The Second Leap

ABSTRACT

This opening paper sets out the thesis of the series and discloses how it will be tested. The thesis: artificial and organizational intelligence fail at an identical architectural problem. Specialized capabilities do not become general intelligence merely by existing. They must be connected through mechanisms that maintain context, relevance, feedback and continuity over time.

The paper grounds the thesis in contributions to AGI-26 (Cazzani; Botnick et al.; Orban; Laird et al.) and situates it within organizational research (David 1990; March & Simon 1958; Nelson & Winter 1982; Weick & Roberts 1993). From this follows the integration condition. LEAD21, the author's operating system for leadership, is then tested against it — with an open outcome. Result: the four functions are built into the system and proven in practice for human coordination. There is no evidence of an effect on AI adoption. For that, the paper states a falsifiable prediction.

Keywords: Artificial General Intelligence · Leadership Architecture · Leadership Operating System · LEAD21 · Coordination · Productivity Paradox · Human Coordination at Scale

Suggested citation: Huesch, G. (2026). The Second Leap: Why leadership is the architectural condition of AI adoption. Working Paper Series 2026, Series 1 "The Superintelligent Organization", WP 0. Global Leadership Consultancy.

THE THESIS OF THE SERIES

AGI gives every organization a superbrain. It gives it no nervous system.

Artificial and organizational intelligence fail at an identical architectural problem: specialized capabilities do not become general intelligence merely by existing. They must be connected through mechanisms that maintain context, relevance, feedback and continuity over time.

LEAD21 is built as such an operating system. This series tests that claim condition by condition — including where it is not substantiated.

PART 1 · THE FINDING

1 THE QUESTION

The first leap was the machine that computes. The second is the organization that can think with it. The first has been researched for years at great expense. The second hardly at all. Yet it is the second that decides what a company actually gets out of artificial intelligence.

The question of this paper is therefore not how capable AI systems will become. It is: under what condition does purchased intelligence become organizational intelligence?

The answer is an architectural statement. It makes leadership the precondition of the technology and not its background music.

That the question presses in practice is shown by transformation research: 88% of transformations miss their goals (Bain & Company, February 2024). Whoever introduces technology without changing the leadership architecture gets faster answers to unchanged questions and takes that for progress.

2 THE DYNAMO IN THE OLD FACTORY

When factories were electrified, many operators at first replaced only the central steam engine with a large electric motor. Line shafts, belts and the layout of the machines stayed as they were — steam power could only be distributed mechanically, and the hall had been built around that fact. The productivity gain did not materialize. It came only when individual drives made it possible to rebuild the factory around the flow of work. Roughly four decades lay between the first power station and the measurable surge (David 1990).

Paul David transferred this finding to the computer in 1990. Brynjolfsson, Rock and Syverson (2021) generalized it: the return on a general purpose technology hinges on complementary, hard-to-see investments in processes, skills and organization. As long as these investments are under way and are not recorded as capital, the technology is systematically underestimated.

Connecting a language model to an unchanged decision process corresponds to the electric motor on the old line shaft. The power source has been swapped. The floor plan stands.

What is missing is not a bigger brain but a nervous system: signal pathways that connect specialized capabilities. Speaking of a nervous system does not claim that organizations are living beings. It names a function — precisely the one missing between the capabilities.

3 WHAT AGI-26 SHOWS

It is notable that AGI research itself treats general intelligence as an integration problem. Four contributions make this visible.

Specialization is not generality. Dario Cazzani proposes a measure of pragmatic intelligence that captures efficient entropy reduction across different contexts: I_cd = (−ΔS / W) × H(C) — efficiency times breadth. A dedicated term for context breadth prevents narrow specialists from scoring high. His conference poster carries the sentence that decides the matter:

"Efficiency alone isn't pragmatic intelligence. Demonstrated breadth × efficiency is."

Beside it, the image he chooses: the crystal problem. A crystal reduces disorder with perfect efficiency — in exactly one context. H(C) = 0, therefore I_cd = 0. Cazzani's conclusion: "Breadth isn't a bonus. It's a gate." Across four test domains the most efficient agent ranks only fourth on the overall measure.

The image is not a compliment. A crystal is perfect and incapable of learning. That is exactly the point. Transferred: the most efficient unit in your company may well be the least intelligent one — not despite its efficiency, but because that efficiency hangs on a single context.

Continuity is an architectural achievement. Botnick, Hammer, Isaev, Goertzel and Crawford describe, with OmegaClaw, an agentic architecture for continual operation under bounded resources. Their decisive sentence is one about jurisdiction: the language model coordinates memory access, tool use and formal reasoning — "rather than serving as the sole reasoner". What carries over time is the loop, not the model. The architecture distinguishes five memory tiers of differing persistence: working state, episodic trace, semantic memory, persistent symbolic structure, durable artifacts.

Coordination becomes the bottleneck. David Orban models coordination as an engineering problem. If AI drives decision speed beyond human verification capacity, inaction can become the best option for rational actors. He calls this the Freezing Equilibrium. What is meant: when verifying a recommendation costs more than its expected benefit, both stop — the verifying and the acting. His model predicts sharp transitions between self-healing and self-destabilizing institutional states.

Metacognition is a component in its own right. Laird, Lebiere, Rosenbloom and Stocco — the originators of the Common Model of Cognition — treat metacognition as a component of the architecture, not as a by-product of capability. A system that does not have access to its own state does not have access to its correction either.

These findings stand in a longer line. Newell (1990) called for unified theories of cognition because isolated partial theories do not explain behavior. Kotseruba and Tsotsos (2020) show, surveying forty years of cognitive architectures, that perception, attention, memory, learning and action selection carry weight only in interplay.

4 WHAT ORGANIZATIONAL RESEARCH HAS SAID FOR DECADES

The same matter, under other names, considerably older.

March and Simon (1958) describe organizations as a response to bounded rationality: no one can survey the whole, which is why programs and routines connect the parts. Nelson and Winter (1982) understand routines as the memory of the organization — what does not reside in routines leaves with the people. Weick and Roberts (1993) show, using aircraft carriers, that collective capacity for judgment arises from the way actions are related to one another, not from the brilliance of individuals.

From both literatures, four functions can be derived that a connecting mechanism must deliver.

Transferring cognitive architectures to organizations is a structural analogy. It asserts sameness of function, not of material. An analogy proves nothing. It generates hypotheses that must be testable on organizations. That is exactly what this series demands of itself.

5 ONE EXAMPLE, WORKED THROUGH

Illustrative example, not a case report.

An industrial company introduces an AI-supported forecasting system in sales. Forecast quality rises measurably. Yet little changes in the result. Four places explain why.

Context. Sales and production planning understand "order" to mean different things — in one case order intake, in the other invoicing. Both figures are correct. They describe different things. The forecast is precise about a quantity on which there is no agreement.

Relevance. It is not settled who decides when forecast and plan contradict each other. The contradiction belongs to no one. It is left lying, and both sides carry on with their own figure.

Feedback. After the quarter no one checks which forecasts were wrong and why. The system learns from data. The organization learns from nothing.

Continuity. When the project manager leaves, no one knows any longer which assumptions are built into the model. The state was in a head, not in a routine.

The system was intelligent. The organization did not become so.

6 HOLD ON

Objection 1: Won't AGI simply solve the coordination problem along the way?

The AGI-26 contributions argue against it. Faster decisions beyond verification capacity can make organizations freeze rather than coordinate them (Orban). And even coordination solved technically does not answer who is responsible for a decision and who follows it. That remains a question of the institution (Selznick 1957).

The advantage: whoever installs the four functions is not waiting for the next model generation. They already work for the coordination of people — immediately. The work is not wasted in any case.

Objection 2: Isn't this just change management?

Change management accompanies the introduction of a tool. The condition concerns the standing architecture that every tool meets: how matters are discussed, decided, delegated and followed up. A project ends. An operating system keeps running.

The advantage: a project has to be set up anew for every technology. An architecture carries the next adoption — and the one after that.

7 THE CONDITION

Condition 0 · Integration An organization gains only as much general intelligence from artificial intelligence as its leadership architecture maintains context, relevance, feedback and continuity between its specialized capabilities.

The architecture precedes the technology. Where it is missing, AI scales the organization's existing degree of maturity — including its immaturity.

PART 2 · WHAT LEAD21 ANSWERS

LEAD21 is a system of 21 methods that the author developed over more than 25 years for the coordination of leadership in transformations, strong growth and post-merger integrations. It originated before the AGI debate and was not built for it. The test is therefore only this: whether its instruments fulfil the four functions of Condition 0.

CRITICAL APPRAISAL

For each of the four functions LEAD21 has a named instrument with a described mechanism. That is more than a stance and less than a proof. Three limitations belong with it.

First, the mapping comes from the developer of the system and has not been independently verified. The author has a commercial interest in its outcome.

Second, the available evidence relates to human collaboration, not to AI adoption. One survey is scientifically accompanied and measured: a one-year study with 567 leaders of a European industrial group, 78% response rate, led by Prof. Dr. Beate Hanson and Prof. Dr. Mohit Bhandari (McMaster University); it showed significant improvement in 8 of 9 dimensions of the transformation and performance climate. The remaining surveys listed at gerald-huesch.com/evidence-en are expectation surveys with self-report — those are not measurements, and this series does not present them as such.

Third, no one has yet examined whether organizations using LEAD21 bring AI applications into routine operation faster or better. This part of the claim is a hypothesis, not a result.

WHAT THE SERIES GIVES BACK TO LEAD21

The test exposes two gaps the system can close.

Feedback captures behavior but not stored judgments. When the situation changes, the behavior is adjusted — the judgment once formed remains untouched. WP 3 works this out and proposes a metric for it: Correction Latency, the time between counter-evidence and a changed decision.

Continuity is substantiated for the duration of the programme, not beyond it. How long the mechanisms carry after it ends has never been surveyed. WP 1 takes this up.

FALSIFIABLE PREDICTION

Prediction 0. Organizations that have installed a shared meeting, decision and delegation standard before adopting AI bring an AI use case into routine operation faster and show a shorter Correction Latency than comparable organizations without such a standard.

The prediction is falsifiable. If a clean comparison shows no difference, the integration condition is wrong in this form — or LEAD21 does not fulfil it. Either would be a result.

PART 3 · ABOUT GERALD HUESCH

The question of this paper arises only for someone who has seen both sides: the organization that adopts something, and the one that fails at it. More than 30,000 executives in over 30 countries and 300+ companies have worked with Gerald Huesch — from NASA to Rolls-Royce, with a combined responsibility volume of over €550 billion.

Systems shape behaviour. Through the selection of the right instruments, mechanics and leadership codes — derived from values and leadership guidelines — the behaviour of everyone involved is synchronized and aligned. Not through motivation, but through a system that produces the right behaviour when applied. Sales has Salesforce. Finance has SAP. Leadership has LEAD21.

About the author. Gerald Huesch is the creator of the leadership operating system category – "one of the most professional and creative minds in the domain of leadership and difficult change projects" (Boston Business School). With LEAD21 he built the first operating system for leadership – 21 evidence-based methods, the comprehensive solution and the missing link for complete transformation, extreme growth and high-volume post-merger integration: a unique leadership system for human coordination at scale, in which alignment and culture emerge as a by-product. He has trained more than 30,000 executives in over 30 countries (NASA, Rolls-Royce, Airbus, Omnicom and others) and delivers executive leadership training, leadership development and executive mentoring for executives in Germany, the Americas, Asia Pacific and the UAE. Participants from 300+ companies report +87% clarity, +58% efficiency, +50% performance, +46% alignment and −50% negative stress after LEAD21 (participant-reported); a one-year study with 567 leaders (Prof. Beate Hanson & Prof. Mohit Bhandari, McMaster University) showed improvement in 8 of 9 climate dimensions. Dr. Charles Pellerin (NASA) on LEAD21: "Unique in the world, and a capability that nobody else has." gerald-huesch.com/about-en

PART 4 · OUTLOOK

STATUS OF THE SERIES

WHAT COMES NEXT

WP 1 · Organizational Sentience. The capacity of an organization to register and report its own state. Without it, it cannot correct that state. The paper makes the concept testable and presents a test that is cheap to run and hard to pass.

OPEN QUESTIONS FROM THIS PAPER

1. Can Correction Latency be surveyed in running organizations without altering the behavior it measures?

2. Does the integration condition also hold where AI does not prepare decisions but only executes them?

3. Which of the four functions tears first in practice? Experience points to relevance — this is not substantiated.

NOTES

On method. Conception, argument and responsibility rest with the author. An AI system (Claude, Anthropic) was used as a working instrument for research, structuring, the transposition of the AGI-26 contributions and the bilingual drafting. The author instructed the system to assess the relevance and importance of LEAD21 for AGI research critically and with the greatest possible objectivity, including against his own interest. Disclosed for the same reason this series argues for legibility as a leadership principle: whoever demands auditability of others establishes it first for themselves. Readers are expressly invited to evaluate the contribution of LEAD21 to AGI critically, with maximum objectivity of their own, and to draw their own conclusions.

Declaration of interest. The author is the developer and owner of LEAD21 and has a commercial interest in its dissemination. For that reason every statement about LEAD21 in this series is assigned a degree of coverage and an evidence status.

On sources. The author evaluated the posters he documented at AGI-26 as well as titles, abstracts and the proceedings tables of contents — not the full texts. All statements are paraphrased; verbatim quotations appear in quotation marks and come from the posters. The transfer to organizations and to LEAD21 is an analogy by the author and not a statement by the researchers cited. Where a proceedings entry could not be verified first-hand, this is noted.

REFERENCES

AGI-26 (proceedings and conference posters)

Botnick, M., Hammer, P., Isaev, P., Goertzel, B., & Crawford, K. (2026). OmegaClaw: A continually operating agentic architecture under bounded resources. In M. Iklé, A. Franz, A. Kemendo & H. Lowy (Eds.), Artificial General Intelligence. AGI 2026, Proceedings, Part I (LNCS 16854, pp. 338– 353). Springer. https://doi.org/10.1007/978-3-032-33010-9_22 — Poster documented first-hand; page range and DOI from a secondary source.

Cazzani, D. (2026). Efficient entropy reduction as a substrate-independent measure of pragmatic intelligence. In Iklé et al. (Eds.), AGI 2026, Part I (LNCS 16854, pp. 154–171). Springer. https://doi.org/10.1007/978-3-032-33010-9_10

Laird, J., Lebiere, C., Rosenbloom, P., & Stocco, A. (2026). Unified, comprehensive metacognition within the Common Model of Cognition. In Iklé et al. (Eds.), AGI 2026, Part II (LNCS 16855, pp. 1– 15). Springer.

Orban, D. (2026). Civilizational metamaterials: Engineering coordination under capability gradients and structural turbulence. In Iklé et al. (Eds.), AGI 2026, Part II (LNCS 16855, pp. 118–136). Springer.

https://doi.org/10.1007/978-3-032-33195-3_10

Organizational and leadership research

Argyris, C., & Schön, D. A. (1978). Organizational learning: A theory of action perspective. Addison-Wesley.

Bain & Company (2024). Press release, February 2024: 88% of transformations miss their goals.

Brynjolfsson, E., Rock, D., & Syverson, C. (2021). The productivity J-curve: How intangibles complement general purpose technologies. American Economic Journal: Macroeconomics, 13(1), 333–372.

David, P. A. (1990). The dynamo and the computer: An historical perspective on the modern productivity paradox. American Economic Review, 80(2), 355–361.

March, J. G., & Simon, H. A. (1958). Organizations. Wiley.

Nelson, R. R., & Winter, S. G. (1982). An evolutionary theory of economic change. Harvard University Press.

Ocasio, W. (1997). Towards an attention-based view of the firm. Strategic Management Journal, 18(S1), 187–206.

Selznick, P. (1957). Leadership in administration: A sociological interpretation. Row, Peterson.

Simon, H. A. (1971). Designing organizations for an information-rich world. In M. Greenberger (Ed.), Computers, communications, and the public interest (pp. 37–72). Johns Hopkins Press.

Walsh, J. P., & Ungson, G. R. (1991). Organizational memory. Academy of Management Review, 16(1), 57–91.

Weick, K. E. (1995). Sensemaking in organizations. Sage.

Weick, K. E., & Roberts, K. H. (1993). Collective mind in organizations: Heedful interrelating on flight decks. Administrative Science Quarterly, 38(3), 357–381.

Cognitive science

Kotseruba, I., & Tsotsos, J. K. (2020). 40 years of cognitive architectures: Core cognitive abilities and practical applications. Artificial Intelligence Review, 53(1), 17–94.

Newell, A. (1990). Unified theories of cognition. Harvard University Press.