What AI
is doing to your
organization
Three research lenses that measure not just how much your people use AI, but what kind of organization AI is turning you into.
A note on the data shown in this document
The empirical signals presented across these lenses come from an initial survey of 216 design practitioners across 43 countries, not yet from a cross-sector organizational sample. Design practice was chosen as a first population because it sits at the intersection of human-centered work, digital tools, and creative judgment, making AI's effects especially visible. The patterns found are used here as directional evidence, proxy signals that the organizational phenomena we are measuring are real and surveyable. The cross-sector organizational survey (currently in design) will generate the sector-specific, role-specific, and organization-specific data that will make these lenses precise instruments rather than directional ones. This document shows the framework's analytic potential, not its final output.
Interaction
Intelligence
HCI research shows that how people interact with AI, not whether they use it, predicts the quality of outcomes. This lens measures whether your organization is developing genuine human-AI collaboration capacity.
What it measures
Whether people approach AI directively (give commands, expect execution), iteratively (adjust prompts based on responses), or as thinking partners (consider AI input and sometimes change direction).
Why it matters
Directive interaction extracts what AI already knows. Thinking-partner interaction generates outcomes neither human nor AI would produce alone. The difference is not training, it is interaction architecture.
The organizational signal
52% of fully integrated practitioners approach AI as a thinking partner. Only 14% of routine users do. Interaction quality scales with integration depth, but integration without interaction intelligence produces volume, not value.
Three organizational interaction profiles
Usage level: Mostly Routine
Risk: Plateaus at productivity; competitors developing thinking-partner capacity will outpace
Usage level: Mixed Routine + Explorer
Risk: Iteration optimizes the known; it rarely discovers the unknown
Usage level: Mostly Integrated
Strength: Compound intelligence returns; genuine novelty
What this concept enables you to decide
| Decision area | Without this | With this | Strategic implication |
|---|---|---|---|
| AI Training | We train people to use AI tools effectively | We know whether training is changing interaction quality or just tool familiarity | Tool training produces iterative users. Thinking-partner development requires different interventions: reflection protocols, structured AI dialogue practice, permission to be uncertain. |
| Value Measurement | We measure AI output volume and time saved | We can distinguish efficiency value from intelligence value in AI-assisted work | Organizations measuring only efficiency will systematically underinvest in the interaction quality that produces their highest-value AI outputs. |
The question this concept answers for the C-suite
Are we getting productivity from AI, or intelligence from AI? The data shows a 4× difference in thinking-partner interaction between routine and integrated users. The gap is not talent, it is interaction architecture. Organizations that design for thinking-partner interaction get qualitatively different outputs from the same tools as organizations that design for directive efficiency.
What it measures
Whether people can accurately predict AI output quality, recognize when AI is shaping their thinking, and evaluate AI contributions with calibrated judgment rather than automatic acceptance or rejection.
Why it matters
Metacognition is the difference between using AI and understanding how AI is affecting your judgment. Organizations with low metacognitive capacity adopt AI outputs without evaluating them, and can't tell when AI is wrong, biased, or homogenizing their outputs.
The organizational signal
71% of practitioners now spend more time evaluating AI outputs than producing from scratch. But only 40% of multi-tool users can accurately predict AI output quality. High usage without high metacognition is a governance risk, not a capability.
What this concept enables you to decide
| Decision area | Without this | With this | Strategic implication |
|---|---|---|---|
| Quality Assurance | We review AI outputs before publication or use | We know whether reviewers have the metacognitive capacity to catch AI errors, bias, and homogenization | Review processes designed for human error do not catch AI error. Metacognitive capacity, the ability to evaluate AI judgment, is a skill that must be explicitly developed, not assumed from general competence. |
| Risk Management | We have AI use policies | We know whether our people can detect when AI is wrong, biased, or producing convergent outputs | The organizations most at risk from AI errors are those with high adoption and low metacognition, where AI outputs move through workflows without adequate human evaluation. The data shows this combination is common. |
The question this concept answers for the C-suite
Do your people know what AI is doing to their thinking? 71% of practitioners are now evaluating AI outputs more than generating their own. That shift transfers significant cognitive authority to AI. Whether that authority is well-placed depends on metacognitive capacity, which only 31% of the sample demonstrates. The remaining 69% are accepting or rejecting AI outputs without calibrated judgment for doing so.
What it measures
The diversity of AI tools used and the sophistication of use across tool types, from single-tool dependence through multi-tool fluency to integrated AI ecosystems that shape entire workflows.
Why it matters
Single-tool dependence creates cognitive lock-in, people learn to think in the vocabulary of one AI system. Multi-tool fluency builds comparative judgment and broader metacognitive capacity. The tool ecosystem shapes how people think, not just what they produce.
The organizational signal
22% of practitioners use only one AI tool. 17% use four or more. The 4+ tool group shows the highest metacognitive capacity and the strongest thinking-partner interaction approach. Tool diversity and cognitive sophistication correlate.
What this concept enables you to decide
| Decision area | Without this | With this | Strategic implication |
|---|---|---|---|
| Tool Strategy | We standardize on approved AI tools for security and efficiency | We know whether standardization is creating cognitive lock-in that reduces metacognitive capacity | Standardization has real security and governance benefits. But single-tool mandates may produce the lowest metacognitive capacity populations in the organization. The tradeoff must be explicit, not accidental. |
| Talent Development | We train people on the tools we use | We know whether tool diversity is a learning investment or an unmanaged risk | Multi-tool fluency builds comparative judgment that single-tool training cannot. Organizations that deliberately develop tool diversity as a capability are investing in metacognitive infrastructure, not just tooling. |
The question this concept answers for the C-suite
Is your AI toolkit expanding your people's thinking, or narrowing it? The data shows that single-tool users have the lowest metacognitive capacity and the most directive interaction approach. Multi-tool users have the highest. This is not about using more tools for their own sake, it is about whether the tool ecosystem is developing or constraining the human cognitive capacity that makes AI genuinely valuable.
Agency
& Network
STS research shows that technologies are never neutral, they carry scripts that prescribe how people should act, and they restructure the social networks around them. This lens measures whether AI is acting on your organization in ways you haven't decided.
What it measures
How people in your organization characterize AI's role, as a tool they control, an assistant that helps, a collaborator that influences direction, or an unpredictable participant whose influence is unclear.
Why it matters
Agency attribution is not just a perception. It determines accountability. If your people see AI as a tool, they take full responsibility for outputs. If they see it as a collaborator, responsibility is distributed. If they see it as unpredictable, accountability is unclear. Each position has different governance implications.
The organizational signal
56% call AI an "assistant", they contribute but humans decide. Yet daily users attribute collaborator status at 2× the rate of non-daily users (27% vs 14%). Frequency of use is shifting the perceived agency balance, whether organizations have decided that or not.
Three agency positions, and their governance consequences
Accountability: Clear
Risk: Cognitive underuse; AI potential unrealized
Accountability: Nominally clear
Risk: Actual AI influence exceeds acknowledged AI influence
Accountability: Requires new frameworks
Strength: Most accurate reflection of actual AI influence
What this concept enables you to decide
| Decision area | Without this | With this | Strategic implication |
|---|---|---|---|
| Accountability Frameworks | Humans are accountable for all AI-assisted outputs | We know what agency level our people attribute to AI, and whether our accountability frameworks match that level | Governance built for tool-level AI managing collaborator-level AI creates accountability gaps. The question is not whether AI has agency, it is whether your organization has decided what to do about it. |
| Leadership Communication | We communicate that humans remain in control | We know whether that narrative matches how your most integrated users actually experience AI | 27% of daily users already attribute collaborator-level agency to AI. Leadership narratives of full human control ring hollow for this group, and erode trust in the broader AI communication strategy. |
The question this concept answers for the C-suite
Who does your organization think is making decisions, and does your governance reflect the actual agency balance? The data shows that daily AI use shifts agency attribution significantly. Your most integrated employees are already operating with a collaborator model of AI. If your policies, accountability structures, and communication all assume an assistant model, you have a governance gap that is growing with every day of AI use.
What it measures
How people respond when AI produces unexpected output, whether they reject, modify, explore the unexpected direction, or use it as a starting point for something new. Each response reflects a different negotiation with AI's embedded assumptions.
Why it matters
AI systems carry inscribed assumptions about how tasks should be done. When AI produces unexpected output, it is revealing those assumptions. How your people respond determines whether AI's inscribed logic colonizes your organizational practice, or whether human judgment maintains primacy.
The organizational signal
Tool-framers reject unexpected AI output at 35%, they reassert control. Collaborator-framers explore the unexpected at 36%, they negotiate. The inscription response reveals how much your people trust their own judgment relative to AI's embedded logic.
What this concept enables you to decide
| Decision area | Without this | With this | Strategic implication |
|---|---|---|---|
| Professional Judgment | We trust our people's judgment in AI-assisted work | We know whether that judgment is being expressed or suppressed when AI pushes back | Organizations where most people modify AI output to match original direction are organizations where AI is confirming existing thinking. Organizations where people explore unexpected directions are organizations where AI is expanding thinking. Both have value, but they are different strategies. |
| Innovation Culture | We encourage creative use of AI | We know whether "creative use" means exploring AI's unexpected outputs or overriding them | Inscription response is a direct measure of whether AI is expanding or constraining organizational thinking. If 44% of your people modify every unexpected AI output to match original direction, AI is not expanding your thinking, it is being domesticated to confirm it. |
The question this concept answers for the C-suite
When AI surprises your people, do they lead or follow? The inscription response is the most direct measure of the human-AI power dynamic in your organization. 44% modify AI output to match their original direction, a form of control. 25% explore where AI leads, a form of openness. Neither is uniformly right. But most organizations have no data on this distribution, no policy about it, and no deliberate culture around it.
What it measures
Whether AI adoption has changed how work is divided, discussed, or valued among colleagues, clients, and professional communities, and whether those changes were deliberate or emergent.
Why it matters
AI does not just change individual workflows. It changes the social organization of work. When a team member can do in minutes what previously required a specialist, the team's interdependencies change. These changes are often invisible until they create conflict, redundancy, or unexpected accountability gaps.
The organizational signal
71% of fully integrated practitioners report noticeable or significant network restructuring. Only 35% of routine users do. Full AI integration is reorganizing professional relationships at scale, whether organizations have planned for this or not.
What this concept enables you to decide
| Decision area | Without this | With this | Strategic implication |
|---|---|---|---|
| Organizational Design | We deploy AI tools to improve individual productivity | We know whether AI is restructuring team interdependencies faster than organizational design can accommodate | 71% of fully integrated practitioners report significant network change. These are not planned organizational design decisions, they are emergent reconfigurations driven by AI adoption. Leading organizations manage this deliberately; most discover it in retrospect. |
| Talent Retention | We track employee satisfaction with AI tools | We know whether AI-driven network changes are creating role ambiguity, value disputes, or collaboration conflicts | When AI enables an individual to perform tasks previously distributed across a team, the remaining team members' roles become unclear. This is the leading source of AI-related professional anxiety, not job loss, but role dissolution. |
The question this concept answers for the C-suite
Is AI quietly reorganizing your teams, and have you decided what to do about it? 52% of all practitioners in the data report noticeable or significant changes in how work is divided and valued. Among fully integrated users, that rises to 71%. These are not technology changes, they are organizational design changes happening without organizational design decisions. The network reconfiguration lens tells you where in your organization this is happening, how fast, and whether it is creating value or conflict.
Institutional
Change
Organizational theory shows that how institutions adopt technology reveals what they believe technology is for, and determines what it actually becomes. This lens measures whether your organization is leading its AI transformation or discovering it after the fact.
The empirical signals shown throughout this lens come from the first survey, 216 design practitioners across 43 countries, not from a cross-sector organizational survey. That study does not yet exist. This lens was not part of the original survey design; its three concepts (Algorithmic Isomorphism, Ambidexterity, Institutional Logic Shift) have been analytically mapped onto the existing data as directional signals to demonstrate what the organizational lens would reveal if measured directly.
Specifically: design practitioners were asked about their industry, work context, and organizational role, not about their organization's AI strategy, governance, or institutional logic. The percentages shown are proxy readings inferred from adjacent variables, not direct measurements of the organizational concepts. The cross-sector survey described in the Study Strategy tab is what will generate the real organizational-level data.
What it measures
The pressure driving AI adoption in your organization and whether shared AI tools are homogenizing professional judgment, outputs, and competitive positioning across your sector.
Why it matters
When all firms in a sector use the same AI tools trained on the same data, outputs converge. Differentiation, the source of competitive advantage, erodes not through bad strategy but through identical AI-mediated workflows. This is the new market risk that no adoption metric captures.
The organizational signal
Latin American organizations show 38% full AI integration vs. 14% Anglo-Western, driven by different institutional pressures, not different talent. The adoption pattern is not a talent story. It is an institutional pressure story.
What this concept enables you to decide
| Decision area | Without this | With this | Strategic implication |
|---|---|---|---|
| Competitive Strategy | We assume AI adoption gives us an advantage | We know whether AI is converging our outputs toward sector averages | If all firms in your sector use the same AI tools with the same training data, the efficiency gains are table stakes. The real question is whether AI is homogenizing what makes you distinctive, and whether you've decided to let it. |
| Governance Design | We have AI policies aligned with our strategy | We know whether governance is aligned with actual AI integration levels | 75% of fully-integrated financial services practitioners frame AI as an "assistant." Governance designed for assistant-level AI is managing collaborator-level reality. The gap is a liability. |
The question this concept answers for the C-suite
Are you adopting AI strategically, or are you being adopted by it? The difference is between an organization that decides how AI shapes its work and one that discovers, after the fact, that AI has been shaping it without authorization. Algorithmic isomorphism is not a future risk, the data shows it is already operating across sectors and regions.
What it measures
The balance between exploiting AI for efficiency on familiar tasks and exploring AI to develop genuinely new capabilities, at individual, team, and organizational levels simultaneously.
Why it matters
Organizations that only exploit AI achieve short-term efficiency gains and long-term capability stagnation. Organizations that only explore never operationalize insight into competitive practice. The ambidextrous balance, not one or the other, is the organizational AI strategy question.
The organizational signal
67% of government/public sector practitioners use AI only for routine tasks. 57% of academic practitioners are at Explorer level. 77% of large enterprise in-house teams (500+) are at Explorer or Integrated. The explore/exploit balance is determined by organizational context, not individual choice.
What this concept enables you to decide
| Decision area | Without this | With this | Strategic implication |
|---|---|---|---|
| AI Investment | We measure AI ROI through efficiency and output metrics | We can distinguish efficiency ROI from exploration ROI, and know which our organization is actually getting | Pure efficiency measurement creates organizational incentive to over-exploit. Teams that cannot demonstrate exploration value will stop exploring, and the organization's AI capability will plateau at productivity, never reaching genuine innovation. |
| Team Design | We give teams AI tools and freedom to use them | We know which teams are stuck in exploitation and which have team norms enabling genuine exploration | Team norms moderate individual AI use more than tool access does. Changing tools without changing norms produces no behavioral change. The ambidexterity lens tells you which teams need norm intervention, not just tool deployment. |
The question this concept answers for the C-suite
Are you getting both kinds of return on your AI investment, or trading one for the other? 67% of government practitioners are getting only efficiency gains. Their organizations don't know this, because they don't measure it. The ambidexterity lens reveals whether your AI investment is compounding across efficiency and capability, or cannibalizing one for the other.
What it measures
Whether the dominant organizational belief system about AI, efficiency tool vs. learning partner, matches the actual integration level and agency attribution of the people doing the work.
Why it matters
When organizational narrative and policy are designed for a different level of AI than people are actually using, governance fails silently. The gap accumulates risk invisibly, until a decision, an error, or an accountability question surfaces the mismatch.
The organizational signal
55% of Latin American practitioners live a logic gap: their work has been restructured by AI, but organizational language hasn't shifted. Only 9% are logic-aligned, behavior and language match. Asia-Pacific shows the highest alignment at 43%.
What this concept enables you to decide
| Decision area | Without this | With this | Strategic implication |
|---|---|---|---|
| Governance Calibration | We update AI governance annually or when incidents occur | We can measure the gap between governance and practice, and update governance when the gap exceeds a threshold | Annual governance review is calibrated to a policy cycle, not to the actual pace of AI integration. The logic shift metric tells you when your governance is out of sync, before an incident makes that visible. |
| Organizational Identity | We know what kind of company we are | We know whether AI is changing that identity, and whether leadership has decided to lead or follow that change | The logic shift is ultimately an identity question: are we an efficiency organization that uses AI, or a learning organization that thinks with AI? Both are viable strategies. Neither can be sustained accidentally. |
The question this concept answers for the C-suite
Does what you say about AI match what your people actually do with it? The logic shift lens surfaces the gap between organizational narrative and organizational reality, and gives you a measurement system for closing it before it becomes a governance failure. In the data, 51% of Latin American practitioners are living a logic gap. They are managing a collaborator-level transformation with assistant-level frameworks. That gap is growing with every month of AI adoption.
Where is your
organization today?
A unified diagnostic landscape across all nine concepts, read the overview band at a glance in a leadership meeting, then scroll below for the detailed prescription on each concept. Inspired by Keeley & Doblin's Ten Types of Innovation diagnostic structure.
Low use + high language
High use + governed
Low use + low language
Most integrated orgs
Aware, no strategy
Aware + intentional
Unaware conformist
Distinct, unknowing
Detailed concept analysis & prescriptions
Nine concepts · full evidence bands · decision tables · detailed prescriptions
Interaction Intelligence
Directive → Iterative → Thinking Partner
Prescription to close the gap
Design AI workflow protocols that require reflection, not just output. Mandate "AI dialogue review", what did AI suggest that you changed, and why? Reward teams that demonstrate AI-expanded thinking, not just AI-accelerated output.
Metacognitive Capacity
Automatic acceptance → Calibrated judgment
Prescription to close the gap
Introduce structured AI evaluation protocols: before accepting output, employees articulate what they expected and why the output does or does not meet that expectation. Build error-spotting exercises into onboarding. Track and share cases where AI was confidently wrong.
Tool Ecosystem
Single-tool dependence → Multi-tool fluency
Prescription to close the gap
Introduce tool diversity as an explicit learning objective, not just a security risk to manage. Create structured exposure to 2-3 different AI systems for the same task type, to build comparative judgment. Document where different tools produce systematically different outputs on your work.
Agency Attribution
Not a continuum of more-is-better, a question of alignment between how people describe AI and how they actually use it
Prescription to close the gap
Run an agency audit: for each major decision type in the past quarter, ask how much AI shaped the outcome. Map your organization on this 2×2. If you are in the Silent Gap (high integration, low agency language), redesign accountability frameworks before an incident makes the mismatch visible. The Coherent Organization does not deny AI agency, it governs the agency it has already granted.
Inscription Response
Not a continuum, a distribution between two legitimate poles. The advanced position is deliberate choice, not a preferred direction
Prescription to close the gap
The goal is not to shift the distribution left or right, it is to make it intentional. Build a case library: document when following AI's unexpected output produced better outcomes and when overriding it did. Create explicit team protocols for high-stakes decisions: in this context, we default to human judgment; in this context, we explore what AI surfaces. Deliberate distribution is the advanced position, not a particular ratio.
Network Reconfiguration
No change → Minor → Noticeable → Restructured
Prescription to close the gap
Map AI-driven role changes explicitly: which tasks have migrated, which collaborations have dissolved, which new interdependencies have formed. Do this annually. The Coherent Organization does not prevent AI from restructuring teams, it ensures that restructuring is a deliberate organizational design decision, not an emergent accident.
Algorithmic Isomorphism
Not a continuum of adoption, a question of awareness and intentionality. The risk zone is the middle, not the low end
Prescription to close the gap
Step 1, build awareness: audit a sample of AI-assisted outputs against sector benchmarks. Ask: are our analyses, reports, and recommendations becoming more similar to competitors'? Step 2, build intentionality: decide which workflows should converge (efficiency through standardization is fine for commodity tasks) and which must diverge (proprietary data, unique methodology, human judgment layers that competitors cannot replicate). The Coherent Organization does not resist isomorphism everywhere, it manages it strategically.
Ambidexterity
Exploitation only → Balanced explore + exploit
Prescription to close the gap
Introduce an explicit exploration allocation: dedicate a measurable portion of AI use time to genuinely novel tasks, not efficiency optimization. Create team-level exploration norms, psychological safety to use AI for uncertain, open-ended problems. Measure innovation output from AI alongside efficiency output. Reward teams that develop new capabilities, not just faster workflows.
Institutional Logic Shift
Efficiency narrative → Learning narrative → Aligned
Prescription to close the gap
Run an annual logic-behavior audit: compare what leadership says about AI (communications, policy, metrics) against how people actually use it (integration level, agency attribution, network change). When the gap exceeds 20 percentage points on any dimension, trigger a governance review. The Coherent Organization does not update governance on a schedule, it updates governance when the evidence demands it.
What your X-Ray combination means, three organizational profiles
No organization sits at the same point on all nine bars. The combination of positions across all three lenses reveals your organization's AI transformation profile, and the strategic priorities that follow from it.
Who, where,
and how many
A practical sampling strategy grounded in three evidence sources: the first survey's 216 practitioners, a 9,147-connection LinkedIn network, and the 2025 AI adoption landscape across Colombia, the US, and Europe.
Expert Collaborators Sought, Three Lenses + Systems Design
The framework brings three theoretical lenses (HCI, STS, Organizational Theory) into conversation through the practice of systems design. We are actively seeking expert collaborators across all four areas to develop and validate this research.
Academic or industry research background
HCI & emerging technologies
Cross-cultural AI adoption
Critical theory background
Meaning & interpretation
Hermeneutic & speculative design
Innovation & institutions background
Institutional change & ambidexterity
Emerging market research
Systems thinking & design epistemology
CAS & open innovation
Design epistemology & professional identity
The Configurations Lab is seeking expert collaborators across the three lenses and the integrating systems-design practice. The theoretical architecture and sampling strategy are open to refinement. If this framework intersects with your research agenda, we would welcome a conversation about how to develop it together.
Practical Recommendation Matrix
Four scenarios from minimum viable to ideal, evaluated on what each enables analytically and what it requires operationally
Pilot only
Recommended
Ambitious
Full study
Fewer than 10 respondents per company and you cannot distinguish company-level patterns from individual variance, the multilevel design collapses into a flat survey. Fewer than 4 companies per sector and you are comparing individuals who happen to work in different sectors, not sectors themselves. 20 × 15 = 300 is the threshold where both the within-company and between-company analyses become meaningful simultaneously.
The first survey (n=216) was individual-level, practitioners responding as individuals. This survey targets organizational contexts, with multiple respondents per organization. That shift enables a new analytical question: not just what individuals believe about AI, but whether organizations develop coherent collective patterns, and whether those patterns differ by sector, size, and geography.
Three Levels of Measurement
The same instrument generates individual, organizational, and sector-level diagnostics simultaneously, enabling six distinct comparison types from a single survey deployment
Sector Priority, Four Evidence Sources
Each sector rated across three evidence sources: first survey representation, LinkedIn network access, and 2025 AI adoption landscape in target geographies
Digital Products
USA/UK: Global tech corporations, cloud providers, enterprise software
Europe: Digital companies, enterprise software firms (GDPR context)
Services
USA/UK: Global banks, investment firms, financial institutions
Europe: European banks under EU AI Act, highest regulatory pressure
Professional Services
USA/UK: Innovation consultancies, design labs, digital agencies
Europe: European consulting and professional services
Education
USA: Research universities, ed-tech platforms, health networks
Europe: European universities and health institutions
Note: Bureaucratic access barriers, institutional partnerships essential for access
Agricultural Sectors
Geographic Targeting, Why Three Locations Matter
Geography is not a control variable here, it is a theoretical lever. Different institutional environments produce different AI adoption logics. Same sector, different country = isolating institutional effects from professional ones.
High integration, rapid adoption, mimetic pressure dominant
Normative pressure, cautious Explorer pattern, strong governance
GDPR regulatory pressure, coercive isomorphism visible
Structured AI governance emerging, logic shift measurable
Regulatory + normative pressure · highest governance investment
GDPR + EU AI Act pressure, most regulated AI context globally
Client-facing AI integration highest · normative pressure from clients
Design-adjacent, connects to first survey population
Lower priority, similar pattern to USA/UK expected
High social mission, learning logic often dominant
High AI maturity · learning-logic dominant
Lower priority · ethics approval slower in EU
Brazil: Academic and professional networks in São Paulo open access to the second-largest AI market in Latin America, with a mature tech ecosystem (Nubank, Totvs, iFood, Embraer) and a distinct institutional context from Colombia. A Colombia vs. Brazil comparison isolates within-Latin America institutional variation, testing whether the behavioral-linguistic mismatch is a Colombian effect or a broader LatAm phenomenon. This is only possible with the proposed team's combined geographic reach.
Spain: A European Spanish-language comparison the first survey couldn't make. If the mismatch appears in Spanish practitioners, it's a language effect. If it doesn't, it's a Latin American institutional effect. Spain is also under the EU AI Act, a regulatory contrast (LatAm's softer governance vs. EU's binding framework) within the same language group.
Participants Per Company, Why 12–15 Is the Threshold
The minimum number of respondents per organization that makes the multilevel design analytically valid
With fewer than 10 respondents per company, individual variation dominates any organizational signal. One senior director with unusual views can shift the company's average by 15 percentage points. You cannot confidently say "this company has an efficiency logic", only that some people in it do.
At 12–15 respondents, the organizational signal stabilizes. You can disaggregate by role level (individual contributor vs. manager), detect within-company variance on the three lenses, and compare companies to each other with reasonable confidence. This is the minimum for the X-ray to be meaningful at the company level, not just the sector level.
Mix role levels intentionally: 4–5 individual contributors, 4–5 team leads or managers, 2–3 directors or above. This gives you the vertical slice needed to test whether the logic gap (between what leadership says about AI and what practitioners experience) operates within each organization, not just across organizations.
How the Design Enables Triangulation
The value of this sampling strategy is not the total n, it is the intersection of three analytical dimensions that generate insights no single-dimension study can produce
Different Geography
Different Sector
Different Role Levels
Insights only possible with this design, not available from the first survey or any single-dimension study
Recruiting Sequence, Five Steps Per Company
The organizational survey fails not because companies refuse to participate, it fails because the internal distribution breaks down. Each step is designed to prevent a specific failure mode.
The theoretical
architecture
A brief orientation for academic collaborators and reviewers, the theoretical foundations, the epistemological stance, and the research contribution this study makes to the literature.
What kind of professionals is AI generating in organizations, and what kind of AI are professionals generating in return?
This question is deliberately symmetrical. It resists the dominant framing of AI adoption research, which treats AI as an independent variable acting on passive human recipients. Drawing on Actor-Network Theory and Science and Technology Studies, we treat AI and professional practice as co-constitutive: each shapes the other through ongoing negotiation. The survey instrument operationalizes this symmetry across three analytical lenses and nine measurable concepts.
The first empirical study (Rivera & Russi, 2026, under review at Base Diseño e Innovación) established this framework with 216 design practitioners across 43 countries. The current study expands the population to cross-sector organizational contexts, testing whether the framework's findings, particularly the behavioral-linguistic mismatch, generalize beyond a single professional community.
Why Design Studies holds the framework together
The three measurement lenses, HCI, STS, and Organizational Transformation, each have established theoretical traditions and validated instruments. What they lack is an epistemological framework for holding them in productive tension without collapsing one into another. HCI's engineering tradition pulls toward optimization and individual behavior. STS's critical humanities tradition pulls toward meaning and power. Organizational theory pulls toward institutional performance. Treated as additive, these three perspectives produce conflicting claims.
Design Studies provides the integrating epistemological stance through its alignment with Complex Adaptive Systems (CAS) thinking. Specifically: CAS frameworks understand organizations as adaptive nonlinear networks with no global controller, crosscutting hierarchical interactions, and emergent properties not predictable from component analysis alone (Arthur 1999; Teixeira & Forlano 2016). This is precisely the theoretical space where the three lenses must operate, not as parallel measures of independent variables, but as complementary readings of an emergent organizational phenomenon.
Design Studies contributes two specific CAS capacities: the tolerance for irreducible complexity that allows behavioral and meaning data to coexist without forcing resolution, and the systems thinking tradition (Charles L. Owen, Buckminster Fuller, Horst Rittel) that understands problems as nested across scales, individual, organizational, sectoral, each generating emergent properties invisible at adjacent levels. This is why the framework operates simultaneously at three levels of analysis, and why the Coherent Organization is defined as a profile of internal consistency rather than an optimized score on any single dimension.
This CAS inheritance is necessary but not sufficient. Where CAS treats agents as separable units whose interactions produce emergence, the lab's wave-four lenses make a deeper commitment: the agents themselves emerge through intra-action (Barad 2007; Frauenberger 2019). The integrating practice, Entangled Systems Design, extends the systems-thinking tradition into this register. It lets the three lenses be read together not as parallel measurements of pre-given variables, but as complementary readings of an ongoing materialization.
How each concept is grounded and what it adds
The behavioral-linguistic mismatch, and whether it generalizes
The first study's most significant finding was a behavioral-linguistic mismatch: Latin American designers (predominantly Colombian) showed high behavioral AI integration but low agency-attribution language, a gap that narrowed significantly with seniority (40-point gap at junior level, shrinking to approximately 9 points at senior level). This finding was only visible through the three-lens analytical combination; no single lens would have surfaced it.
The cross-sector study's primary empirical question is whether this mismatch generalizes beyond design practice. Three competing hypotheses are possible: the mismatch is a design-culture effect (specific to creative professional identity); a Latin American institutional effect (organizations adopting AI under mimetic pressure without the governance infrastructure to match); or a universal junior-professional effect (inexperienced AI users everywhere lack the vocabulary to accurately describe their relationship with AI). The three-geography, four-sector design is specifically constructed to discriminate between these hypotheses.
The Coherent Organization framework emerges from this analysis as the study's applied contribution: an empirically grounded organizational profile defined not by maximum AI adoption but by the alignment between integration behavior, agency language, and institutional governance, a distinction the existing AI organizational literature does not make.
Simkute et al. (2023) CAILS scale · Vaccaro et al. (2024) CAIMS · Bhargava & Gopal (2022) three-level AI usage model · Luan, Kim & Zhou (2025)
Latour (2005) Reassembling the Social · Akrich (1992) inscription theory · Callon (1986) translation · Akrich & Latour (1992)
DiMaggio & Powell (1983) · March (1991) · Thornton & Ocasio (1999) · Caplan & Boyd (2018) · Wang & Long (2025) · Gutierrez et al. (2025)
Cross (1982) designerly ways of knowing · Frayling (1993) research through design · Ruecker & Radzikowska (2011) humanities visualization · Teixeira & Forlano (2016) CAS and innovation systems · Arthur (1999) complexity economics
Rivera & Russi (2026, under review), Stage 1: 216 design practitioners, 43 countries. Stage 2 (current): cross-sector organizational survey, target n=300, 20 companies, 4 geographies, 4 sectors
Management & innovation journals · Design research journals · Digital Humanities venues · HCI venues (CHI, CSCW) · Industry translation as organizational diagnostic tool · Venue selection to be determined collaboratively