic AI service management, and individual rules give rise to an organization that can control new AI capabilities and deploy them in a transparent manner, with people at the center.
Most Recent
Would you like to know how HUCAISM translates general AI governance requirements into specific roles and responsibilities for day-to-day service operations? Then be sure to read the previous post:
Training Tip: HUCAISM AI Service Professional at SERVIEW
If you want to understand how to systematically build and further develop AI service management with HUCAISM, this is the HUCAISM Professional Training at SERVIEW is the perfect place to start. You’ll learn about the HUCAISM model, its maturity levels, and the key components for the responsible operation of AI-powered services.
Learn more:
HUCAISM training courses at SERVIEW
