CQTA Robert Krupp (del Valle)

DE EN

Focus area

AI rarely fails on the technology.

The technology has long since arrived. What is missing is the organization to carry it.

Four pieces of work from recent months — a white paper, two articles and a survey of our own — arrive at the same conclusion. Here it is in short.

In six sentences

  1. The technology is here.Nine in ten industrial companies use AI or plan to start within 24 months.
  2. The ambition stays small.Half want to cut costs with it. Innovation is named by 13 percent.
  3. The gap keeps widening.Technology grows exponentially, organizations logarithmically. The backlog sits in between.
  4. Failure happens with people.Around 70 percent of transformation projects fail — mostly on culture, communication and change management.
  5. Piecemeal does not work.A tool here, a pilot there. Impact only appears when five levels work together.
  6. And the business model follows.Once you have usage data, you no longer sell machines, you sell uptime.

People before machines

Technological progress runs exponentially, an organization's capacity to adapt runs logarithmically. That discrepancy — known as Martec's Law — is the real bottleneck in every AI rollout.

The gap between what is possible and what gets done Technology Organization The gap Time
The upper curve is not yours to change. Working on the gap means working on the lower one.

A handful of conditions decide whether an organization lifts that lower curve. They sound soft and are anything but: without them, even ambitious initiatives come to nothing.

  • A change mindset

    The willingness to question existing assumptions. "That's how we've always done it" is the single biggest brake on innovation — no other sentence captures it as precisely.

  • Real participation

    Harvard research puts the success rate of projects with serious employee involvement around 50 percent higher. In one in five German companies it still does not happen.

  • Psychological safety

    People who fear consequences defend the status quo. Up to 60 percent of managers worldwide see honest conversations about what AI means for jobs as the most effective lever.

  • And only then, timing

    A transformation should start only once there is a shared picture of the goal. Moving fast without that produces short-term effects, not change.

In full in the article "People before machines" — read it and download the PDF (German).

AI can calculate change. Only people can shape it.

Five levels instead of isolated measures

Isolated optimization falls short. In the white paper, Dr. Norbert Lühring, Prof. Dr. Dr. h.c. Cornelius Herstatt and I describe an operating model that treats product development as one connected stack. AI has no level of its own in it — it accelerates processes and methods.

Five levels, one system 1 · Product strategy Set direction — markets, platforms, innovation fields AI indirect 2 · Controlling and measurement Make it measurable — KPI cascade, dashboards Automation 3 · Process design Steer the flow — stage-gate with agile sprints, frontloading AI acts here 4 · Methods and tools Secure quality — QFD, FMEA, DFM/DFA, target costing AI acts here 5 · Organization and roles Enable delivery — cross-functional teams, clear ownership New roles If one level drops out, the whole system loses its effect.
The five-level operating model from the white paper "Product development and artificial intelligence" (2026).

Why this has to happen early is shown by one of the oldest calculations in engineering: up to 80 percent of later cost is locked in during the first phases. A change made in production costs a multiple of the same change made at concept stage. That is exactly where the economic leverage of AI sits — it makes early testing, simulating and discarding cheap.

What a late change costs ×1×3×7 ×15×40×100 ConceptDesignDevelopment PrototypeProductionMarket Relative cost of change, concept = 1
Boehm's cost-of-change curve: between concept and production a factor of up to 1 : 300.

In full in the white paper — request the PDF from Lischke Consulting (German).

From product to platform

The most interesting consequence of digitalization is not faster development, it is the changed business model. Sensors in industrial goods capture usage and condition data in real time. Once you analyze that data, you no longer sell a machine, you sell its availability.

Four stages of value creation Product sold once Usage sensors, data digital twin Service predictive maintenance pay-per-use subscription Platform customers, suppliers, partners connected benchmarking co-innovation Loyalty and value contribution rise with every stage
Data-driven business models do not come from technology alone. They come from access to usage data — and the willingness to rethink your own offering.

In full in the article "Digitalization opens up new business models" — lesen und als PDF herunterladen.

Backed by our own survey

In April 2026 we surveyed 115 specialists and executives at German industrial companies across twelve sectors. The findings support every one of the points above.

90 %use AI or plan to start within 24 months
42 %name the skills shortage as the biggest operational bottleneck
26 %cannot assess their own AI strategy
13 %see innovation as the main benefit of AI
What companies expect from AI Cost savings 50 % Faster processes 40 % Higher quality 38 % Innovation 13 % Multiple answers possible, N = 115
AI is being used as a tool for optimization, not for transformation. That is precisely where the untapped potential sits.

Further findings: 72 % rate product development as highly or very highly strategic · 58 % plan to expand AI within 12 to 24 months · 35 % work hybrid, only 4 % purely agile · concept development (33 %), idea generation (26 %) and testing (24 %) lead AI adoption. Study "Product development in transition", Lischke Consulting, April 2026 — request the PDF (German).

What counts now

The legal frame is set: since August 2026 binding obligations apply to high-risk AI — risk management, traceable data, human oversight, documentation. That is effort, not an obstacle. Set up governance early and you deliver in regulated markets while others are still working out who signs.

What matters far more is the attitude behind it. AI recognizes patterns and prepares decisions. It can work through variants no person would get through in a working lifetime. What it cannot do is explain to a team why their work is changing, and get them excited about it.

The most interesting phase of product development has only just begun. It belongs to those who bring both together — the computing power of the machine and the judgment of the people working with it.

Further reading

These publications are currently available in German only.

  • Product development and artificial intelligence

    White paper with the complete five-level operating model, the method chain from QFD to target costing, and an ROI calculation. Together with Dr. Norbert Lühring and Prof. Dr. Dr. h.c. Cornelius Herstatt, June 2026.

    Request the PDF from Lischke Consulting (German)

  • People before machines

    Why a change mindset is the decisive success factor for AI — with participation and psychological safety as the two other conditions. February 2026.

    Read the article and download the PDF (German)

  • Digitalization opens up new business models

    Data-driven, service-oriented, growth-focused: from sensors and digital twins through pay-per-use to platform ecosystems. May 2025.

    Read the article and download the PDF (German)

All of these were written in the course of my work for Lischke Consulting.

Where does your company stand?

Whether AI stays a pilot at your company or actually reaches development is rarely decided by the technology. Tell me where it is stuck — a conversation usually makes quite quickly visible which of the five levels is the problem.

Get in touch