AI in Projects: How Well Is It Really Being Implemented in Germany?

First, the good news: The catch-up effect is real

The Bitkom figures from early 2026 are remarkable. 41 percent of German companies with at least 20 employees are actively using AI – compared to 17 percent the previous year. This is not a linear increase. It is a leap.

And it doesn't stop at adoption: 77 percent of AI-using companies report an improved competitive position, and 52 percent say AI contributes measurably to business success. These are no longer mere pilot projects riding a wave of hype expectations – these are concrete results.

“AI is the most important future technology for 81 per cent of German companies – up from 73 per cent in 2024.” – Bitkom, September 2025

Even more interesting: only 17 per cent now describe AI as “overhyped” – in 2024 it was 26 per cent. The scepticism of the early 2020s is giving way to a more sober, yet decidedly more positive attitude.

Shorter project timelines – but only when AI is genuinely used

What happens in practice when AI is deployed on projects? Most of the figures in circulation are self-assessments from surveys. But there is a measurement. Harvard Business School ran a pre-registered field experiment with Boston Consulting Group: 758 consultants, randomly assigned to three groups – no AI, GPT-4, or GPT-4 plus an introduction to prompting – then set 18 realistic consulting tasks. Those working with AI completed 12.2 per cent more tasks and took 25.1 per cent less time over them, at more than 40 per cent higher quality.

The second half of the result is the more important one. On a task deliberately placed beyond the model’s capabilities, the AI users were 19 percentage points less likely to reach the correct answer than the control group. The authors call this the jagged frontier: inside what the model can do, you get faster and better. One step outside it, you get faster and wronger – and you do not notice.

This does not transfer one to one: what was measured were consulting tasks, not project timelines, and not in Germany. For German project work there is only survey data so far. In a Capterra survey of 200 German project managers, 44 per cent use AI in their project management tools – so the narrow majority do not, and see no gains accordingly.

This aligns with what I observe in projects. When AI is used consistently – for status reports, for processing stakeholder feedback, for risk analyses – it genuinely saves time. But “using it consistently” requires more than a tool subscription. It requires the entire team to share a common understanding of where and how AI can be meaningfully applied.

Numbers that matter

41%

of companies are actively using AI – twice as many as the previous year (Bitkom 2026)

25.1%

less time per task – measured in a field experiment, not self-reported (Harvard Business School / BCG, 758 participants, 2023)

~80%

of AI initiatives fail to deliver the intended business value (RAND Corporation, 2024)

65%

of executives worldwide are unable to define a clear business case for AI (DXC Technology, August 2025)

The uncomfortable side: Why so many AI projects fail

Now for the figure that never appears in any presentation, but should have featured in every retrospective. Various industry reports – including analyses from MIT circles – suggest that the majority of AI pilot projects fail to achieve scalable value. The exact figures vary depending on the definition, but the published surveys place them between 70 and 90 per cent: the RAND Corporation puts it at around 80 per cent, and a widely cited MIT analysis of generative AI at 95 per cent of pilot projects. Failure here does not mean chaos – it means that the hoped-for outcomes fail to materialise despite the investment.

What lies behind this? Three patterns keep recurring.

1. The first is the missing “why.” A DXC Technology survey of 2,496 executives across 23 countries shows that 77 per cent view AI as a strategic priority – yet 65 per cent cannot articulate specifically what AI is supposed to achieve for their organisation. That is not an AI problem. That is a strategy problem.

2. The second pattern is poor data maturity. AI systems are only as good as the data they work with. Many companies launch pilots without first clarifying their data foundation. The result: the AI delivers – but what it delivers is worthless.

3. The third pattern is the skills gap – and it is compounded further in an AI context. McKinsey shows that 33 percent of German employees lack the skills required for their current role, and 44 percent did not spend a single day on training in the past year. When teams do not know how to work with AI outputs, the benefits fail to materialise – regardless of how good the tool is.

What I keep seeing: when only individual team members actively use AI while the rest wait and see, no efficiency gain emerges – instead, a new coordination problem is created.

Philip Müller

Where AI in projects actually lands today

A 2025 GPM study with 176 participants shows where AI in project management is actually being used today: primarily in communication. Creating summaries, preparing minutes, drafting standard responses. That is not nothing – but it falls far short of what would be possible.

Respondents see the greatest potential in risk management. This makes sense: risk assessments often depend on structured volumes of information that AI can process well. Yet this is precisely the area where the fewest teams are currently working with AI in a systematic way.

More than half of German companies cite AI capabilities as the main reason for investing in new project management software. The market is moving – the question is whether actual usage will keep pace with investment.

What works in practice – and what I take away from it

I am not surprised that failure rates are high. What does surprise me: how many companies carry on afterwards, repeat the same mistakes, and label the outcome an “AI problem.”

There is a decisive difference between treating AI as a technology initiative and treating it as a working method. Not “We are piloting AI” – but “How do we solve this specific problem, and can AI help?” That sounds like a semantic nuance. It is not.

Three things I have taken away for myself:

1. Small, concrete use cases beat grand visions. Those who start with “AI should transform our entire project work” fail. Those who start with “AI should summarise our sprint retros” have usable results within two weeks.

2. Humans decide – AI proposes. The GPM study confirms what I know from practice: hybrid models work. AI for information processing and structuring, humans for judgement and context.

3. Data protection is a genuine factor in Germany, not an excuse. 43 per cent of German project managers name data protection and security as one of their three biggest challenges in adopting AI – against 32 per cent worldwide. In both groups the top spot goes to data quality. This needs to be part of every AI strategy from the outset, not an afterthought.

Conclusion: Germany is moving – but not uniformly

The catch-up effect is real. The successes among those who use AI consistently are real. But the gap between “naming AI as a priority” and “using AI in a way that makes projects run better” remains wide.

What makes me optimistic: the companies getting it right do exist. They are rare, but they are real. And what they have in common is not a larger budget or better tools – but clearer questions at the outset and more patience in the process.

The most honest answer to the question of how well AI is being implemented in German projects: better than two years ago, worse than the presentations suggest – and with a recognisable pattern for anyone who looks closely.

Sources

• Bitkom: Digitalisierung der Wirtschaft 2026 — 604 companies with 20 or more employees, surveyed by telephone in calendar weeks 2 to 6 of 2026, published 11 March 2026. Source of: 41 per cent AI adoption (previous year 17 per cent), 77 per cent improved competitive position, 52 per cent measurable contribution to business success.
• Bitkom: Durchbruch bei Künstlicher Intelligenz — 604 companies with 20 or more employees, calendar weeks 27 to 32 of 2025, published 15 September 2025. Source of: 81 per cent regard AI as the most important future technology (2024: 73 per cent), 17 per cent as a passing hype (2024: 26 per cent).
• Capterra: 2024 Most Impactful Project Management Tools Survey — 2,500 respondents worldwide, 200 of them project managers in Germany, fielded May 2024, published 4 September 2024. Source of: 44 per cent using AI in project management tools, 89 per cent reporting a positive return, and data protection as a challenge for 43 per cent in Germany against 32 per cent worldwide.
• DXC Technology: survey of 2,496 technology and business leaders across 23 countries, August 2025. Source of: 77 per cent naming AI a priority, 65 per cent without a clear business case, 94 per cent facing substantial barriers to scaling. The survey is international, not German.
• McKinsey: HR-Monitor 2025, published 21 July 2025 — around 2,000 companies and more than 4,000 employees across Europe and the USA, roughly 1,000 of them in Germany. Source of: 33 per cent lacking the skills their current role requires, 44 per cent without a single day of training in the previous year.
• GPM: 2025 survey of 176 participants on the use of AI in project management.
• Dell’Acqua, F. / McFowland III, E. / Mollick, E. / Lifshitz-Assaf, H. / Kellogg, K. / Rajendran, S. / Krayer, L. / Candelon, F. / Lakhani, K. (2023): Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality. Harvard Business School Working Paper 24-013 — a pre-registered, randomised field experiment with 758 Boston Consulting Group consultants across 18 tasks. Source of: 12.2 per cent more tasks completed, 25.1 per cent less time, more than 40 per cent higher quality inside the capability frontier; 19 percentage points fewer correct solutions outside it.
• On the failure rate of AI initiatives: RAND Corporation (around 80 per cent not delivering the intended business value) and a widely cited MIT analysis of generative AI (95 per cent of pilot projects).
• Not used, because not verifiable: the “PMI Global AI Report 2025” cited here previously does not exist under that name. Nor does the widely repeated claim that PMI found AI-using organisations delivering 61 per cent of projects on time against 47 per cent without AI appear in the Pulse of the Profession 2025 — that report is about business acumen, and the figure 47 does not occur in it.