netzstrategen AI Operations.
Insights Glossary Resources Deutsch Self-Check
← All posts
Strategy

Measuring AI ROI: Why Broad Adoption in Germany Does Not Prove Value Yet

Published on 8/17/2026 · André Hellmann

ROI Implementation Gap DIHK Deloitte Mid-Market

German companies deploy AI more broadly than companies in any other market studied. 41 percent report that more than 60 percent of their staff use AI tools, against an international average of 29 percent (Source: Deloitte, The ROI of AI — German Cut, 2026). On return, the picture flips. Anyone unable to measure AI ROI defends the budget in the next round with anecdotes. That rarely ends well.

Correction, 19 August 2026

The first version of this article cited two figures we attributed to the DIHK Digitalisation Survey 2026: that 81 percent could not quantify AI value and that more than 75 percent saw no measurable return. A check against the original source confirmed neither figure. They came from a secondary source that blends numbers from several studies. We have rebuilt the article on verifiable data. The core argument is unchanged — the evidence behind it can now be checked.

Where do you stand?

Discuss your next step in a free diagnosis call. Book a slot →

Contents

What the German figures show

The DIHK Digitalisation Survey 2026 polled 4,686 companies between 10 and 28 November 2025. The report is titled “Artificial Intelligence, Sovereignty and Resilience” (Source: DIHK, 2026).

One finding from it sets up this article. Among companies already running AI in practice, 41 percent rate the productivity effect as high (Source: DIHK Digitalisation Survey, 2026). So the majority of active users do not.

The second figure comes from Deloitte. More than two thirds of organizations reach their typical AI return only after two years or later (Source: Deloitte, The ROI of AI — German Cut, 2026). Deloitte calls this Germany’s own paradox: high adoption, limited strategic value.

Both numbers explain each other. Broad adoption on its own produces no return. It produces activity. Return appears once workflows are rebuilt and outcomes are attributed.

That is notable because the two numbers do not contradict each other — they explain each other. Companies that do not measure find nothing. And what cannot be found cannot be steered.

The international picture matches: 95 percent of GenAI pilots show no measurable effect on the profit and loss statement (Source: MIT NANDA, 2025). Germany is not a special case. The German figures are evidence for a global pattern — the Implementation Gap.

The free Self-Check maps where a company stands in a few minutes. It delivers the first defensible estimate.

Why measurement fails on structure

The obvious explanation would be a tooling problem: the wrong dashboard, missing tracking. It falls short.

AI value appears in workflows, not in applications. A team saves 20 minutes per quote. Whether that becomes money depends on what happens to those 20 minutes. Invested in more quotes, they create revenue. Spread across the day, they create nothing measurable.

That explains why measurement fails. The counter is not missing — the attribution is. Three gaps show up almost every time:

  • No named workflow. “We use AI in marketing” is not a unit of measurement. “We produce product copy” is.
  • No named ownership. Without an owner, nobody delivers or defends a number.
  • No link to the P&L. Time saved is not a P&L line. It has to be translated — into revenue, cost or cycle time.

Not measurable does not mean worthless. It means: not defensible in the next budget round.

The three building blocks of measurable operations

Measurability comes from three decisions, not from a tool. They are unspectacular, which is exactly why they get skipped.

Two paths, one difference Why AI value either becomes measurable — or does not The usual path Tool rolled out Usage grows Success claimed not valuable The measurable path Workflow named Owner named P&L metric set valuable Illustration: netzstrategen · Schematic, not survey data netzstrategen
The difference between valuable and non-valuable AI is decided before the first prompt — when workflow, ownership and target metric are set.
For presentations:

Block 1: the named workflow. A process with a start, an end and a volume. How many quotes per month? How long does one take today? Without a baseline there is nothing to compare later. Why the workflow comes before the tool is covered in workflow-first.

Block 2: the named owner. One person accountable for the number. Not IT, not “the team” — a person with a name. That person also decides what happens to the time freed up.

Block 3: the P&L metric. One line from the profit and loss statement the workflow feeds into. Revenue per sales hour, cost per case, cycle time to invoice. Exactly one metric per workflow, not five.

With all three in place, measurement becomes almost trivial. With one missing, no dashboard helps.

What mid-market companies should do now

Getting started takes no programme. It takes one workflow and six weeks.

  1. Pick a workflow that runs often and touches money. Quoting, complaint handling, tender review.
  2. Measure the baseline before any AI use. Two weeks of records are usually enough.
  3. Assign owner and P&L metric and write both down. Half a page is enough.
  4. Run it for four weeks and record the same value weekly.
  5. Decide: expand, adjust or stop. All three outcomes are valid.

After six weeks this cycle delivers a number that holds up in a budget review. It delivers something else too: practical experience of how measurement works inside the company. How that grows into a full business case is covered in the ROI of AI Operations.

Conclusion: measurability is a budget question

The German figures do not describe a technology problem. They describe an operating problem. AI runs in many companies — it is just not set up so that its contribution becomes visible.

That has an uncomfortable consequence. Budgets are not cut by impact but by provability. Companies that want to keep investing in 2027 need numbers in 2026. Not many — defensible ones.

Measurability is therefore not a reporting topic but part of operations. It belongs in the design, not in the follow-up. That is exactly what AI Operations means: running AI as a permanent function instead of evaluating it as a project.

Frequently asked questions

How do you actually calculate the ROI of AI?

Through a named workflow with a baseline. Cost of the solution against the change in one P&L line — revenue, cost or cycle time. Without a measured baseline, every calculation stays an estimate.

Is time saved enough as proof?

No. Time saved is an intermediate step, not a P&L metric. It counts only once the freed-up time visibly turns into revenue, lower cost or more throughput.

Why does the majority of AI users report no high productivity effect?

Because workflow, ownership and target metric are usually not defined. Only 41 percent of companies with AI in active use rate the productivity effect as high (Source: DIHK Digitalisation Survey, 2026). Measurement fails on structure, not on tooling.

How long until the first defensible number?

For a clearly scoped workflow, around six weeks: two weeks of baseline, four weeks of operation. What matters is recording the same value in both phases.

Where is the best place to start?

With the workflow that runs often and touches money. We map out which workflow makes the best first proof point in a free diagnosis call.

Sources

What's next