Autonomous Enterprise Journey: How AI value engineering creates measurable business value

Where is AI worth implementing in your own company?

Whether it’s order processing, procurement, human resources, or production planning: With complex workflows and heterogeneous system landscapes, it’s often unclear where new technologies like AI can be effectively applied to operate more efficiently and profitably.

How can AI value be reliably assessed, and how can the return on investment of AI use cases be measured? How can companies identify the key strategic levers and critical processes for their business success?

Katrin Luft, Senior Expert at valantic and co-developer of the “Autonomous Enterprise Journey”, provides answers and recommendations in a conversation with Darya Basarhina, Vice President of Sales & Marketing and Manufacturing Practice Lead at valantic.

Value First: Business Value as the starting point for the Autonomous Enterprise Journey

Katrin, nearly every company is now exploring AI. But many don’t know where to start. Where do you begin?

Our approach is the Autonomous Enterprise Journey, where we start with the company’s strategic goals and the business processes that are most critical to achieving them. This is based on the “value first” principle: Together with the client, we identify the processes that are particularly relevant to competitiveness and business success, clarify the outcomes these processes must deliver, compare suitable solutions, and evaluate their expected business value in relation to costs, risks, and feasibility. In other words, we focus on use cases with demonstrable value potential and realistic conditions for implementation.

You also talk about “AI Value Engineering”. What do you mean by that?

This term describes a systematic approach within the Autonomous Enterprise Journey. In AI Value Engineering, we identify, quantify, and prioritize the opportunities and value potential of AI and automation; we provide support from the initial assessment and the definition of target values and KPIs through to implementation and ongoing verification of the actual impact achieved.

Which business processes are the focus of AI Value Engineering?

The selection depends on the customer’s strategy and business model. In the manufacturing industry, these are typically the chains from planning to production (Plan-to-Produce), from customer order to cash receipt (order-to-cash), from procurement to payment (procure-to-pay), and from product development to plant operation (design-to-operate). We examine these processes across system and departmental boundaries to identify where the greatest levers for business success lie and how a change at that point affects the entire chain.

Does the value-first approach always involve AI?

No, traditional automation can also be useful for repetitive, rule-based tasks, such as data transfer, data reconciliation, or standard workflows. AI is particularly helpful with unstructured information, classification, forecasting, and decision-making. Agentic AI plans and executes multiple steps in a targeted manner across connected systems. To do this, we define in advance which actions can be performed autonomously and which require human review or approval. Sometimes it also makes more sense to rethink a process from the ground up using AI, rather than optimizing an existing one in individual steps.

What role does the choice of technology play in the Autonomous Enterprise Journey?

The technology aligns with the business goals and requirements of the specific use case. valantic provides system-agnostic consulting, meaning we remain open to all technologies. We take into account the existing system landscape, integration options, and dependencies on vendors. To this end, we maintain partnerships with leading hyperscalers and specialized providers of data and AI solutions. This allows us to determine for each use case whether traditional automation, AI, or Agentic AI is the right approach. We combine deep process knowledge with solid industry expertise; we have also identified use cases for numerous industries and made them available to our clients. Prerequisites include a suitable data foundation, clear responsibilities, and an architecture that ensures secure and cost-effective operation.

Quantify and prioritize value potential

How do you identify the biggest value drivers?

We assess the strategic importance of business processes and their potential. Areas for improvement include, for example, situations where errors compromise quality, long lead times cost revenue, or a shortage of skilled personnel limits growth. Many manual steps, makeshift Excel solutions, and a lack of transparency are also typical warning signs.

To assess the expected business value and prioritize processes and use cases, we also consider what organizational changes are necessary to generate measurable benefits. Where data availability permits, we use process mining: Here, system data reveals real-world workflows, which helps us identify bottlenecks and deviations.

The result is a value heatmap that highlights the greatest potential and a roadmap that prioritizes initiatives based on business value and feasibility. It is important to distinguish between the projected value potential and the proven impact.

How do you quantify the expected business value?

In the first step, we examine the initial situation, such as case volumes, processing times, costs per transaction, error rates, and turnaround times. Building on this, we develop target values and KPIs and determine the timeframe and data we’ll use to measure progress. We compare the expected benefits with the total costs. This includes expenses for development and integration, onboarding, enablement, and organizational adjustments, as well as ongoing costs for models, licenses, hosting, maintenance, and further development.

A practical tool supports us in this process: The valantic Impact Miner allows us to compare our findings with appropriate benchmarks, which serve as a guide for target values. We verify whether the benchmarks are truly comparable in terms of process scope, case structure, and measurement methods.

The result is a business case with hypotheses that need to be validated. For example, if working time is freed up, we initially view this as a gain in capacity, but we clarify with the client whether it actually reduces bottlenecks, lowers costs, or enables additional business. Where appropriate, we calculate the ROI and the payback period. A valuation in euros is generally performed only based on reliable assumptions.

Do you have an example of what AI value engineering looks like in practice?

A classic example found in almost every company, regardless of industry, is the invoice processing workflow. It becomes particularly time-consuming when invoices don’t match orders or goods received and have to be reconciled manually. The use case is intentionally kept small: It delivers quickly measurable benefits, builds trust in AI, and lays the foundation for gradually scaling up to the most strategically important use cases.

For a client with approximately 120,000 invoices per year—of which about 20 percent contain relevant discrepancies—we implemented the solution in stages. It automatically detects discrepancies, prepares clarification messages that are sent after human approval, and assists in processing incoming responses. Technical review and posting remain the responsibility of the business unit.

The business case projects a 45 to 60 percent reduction in manual processing time per clarification case. In the mid-range scenario, this results in an annual net benefit of approximately 138,000 euros after operating costs, with a payback period of about 13 to 14 months.

Realize business value and measure impact

How does a business case turn into a practical, actionable roadmap?

Our Autonomous Enterprise Journey typically consists of three components that build on one another.

The starting point is a Discovery Snapshot: In this workshop with senior management or key stakeholders from business units and IT, we define the goals and relevant business processes, examine three to five use cases, and provide an initial assessment of the potential’s scale. By the end, it’s clear whether a more in-depth analysis is warranted.

Following this, in a series of Impact Case workshops, we collect data from the systems, work with the client to refine target values, KPIs, and evaluation assumptions, and evaluate eight to twelve use cases, including a business case.

The Impact Roadmap complements a data and architecture review, the design of the first use case, and an implementation plan spanning two releases. This provides the company with a solid basis for decision-making that includes the key prerequisites and dependencies.

How quickly will the impact of the Autonomous Enterprise Journey become apparent?

We estimate three to six weeks for an in-depth analysis and implementation planning. In the first week, we focus on the goals and business processes. In weeks 2 through 4, we evaluate the potential of the use cases, clarify the data landscape, and define the target architecture. In weeks 5 and 6, we’ll work on the solution concept, KPIs, and the implementation roadmap. The specific scope and duration depend on the initial situation, data availability, and the complexity of the use cases.

And how long does it take to launch the first production AI use case?

For a clearly defined use case, we aim for implementation within 30 to 90 days—depending on the complexity of the use case. Once the system is live, we compare the results of the agreed-upon KPIs with the baseline, examining changes in case counts, error rates, and other factors affecting process performance. After going live, we begin iterative refinement and optimization based on the measurement results or the rollout to other business processes.

What else is needed besides the technology?

First, we need people who are committed to the solution. Part of our Autonomous Enterprise Journey involves engaging employees and user groups early on and adapting the organization, roles, and responsibilities—including accountability for the business value achieved.

Second, it requires holistic governance: We review the legal requirements for each use case with every customer, with a particular focus on the EU AI Act. We embed data protection, information security, and human oversight throughout the entire lifecycle, from design through ongoing operations. This includes defined access and authorization rights, traceable system actions, quality assurance, human oversight, and approvals.

How do you support clients during ongoing operations?

After implementation, we continuously measure the value achieved based on the agreed-upon KPIs. Together with the business unit, we continuously review the processes, optimize them, and use the value heatmap to identify new use cases, which we implement according to their value contribution as prioritized in the roadmap.

As part of operational monitoring, we track costs, models, and architecture, and compare costs to the value achieved. If a new model delivers the same quality at a lower cost, we facilitate the transition. In this way, we continuously refine the processes in line with the agreed-upon goals.

What advice do you have for companies that want to get started now?

They should begin with the most strategically important business processes and identify the greatest levers for business success there. Clearly defined target values and KPIs lay the foundation for selecting the first use cases and for measuring success later on. From the very beginning, it should also be clear who is responsible for the benefits within the business unit, how work processes will change, and how the solution will be operated long-term. Even an ongoing SAP transformation is no reason to wait. In fact, this is precisely the time to identify automation readiness and quick wins in parallel, aligned with the transformation’s goals, resources, and dependencies.

Thank you very much for your insights and recommendations, Katrin.