Session 05 — Exam Preparation

Theory 01

Strategic Search: Two Models

Gavetti & Rivkin (2007) argue that strategy originates from a two-part search process — one unfolding in the world of cognition (mental models, values, heuristics) and one in the world of action (routines, activities, stocks). Their key insight is that both plasticity (how changeable a firm is) and rationality (how sophisticated the search mechanism is) vary with time.

🧬

Evolutionary Search

Boundedly rational managers. Local search — incremental, trial-and-error change of existing routines and stocks. Limited deliberation.

🎯

Positioning Search

Highly plastic firm. Rational deduction from economic principles. Focus on distinctive, interconnected activities that differ from rivals.

🧩

Case-Based Reasoning

The "middle ground" — analogy and imitation. Applying lessons from past cases without fully deductive logic. Falls between local search and deduction.

📡

Information Sensors

Physical elements of strategy that selectively channel environmental feedback to managers. Data is only meaningful through the lens of representations & values.

Key Tension By the time rational search mechanisms become available (as the industry matures), a firm may have already lost the plasticity needed to act on that intelligence.

Theory 02

The Gavetti & Rivkin Framework

The framework links plasticity of elements to firm age and rationality of search mechanisms to industry maturity.

Fig. 1 — The Assumption Space (adapted from Gavetti & Rivkin, 2007, Fig. 1). The typical trajectory moves from high-plasticity/low-rationality (new industry) toward lower plasticity as rationality rises with industry maturity.

Fig. 2 — Hierarchy of Strategy Elements. Representations and values shape heuristics, which guide activities and stocks. Sensors relay environmental feedback upward through cognitive lenses.

Industry Maturity & Search Mechanisms

Stage	Information Environment	Available Search	Example
1 · Infant / Post-shock	Full structural ignorance; states & priors undefined	Local search only	Lycos 1995 — no model to follow
2 · Intermediate	Cues emerge; winners identifiable	Case-based reasoning (analogy, imitation)	Lycos adopts "media company" analogy from Time Warner
3 · Mature	Stable; cause-and-effect clear	Deductive logic from economic principles	Lycos board uses scale economics to justify "get big fast"

Theory 03

Constantiou et al. — Digital Transformation Theory

Building on Daft & Weick's (1984) model of organizations as interpretation systems, Constantiou et al. (2023) theorize DT as the progressive replacement of humans by digital technologies in the three core organizational processes: scanning, interpretation, and learning.

Fig. 3 — From Human Interpretation Systems to Digital Enactment Systems. In DT, the three processes collapse into a tightly integrated, algorithmically driven loop (adapted from Constantiou et al., 2023, Fig. 2).

The Three Digital Processes

Digital Scanning

Sensors and algorithms replace frontline employees in collecting environmental data. Scanning becomes preprogrammed — more precise but narrower. Senior managers must now practice strategic foresight to define what gets scanned.

Digital Interpretation

ML systems give meaning to data by uncovering statistical patterns autonomously. Interpretations arrive pre-coded, removing shared human sensemaking. Equivocality is resolved ex ante, not socially in real time.

Digital Learning

AI systems act on interpretations autonomously — rule-based or ML-based. Human learning (putting cognitive theories into action) is replaced. Senior managers must pre-imagine the range of possible algorithmic actions.

Theory 04

Digital Enactment Systems

Proposition 3 (Constantiou et al., 2023)

"Digital transformation results in organizations ceasing to be human interpretation systems and becoming digital enactment systems, whereby they not only interpret information about their environments but also digitally enact the environments by creating information nearly autonomously."

In digital enactment, the causal arrow reverses: Instead of the organization's perception of analyzability driving its mode of interaction, the organization's actions (enabled by digital technologies) now shape perceived analyzability. Organizations don't just read the environment — they construct it.

Fig. 4 — Ontological reversal. In traditional organizations, perceived analyzability drives action. In digital enactment systems, algorithmic action constructs the perceived environment.

HFT as the Illustrative Case

High-Frequency Trading organizations are the paper's central example, progressing through three historical periods:

Analog Trading (1900–1950) — phones & telex; human traders; undirected viewing mode

IT-Enabled Trading (1950–2010) — OMS, EMS, DSS; conditioned/discovering mode; humans still interpret

HFT / Digital Enactment (2010–) — algorithms scan, interpret & learn in nanoseconds; senior managers only set strategic parameters

Implications for Strategy & Human Roles

📉

Frontline Displacement

Traders, analysts, data collectors replaced by algorithms. Born-digital firms trend toward flat, thin-payroll structures.

🛠

New Professionals

Data scientists and ML developers emerge to build & maintain systems. Human expertise moves from frontline to backroom.

🧭

New Senior Manager Role

Senior managers shift to: data vigilance, scenario-based strategizing, leading algorithm developers, digital experimentation.

⚠️

Residual Equivocality

Managers focus on information that doesn't enter the organization. When things go wrong, the only option is pulling the plug.

Theory 05

AI & Decision Redistribution

Constantiou, Joshi & Stelmaszak (2024) introduce the concept of decision redistribution: AI does not simply eliminate human decision-making — it redistributes it across three interconnected facets.

Fig. 5 — Integrative Framework of AI and Organizational Decision Making (Constantiou et al., 2024). Decision redistribution is the central dynamic linking all three facets.

Decision Redistribution — Three Scenarios

A
AI in → decisions about AI rise. Zillow had to hire data scientists and make complex decisions about data sourcing after deploying Zestimate.
B
Decisions about AI shape how AI is used. Algorithm design choices constrain future decision-making with AI (e.g., which data Zestimate trained on).
C
Leaving implications to AI = high risk. Zillow's algorithm continued buying houses as market conditions changed — no human oversight corrected the course.

Decision Retention Choosing not to implement AI is also a decision. Organizations may rationally retain human decision-making when they cannot account for the full redistribution AI would cause.

Case Application

Novo Nordisk × OpenAI — Analytical View

Your synopsis applies the session's frameworks to Novo Nordisk's April 2026 partnership with OpenAI. Here's a structured summary of the key analytical moves:

Framework	Application to Novo Nordisk	Key Tension
Gavetti & Rivkin	Partnership appears evolutionary (efficiency gains) but has a latent positioning dimension if AI accelerates drug discovery in ways rivals cannot match	Can they convert efficiency gains into a genuinely different competitive activity before rivals catch up?
Digital Scanning/Interpretation/Learning	AI aggregates genomic datasets, clinical trial data, supply chain metrics — replacing fragmented human data work. OpenAI upskilling = managing digital learning	Prop. 1b: constrains ability to capture unexpected signals (e.g. pandemic-like shocks)
Decision Redistribution	Decisions about data sourcing & algorithm design now critical; scientists shift from doing to validating AI outputs	Algorithmic bias in clinical datasets could systematically exclude patient subgroups
AI as Organizing Capability	Connectivity across siloed R&D, manufacturing, supply chain. Codependence: scientists train models; models extend analytical reach	Emergence: must remain open to unexpected drug-target findings rather than constraining AI to predefined tasks

Zillow Parallel Like Zillow, Novo Nordisk's AI models will be trained on historical data. If exogenous shocks (e.g. new disease variants, supply disruptions) render that history unrepresentative, the models may fail in exactly the ways the Zestimate did — with much higher stakes.

Drug discovery Supply chain AI Decision redistribution Algorithmic bias Positioning vs. evolutionary AI factory Data sovereignty Workforce upskilling

Exam Preparation

Likely Oral Exam Questions

Click each question to reveal the key points examiners expect you to address.

Core What are the two models of strategic search in Gavetti & Rivkin, and what assumptions differentiate them? ▶
- Evolutionary: Limited plasticity, boundedly rational managers, local search on routines and stocks. Trial-and-error, incremental change.
- Positioning: High plasticity, deductive logic applied to distinctive interactive activities. Highly rational top-down formulation.
- Key differentiators: plasticity (how changeable elements are) and rationality (how deliberate the search mechanism is).
- Mention the "middle ground" — Gavetti & Rivkin aim for behavioral plausibility between these extremes, adding case-based reasoning, representations, values, and sensors.
Core What is the fundamental tension Gavetti & Rivkin identify regarding time, plasticity and rationality? ▶
- Plasticity decreases as a firm ages (routines solidify, commitments accumulate).
- Rationality of available search mechanisms increases as an industry matures (more information, identifiable winners, deducible principles).
- The tension: By the time rational search becomes available, the firm may have already ossified and lost the plasticity to act on its insights.
- Lycos example: it managed to shift from a technology to a media representation and act on that insight before losing plasticity — but barely.
Apply How does Constantiou et al. define digital transformation, and how does it differ from prior definitions? ▶
- DT = progressive replacement of humans by digital technologies in performing the fundamental activities of scanning, interpretation, and learning.
- Prior definitions focus on business model change, value propositions, or IT-enabled organizational change — they do not specify the nature of transformation.
- Constantiou et al. place digital technologies at the center and focus on the reconfiguration of the organizational "central nervous system."
- Unique contribution: theorizes DT as ending in a qualitatively different organization — not just the same organization with better tools.
Core Explain the concept of "digital enactment" and the ontological reversal it represents. ▶
- In digital enactment, organizations do not merely interpret their environment — they construct it through information produced nearly autonomously by algorithms.
- Ontological reversal: Traditionally, perceived analyzability causes organizational action mode. In DES, it is reversed — digital actions (algorithmic enactment) shape what the organization perceives as analyzable.
- HFT example: algorithms don't just react to market prices — they actively generate price movements through high-speed trading, thereby constructing the market environment for everyone.
- Flash crash (2010): no organization intended to crash markets; yet collectively their DES created the conditions for it.
Apply What are Propositions 1a and 1b in Constantiou et al., and why do they create a strategic paradox? ▶
- Prop. 1a: Replacing humans with digital technologies may increase efficiency and precision in scanning, interpretation and learning.
- Prop. 1b: The same replacement may constrain the organization's ability to function as a human interpretation system.
- Paradox: efficiency gains come at the cost of adaptability. Digital scanning captures only codifiable information — non-codifiable signals (e.g. weak signals of disruption) are systematically missed.
- HFT: algorithms are faster but cannot interpret an exogenous economic shock as anything other than a trading opportunity.
Apply How does the Zillow Zestimate case illustrate the concept of decision redistribution? ▶
- Zillow deployed AI (Zestimate) to make decisions with AI — pricing houses automatically.
- Decisions about AI (data sourcing, model design) were not adequately redistributed: engineers decided to train on historical stable-market data, but did not account for pandemic-driven volatility.
- Implications of AI decisions were also left to the algorithm — no sufficient human oversight corrected the course when conditions changed.
- Result: $800M loss, 2,000 layoffs. Illustrates why all three facets of the framework must be considered simultaneously.
Synthesis Analyze Novo Nordisk's OpenAI partnership using Gavetti & Rivkin's framework. Is this evolutionary or positioning strategy? ▶
- Evolutionary dimension: Stated focus is on improving existing processes (manufacturing efficiency, supply chain, corporate operations) — incremental enhancement of current activities.
- Positioning dimension (latent): If AI accelerates drug discovery pipelines in ways rivals using traditional timelines cannot match, Novo Nordisk would be doing genuinely different activities — a positioning-type competitive advantage.
- Key question: does the AI factory become a distinctive, hard-to-imitate activity bundle, or is it a commodity tool available to all pharma firms?
- Bring in G&R's tension: Novo Nordisk is an established firm (low plasticity) entering new AI industry territory (low rationality of search mechanisms available). Matches Proposition 3A: must resist routinized deductive search and allow local/exploratory search in the AI space.
Synthesis How does digital transformation change the role of senior managers, according to Constantiou et al.? ▶
- From: Receiving and synthesizing employee interpretations → interpreting information → deciding strategy.
- To: Three new roles — (1) Data vigilance: overseeing information strategy and production; (2) Digital-first strategizing: conceptualizing strategy through digital technologies rather than using digital to support business; (3) Leading algorithm developers: deciding which algorithms to deploy, retire, what data sources to use.
- Also: expertise in scenario-based strategizing (to intervene when algorithms fail), seeing data in interactions (social scientist mindset), and digital experimentation (distinguishing firm-caused vs. organic customer behavior).
- Sovereignty paradox: senior managers retain strategic authority but are increasingly dependent on — and constrained by — autonomous digital technologies.
Core What is "tight coupling" in the context of digital enactment, and why is it strategically significant? ▶
- When all activities underpinning scanning, interpretation and learning are performed by an integrated ensemble of algorithms, the three processes collapse into one — they cannot be separated, modified individually, or intervened upon by managers in real time.
- Strategic significance: a minor error in one part (e.g. not scanning a particular environmental dimension) cascades through interpretation and learning — amplified, with limited ability to correct mid-stream.
- Benefits: efficiency, speed, reduced errors in well-understood domains.
- Risks: loss of human sovereignty; inability to adapt to novel or unpredictable conditions; the only recourse is the "kill switch" (pulling the plug).
Synthesis Connect Gavetti & Rivkin's concept of "information sensors" to Constantiou et al.'s "digital scanning." What is gained and lost? ▶
- G&R sensors: Physical elements of strategy that selectively channel environmental feedback to managers. Shaped by representations and values — data only becomes meaningful through cognitive lenses.
- Digital scanning: Sensors are algorithmically automated — faster, broader, more precise — but now the "lens" is pre-coded in algorithms rather than enacted by human cognition.
- Gained: Scale, speed, consistency; ability to detect patterns humans would miss.
- Lost: The human capacity to notice unexpected signals that fall outside predefined scanning parameters; the ability to reinterpret the same data through new cognitive frameworks (like Lycos shifting from technology to media lens).
- Link this to digital transformation challenging managerial cognition — the what to scan question becomes one of the most critical strategic decisions a senior manager makes.