What companies should know about AI and better processes.

When a task requires language, unstructured information, pattern recognition or flexible support and this creates a measurable benefit. AI is unsuitable when a simple rule, an existing system or eliminating the task is a better solution.

WORK:RE helps: We break the process down into tasks and decide between humans, systems, automation and AI for each task.

Not our promise: We don't use AI just because it is available or modern.

You don't need a ready-made AI idea. Choose a real process with noticeable effort, long lead times, quality problems or capacity bottlenecks. First define the desired outcome.

WORK:RE helps: We start with impact and work, not with a tool, and use this to develop the Future Process and the sensible distribution of tasks.

No. For an initial process, a clear sponsor, a process owner, access to specialist knowledge and the willingness to question the current process are sufficient. A company-wide strategy becomes important when capabilities are to be reused across multiple processes.

WORK:RE helps: The first Blueprint provides concrete insights for a later AI strategy, but does not replace company-wide strategy work.

A collection of ideas can provide orientation, but often leads to isolated functions with no process impact. Value only arises when it is clear how the work should work from end to end.

WORK:RE helps: We do not evaluate AI as a detached use case, but as a possible capability within a newly designed process.

No. A well-chosen process is usually a better start than a company-wide transformation program. The scope remains manageable, the benefits measurable and the learning loop short.

WORK:RE helps: We define a realistic target level and a pilot path. We do not hastily derive an enterprise platform from a single process.

No. What matters is not the size of the company, but rather the process volume, repeatability, data access and economic benefit. Smaller companies often benefit from standard platforms and a focused scope.

WORK:RE helps: We dimension the approach to the company and avoid unnecessary platform complexity.

Technology often automates waiting times, unnecessary handovers and historically developed rules. Only when the goal, result and responsibilities are clear can it be meaningfully decided what people, systems or AI will take on.

WORK:RE helps: RE:THINK clarifies the effect, RE:DESIGN designs the process, RE:ALLOCATE distributes tasks and RE:ENGINEER makes the implementation concrete.

A relevant, definable process with a committed owner, available specialists, observable problems and measurable results is suitable. It should be important enough, but not encompass the entire company.

WORK:RE helps: We examine the benefits, feasibility, risk, data situation and willingness to change and recommend a suitable scope.

Not necessarily. A support process can be a great place to start. However, for a company-wide AI infrastructure, several representative processes or entire process domains must be considered.

WORK:RE helps: We distinguish process solution, reusable capability and Enterprise Foundation.

Yes, if volume, turnaround time, error costs or customer impact are relevant. What is crucial is a reliable baseline so that the effect can be measured after the change.

WORK:RE helps: We define business KPIs and connect the target process with a measurable value proposition.

This is often the case and is not an exclusion criterion. The real process is made visible through interviews, case studies, observation and existing data.

WORK:RE helps: We only reconstruct as much of the actual process as is necessary for understanding. The focus remains the future process.

Optimization improves the existing process. Process engineering first asks what work and what results will actually be needed in the future and redesigns the process from there.

WORK:RE helps: We can optimize if it makes sense, but we are not tied to today's process logic.

At least sponsor, process owner, two to five expert employees as well as IT, data and security at an early stage. With regulated processes, risk, compliance or data protection are added.

WORK:RE helps: We structure decisions and workshops. The technical responsibility remains with the customer.

Assist: AI supports people. Embedded: AI works in the existing tool with context. Workflow: AI takes defined steps in a controlled process. Agentic: AI plans partial steps and acts in a limited manner. Native: Work and systems are designed from the ground up for Human + AI.

WORK:RE helps: We determine the appropriate level for each task - not across the board for the entire company.

Yes. The number of processes does not determine maturity. A single process can be agentic if systems are accessible, actions are limited, rights are clean, and controls are effective.

WORK:RE helps: We define the permitted action space and human control points. The agent platform and technical implementation require appropriate specialists.

No. The highest maturity is not automatically the best. A simple Assist or workflow approach can be more economical, secure and robust.

WORK:RE helps: We optimize for better work and measurable impact, not for maximum AI autonomy.

Tasks with high responsibility, normative decisions, relationship work, consequences that are difficult to reverse or context that cannot be adequately controlled usually require human decision or approval.

WORK:RE helps: We make responsibility and checkpoints explicit. The legally binding risk assessment remains with the responsible customer functions.

When inputs are structured, rules are stable and outcomes are deterministic. Classic automation is often cheaper, more understandable and more reliable.

WORK:RE helps: We consciously differentiate between system, automation and AI and combine them in the target process.

When a goal requires several variable steps, information sources or tools and the scope for action can be clearly defined. An agent is unsuitable if errors cannot be detected or reversed.

WORK:RE helps: We define tasks, tools, rights, boundaries, approvals and success criteria. We don't sell an agent as an end in itself.

Yes, that's actually typical. For example, AI can summarize information in an assistive manner, analyze it embedded in the CRM and create a draft within a workflow while humans make the decision.

WORK:RE helps: The classification takes place at the task level and is visible in the Task Allocation Model.

A clear problem, responsible roles, accessible information, approved AI access, identity and authorization concepts, data protection rules, quality checks and an operational owner.

WORK:RE helps: We check the requirements for each target level and document gaps in the implementation Blueprint.

No, but they must be sufficiently up-to-date, discoverable, understandable and legally accessible for the specific purpose. Bad data cannot be compensated for with a better model.

WORK:RE helps: We determine which data the target process really needs. Cleansing and technical data pipelines are separate delivery services.

Often not for Assist. Stable APIs, connectors or controlled browser actions are usually necessary for workflow and agentic. Critical transactions should preferably run over robust interfaces.

WORK:RE helps: We identify the required system actions and interfaces. The detailed design and API development are carried out by IT or partners.

At least basic rules on allowed tools, data, responsibility, auditing and incidents. Governance must match the risk of the process and can grow with the first controlled applications.

WORK:RE helps: We define process-related control requirements. Complete enterprise AI governance is a mission in itself.

It's central. Without specialists, exceptions, quality criteria, risks and implicit decisions cannot be recorded reliably.

WORK:RE helps: We make this knowledge explicit in the process and requirements. We do not replace professional responsibility.

Then Assist may still be possible. For greater integration, existing workflow tools, APIs, RPA or an enterprise browser can create bridges. However, the long-term solution must not make technical debt invisible.

WORK:RE helps: We separate short-term pilot paths and sustainable target architecture.

About criteria for process clarity, data, interfaces, identity, security, operations, specialist capacity and willingness to change. The requirements increase significantly from Assist to Agentic.

WORK:RE helps: We use an Engineering Readiness Assessment and a Security & Governance Gate.

WORK:RE defines from the target process which technical skills are required. A single process typically creates a process-focused solution architecture, not an enterprise-wide AI platform.

WORK:RE helps: Target architecture, requirements, maturity, data flows, controls and roadmap. Not automatically included: Installation, software development, cloud operation and license procurement.

A Solution or AI Architect translates the WORK:RE Blueprint into a Detailed Solution Design. Data, Integration, Security and Operations flesh out their respective areas.

WORK:RE helps: We create a common, professional and technical starting point and can coordinate the architectural work. Customer or partner experts are responsible for special disciplines.

A reusable pattern that describes required capabilities, components, data flows, and controls. It's not yet custom enough to be built out of the box.

WORK:RE helps: We use reference architectures for Microsoft, AWS, Google Cloud, Salesforce, ServiceNow, SAP and Sovereign/Open Stack and adapt them to the process.

The Blueprint describes the goal, skills and requirements. The Solution Design specifies specific tenants, regions, networks, products, APIs, roles, configurations, deployments and operating parameters.

WORK:RE helps: We supply and represent the Blueprint. The detailed design is created with a qualified architect and the responsible IT functions.

In most cases, the existing strategic platform is the best starting point. Workplace, data, specialist systems, cloud strategy, skills, security, regions and costs are relevant. A hybrid solution often arises.

WORK:RE helps: We first define skills and then compare suitable characteristics. We remain manufacturer neutral.

They can be the defining process world: Salesforce for customer-oriented work, ServiceNow for service and workflow processes, SAP for core operational processes. Your data models, rights and automation options heavily influence the architecture.

WORK:RE helps: We take system-of-record responsibility into account and develop an appropriate, not blanket architecture.

Usually no. macOS is an endpoint and workplace environment, but not a complete enterprise AI platform. MDM, identity, browser, DLP, endpoint security and access to the selected cloud or business platform are relevant.

WORK:RE helps: We consider end devices as an access layer. A separate Apple reference architecture is only necessary for special device or on-device AI requirements.

Yes. A controlling layer between employees, AI services, browsers and business applications can enforce identities and access, protect sensitive data, control permitted AI models, log interactions and limit agentic actions. It is particularly relevant if many web-based applications are used or existing systems do not offer their own AI controls everywhere.

WORK:RE helps: From the process, data and target maturity, we determine which control capabilities are needed and where they need to be effective. The selection and technical configuration of a specific security product is carried out using the customer's IT and security architecture. Such a layer does not replace process design or clean data, API and operational architecture.

No. Individual processes should preferentially reuse existing business services. Only recurring requirements from several processes justify a common Enablement Platform.

WORK:RE helps: We mark requirements as process-specific, reusable capability or enterprise foundation.

Yes. Hybrid architectures are common, but increase requirements for identity, networking, data classification, latency, logging and operations.

WORK:RE helps: We make the required skills and data flows visible. The corresponding architects are responsible for network and platform detail design.

The organization remains responsible. Responsibility must not be delegated across the board to one model or provider. Named owners are needed for process, data, system, risk and operations.

WORK:RE helps: We define decision rights and human control points. Legal responsibility is assigned by the responsible customer functions.

Through approved enterprise access, clear data classes, DLP, role rights, tenant boundaries, controlled connectors, redaction, logging and appropriate contractual terms.

WORK:RE helps: We identify data and necessary controls. Configuration, penetration testing and legal testing are special services.

A human reviews, complements, or approves AI output before any relevant action occurs. The checkpoint must provide real information, time and decision-making power - a meaningless click is not enough.

WORK:RE helps: We design the checkpoint as part of the process and determine when it is necessary.

A custom identity, least privilege rights, tool allowlist, input and output validation, prompt injection protection, action and cost limits, full tracing, approvals, Kill Switch and incident response.

WORK:RE helps: We define the professional boundaries. Security architecture and technical enforcement are carried out together with specialists.

Not across the board. Logging must balance auditability, privacy, confidentiality, retention and operational needs. For agentic actions, a traceable action track is particularly important.

WORK:RE helps: We define the process-related proof requirements. The specific logging policy is defined with Security, Privacy and Operations.

With grounding, source evidence, structured outputs, technical test sets, validations, confidence and escalation rules as well as human review for relevant consequences.

WORK:RE helps: We define error consequences and quality criteria. We do not guarantee an error-free model answer.

They determine which data may be processed, in which region, for what purpose, for how long and under which controls. Depending on the industry, additional requirements may be added.

WORK:RE helps: We integrate these requirements early. We do not provide legal advice and are not a substitute for formal compliance clearance.

A target image of the work, a Future Process, a Task Allocation Model, the appropriate maturity for each task, organizational and technical requirements, control points, business KPIs and a prioritized implementation roadmap.

Border: The Process Blueprint is not yet a fully configured system and is not a Solution Design that has been worked out down to the last API field.

The internal IT, an existing implementation partner or a specifically assembled team of solution architect, data/integration engineer, AI engineer, security and operations.

WORK:RE helps: Fäh Consulting can accompany mobilization, architectural handover, partner coordination, pilot and anchoring. Special development is allocated transparently.

Fäh Consulting can hold together the overall path, the process intention and the decision logic. Depending on the scope, technical delivery and operational responsibility must lie with internal IT or qualified delivery partners.

WORK:RE helps: Clear roles, common acceptance criteria and less loss between business and technology.

This depends on process complexity, data access, platform, procurement, security and integrations. Assist can be possible in a few weeks; Integrated or agentic solutions often require several months.

WORK:RE helps: According to the Blueprint, the effort can be estimated much more seriously. We do not mention apparent accuracy before analysis.

By taking identity, data rights, monitoring, ownership, technical reuse and a realistic operating path into account in the pilot. At the same time, the pilot must not be overloaded with an enterprise platform.

WORK:RE helps: We separate pilot scope, reusable capabilities, and later Enterprise Foundations.

You need a technical product/process owner and a technical service owner. In addition, support, monitoring, incident response, model and prompt changes, costs and recertification must be regulated.

WORK:RE helps: We define the required operating model. Ongoing IT operations are only part of the service if expressly agreed.

With process metrics such as effort, lead time, error rate, first-time right, customer outcome, capacity and risk - supplemented by technical values ​​such as quality, latency, usage and costs.

WORK:RE helps: Impact is measured by the process, not by the number of AI use cases generated.

In addition to model usage, this includes licenses, data preparation, integration, security, testing, change and operation. The biggest costs are often not the model, but rather the reliable integration into the organization.

WORK:RE helps: We reduce unnecessary scope and create the basis for a reliable cost estimate. Binding implementation costs arise in Solution Design.

WORK:RE is Human–AI Process Engineering. AI is an important option, but not the starting point. The focus is on better work, clear processes and a sensible distribution between people, systems, automation and AI.

No. The method is manufacturer-neutral. Existing platforms such as Microsoft, AWS, Google Cloud, Salesforce, ServiceNow or SAP are taken into account if they make sense for the customer.

Not our model: A platform is not recommended because it needs to be sold.

No. WORK:RE connects business, process and technology and creates the ability to make decisions and implement them. Internal IT remains central to architecture, standards, security, integration and operations.

The core service is process and implementation design. Prototyping or delivery can be organized and accompanied; Extensive software development is carried out by internal teams or specialized partners.

No. We make relevant requirements and decision-making needs visible and integrate the responsible specialist departments. Formal legal, data protection and compliance assessments remain with qualified experts.

A classic workshop often ends with ideas. WORK:RE creates a coherent Process Blueprint: target impact, Future Process, task distribution, maturity, technical requirements, governance, KPIs and implementation path.

If a fully specified solution simply needs to be programmed, only a software license is sought, or there is no responsible process owner and no access to specialist knowledge available.

Even if an organization just wants an AI demonstration without process change, WORK:RE is probably not the appropriate approach.

With a non-binding discussion about the reason, process, desired effect, sponsor and general conditions. A decision is then made as to whether a Process Sprint makes sense and who needs to be involved.

Book your first conversation ↗