What remains of consulting when the client can create the concept alone?
The short answer
What remains of consulting is the part a language model does not deliver: selection under uncertainty, conflict and commitment in execution. A field experiment with 758 consultants shows both sides. With AI, they worked faster and at higher quality, but they were significantly more often wrong when the task sat outside the model's capability frontier.
The industry did not see the break coming
The German consulting market reached 49.0 billion euros in 2025. Growth was 0.5 percent, while the association itself had forecast 6.4 percent (BDU, 2026). The gap between expectation and result is roughly 2.8 billion euros.
The gap is more informative than the actual value. A market that grows more slowly than expected is a cycle. A market that its own experts misjudge by that magnitude has changed in a way that is hard to see from inside.
The distribution makes it clearer. Consultancies below one million euros in revenue shrank by 2.5 percent and were the only size class in decline. Larger houses held or slightly improved their position. The missing utilization shows up first where the offer has become most interchangeable.
AI grows as a consulting field and pressures price at the same time
AI consulting grew by 18.8 percent in 2025 and is expected to grow by another 22 percent in 2026 in a market that overall stood at 0.5 percent growth (BDU, 2026). The association calls this two sides of one coin.
On the other side, average billed daily rates fell by 2 percent to 1,300 euros across hierarchy levels (BDU Honorare im Consulting 2025). Nominally that sounds small. After inflation, it is more.
What is striking is how the industry reads the movement. 66 percent of surveyed consultancies see no current pricing effect from AI, and exactly 1 percent raises fees because of it. A separate Luenendonk survey reports a similar picture.
The pricing question has not yet been asked, although daily rates are already softening. The next two years will be decided in each firm separately.
The field experiment shows both sides
758 Boston Consulting Group consultants completed 18 realistic tasks in a randomized field experiment; some had access to GPT-4. The AI group completed 12.2 percent more tasks, needed 25.1 percent less time and delivered more than 40 percent higher quality (Organization Science, 2025).
These numbers matter because of the design: random assignment, comparable tasks, and work that resembled consulting practice. The measured result was output, not the feeling of productivity.
The gain was uneven. Previously weaker consultants improved by 43 percent; the strongest by 17 percent. AI therefore lifts the lower part of the performance distribution, the very part on which classic leverage models rely.
The second finding is cited less often and matters more. On a task deliberately designed outside the model’s capabilities, AI users were 19 percentage points less likely to be right than the control group. Where the model sounds plausible and is wrong, the human follows.
The most robust statement about consulting is therefore this: execution becomes cheaper and better. Judgment at the capability frontier becomes more valuable because more mistakes happen there than before.
A concept in two hours is no longer proof
Resistance to AI in knowledge work is less often pure technology skepticism. Documented drivers are changes in work content and loss of status (Electronic Markets, 2022). A Harvard Business School working paper describes the mechanism as role compression: automation takes the higher-value parts of a role, and what remains feels smaller than what people signed up for.
The BCG experiment points in the opposite direction. The automated part is the part that can be described and repeated. The harder part is recognizing where the plausible result is not true.
That shifts where consulting value becomes visible. As long as consulting was proven by the concept document, the deliverable was the evidence. If a concept can be produced in two hours, it is no longer evidence.
The proof lies in what keeps running after the mandate
A concept that ends with the final report leaves no proof of its impact. The reason is timing: impact appears after the consultants have left, and no one then measures what it came from.
The gap is closed by a commitment at the right moment: What target does this measure serve, what value contribution do we expect, and how would we know that it is not delivering?
BizzPlAI is the Strategy Execution System that keeps running after the mandate and makes consulting recommendations provable. Target, expected value contribution and stop condition are set at the decision and remain findable. Twelve months later, realized value can be placed next to it.
The self-test needs no system. Take your last three completed mandates and ask whether you could prove today what they changed for the client. Where the numbers are missing, the impact may still exist, but the proof does not.
Common follow-up questions
Does AI replace analysis work in consulting projects?
Partly, yes. Research, structuring and first concept drafts can now be created faster and at good quality. What a model does not provide is selection between options under uncertainty, and that is exactly where the BCG field experiment showed a drop in accuracy among AI users.
Why do daily rates fall when AI increases productivity?
Because the productivity gain reaches the client, not automatically the provider. As long as the same service is billed by time, faster work means fewer billable days. Two independent surveys show that the industry has barely addressed the pricing question.
What does the shift mean for small consultancies?
Consultancies below one million euros in revenue were the only size class with declining revenue in 2025. Their question is not whether they use AI, but how clients will notice the difference and what keeps running after the mandate.
Sources
- Facts & Figures zum Beratermarkt, Ausgabe 2026BDU Bundesverband Deutscher Unternehmensberatungen, 2026
- Studie Honorare im Consulting 2025BDU Bundesverband Deutscher Unternehmensberatungen, 2026
- Navigating the Jagged Technological FrontierOrganization Science (Dell'Acqua, McFowland, Mollick, Lifshitz-Assaf, Kellogg, Rajendran, Krayer, Candelon, Lakhani), 2025
- The rise of artificial intelligence - understanding the AI identity threat at the workplaceElectronic Markets (Mirbabaie, Bruenker, Moellmann Frick, Stieglitz), 2022
- Selling Self-Disruptive TechnologiesHarvard Business School Working Paper 26-050 (Narayandas, Zhang), 2026
See how strategy arrives.
In 30 minutes, we show you how to steer your initiatives instead of merely managing them.