Generative AI is becoming a structural collaborator inside startups, allowing small teams to scale capacity and reconfigure roles and workflows; but the shift also creates new dependencies and coordination challenges that reshape how startups create operational, structural, innovation, and market value.

From Prompt To Process: Qualitative Insights On How Genai Use Rewrites Startup Structures

Melanie Schmidt, Sofia Schöbel · Fetched May 23, 2026 · Journal of the Association for Information Systems

openalex descriptive low evidence 7/10 relevance Source PDF

Qualitative interviews with 17 startup employees show that GenAI is being integrated as a structural collaborator, reshaping roles, workflows, capability expectations, and creating leaner work architectures that expand small teams' capacity while introducing new coordination dependencies.

Generative AI (GenAI) is influencing how startups form, operate, and create value. Yet, process-related insights into how GenAI transforms startups are limited. Our study investigates how GenAI reshapes organizational structures and value creation in startups. Using a qualitative approach with 17 expert interviews from employees at startups, we identify how GenAI drives structural recomposition across four domains: shifting roles, AI-embedded workflows, evolving capability expectations, and leaner work architectures. Our results highlight that startups integrate GenAI not as a peripheral tool but as a structural collaborator, enabling small teams to expand capacity while creating new dependencies and coordination logics. Linking these changes to four value dimensions, namely: operational, structural, innovation, and market value, our study conceptualizes AI-augmented orchestration, where human and algorithmic actors jointly configure work and value creation. The findings extend digital transformation theory by showing that GenAI moves organizing from human-driven adaptation toward technology-embedded reconfiguration.

Summary

Main Finding

Generative AI (GenAI) is not merely a productivity tool in startups but becomes a structural collaborator that recomposes organizational architectures. Through four domains of structural change—shifting roles, AI-embedded workflows, evolving capability expectations, and leaner work architectures—GenAI enables small teams to expand capacity and accelerate value creation across operational, structural, innovation, and market dimensions. The authors conceptualize this as "AI-augmented orchestration": human and algorithmic actors jointly configure work and value-creation, shifting digital transformation from human-driven adaptation toward technology-embedded reconfiguration.

Key Points

Four domains of structural recomposition identified:
Shifting roles: role boundaries blur; some tasks migrate to AI, new hybrid roles emerge (e.g., prompt engineering, AI integrators).
AI-embedded workflows: GenAI becomes an embedded actor in core workflows rather than a peripheral tool, enabling iterative, dialogic task execution.
Evolving capability expectations: startups expect different skill mixes (more orchestration, prompt design, model evaluation) and shorter upskilling cycles.
Leaner work architectures: small teams scale output via GenAI, reducing headcount needs for some functions but creating new coordination/dependency costs.
Four value dimensions linked to these structural changes:
- Operational value (efficiency, throughput)
- Structural value (organizational flexibility, leaner architectures)
- Innovation value (faster prototyping, expanded creative search)
- Market value (faster go-to-market, scaled customer interactions)
GenAI introduces new dependencies (on models, vendors, data pipelines) and novel coordination logics (continuous human–AI interaction, feedback loops).
The study reframes digital transformation (DT) theory by showing algorithmic systems take active roles in organizing, extending DT into AI-augmented transformation.

Data & Methods

Research design: exploratory, interpretive qualitative study using inductive thematic analysis.
Sample: 17 semi-structured interviews with startup employees (founders, CTOs, engineers, managers, researchers).
- Startups founded between 2019–2025, size 2–20 employees.
- Industries: software/IT, health/femtech, cleantech, mobility, edtech, entertainment, fintech, building automation.
- Interview duration: 26–64 minutes; total ≈13 hours of recorded material.
Data collection: purposive sampling via LinkedIn, networks, and snowballing; interview guide covered GenAI context, structural/process changes, value consequences, outlook.
Analysis: cross-case thematic coding (Braun & Clarke style); authors report reaching theoretical sufficiency after 17 interviews.
Limitations noted by authors: startup-only context (high fluidity may amplify effects), non-longitudinal design, self-reported data, sample skew (14 male, 3 female), and limited generalizability to larger/mature firms.

Implications for AI Economics

Practical and research implications relevant to AI economics fall into mechanisms, hypotheses, measurements, and policy considerations.

A. Mechanisms and economic consequences - Labor substitution vs. complementarity: - GenAI can substitute routine tasks but complement higher-level orchestration, shifting demand toward hybrid, higher-skill roles (prompt engineering, AI governance). - Net effect on employment depends on task composition and reallocation speed. - Productivity and returns to scale: - Small teams increase output per worker (higher labor productivity); firm-level returns to scale may change if marginal labor cost falls for many tasks. - Could reduce minimum efficient scale for some activities, lowering entry costs and accelerating venture formation—yet vendor/data dependencies may re-concentrate power. - Wage and skill premium dynamics: - Increased demand for AI-orchestration skills may raise wages for scarce talent; routine-task workers may face downward pressure or require reskilling. - Firm boundaries and outsourcing: - Greater reliance on external AI platforms (model APIs, data services) shifts capital from internal human capital to third-party providers, affecting bargaining power and profit shares. - Innovation and market competition: - Faster prototyping and lower iteration costs increase innovation velocity; may intensify competition and quicken market entry, but platform lock-in and data advantages could produce winner-take-most dynamics.

B. Testable hypotheses for empirical economic research - Startups using GenAI will show higher output-per-worker and shorter time-to-market than comparable startups without GenAI, controlling for sector and stage. - GenAI adoption increases employment shares in hybrid orchestration roles and decreases shares in repetitive production roles within 12–24 months. - Greater reliance on externally hosted GenAI correlates with higher vendor concentration and higher supplier bargaining power (measured via contract terms, API usage concentration). - The variance in firm-level productivity gains from GenAI is positively correlated with firms’ ability to design coordination routines (training, feedback loops) rather than raw model access.

C. Suggested empirical measures and data sources - Productivity/throughput: features released per developer, content items per marketer, time-per-task, revenue per employee. - Task allocation: task logs, time diaries, job postings (skill requirements), occupational task databases (e.g., O*NET mapping). - Dependency/coordination metrics: API call volumes, number of external AI vendors used, percentage of workflows mediated by AI. - Wage/compensation: changes in wages by role pre/post GenAI adoption; hiring rates for AI-orchestration roles. - Market effects: time-to-market, customer acquisition cost, churn, number of entrants per market, concentration indices. - Data sources: firm surveys, platform telemetry (with permissions), job-posting analytics, administrative wage data, venture activity datasets.

D. Policy and normative implications - Labor policy and training: need targeted reskilling programs for orchestration and AI-evaluation skills; portable credentials for prompt and model literacy. - Competition and market structure: monitor platform provider concentration and data access asymmetries; consider interoperability and portability requirements to reduce lock-in. - Measurement & statistics: national accounts and productivity measurement should treat large external AI services as a distinct category of intermediate capital; improve attribution methods for AI-driven output. - Governance & risk management: incentivize transparency and auditable feedback loops in startup AI use to mitigate systemic risks from model errors and overreliance.

E. Directions for future AI-economics research - Quantify causal productivity effects of GenAI in startups with quasi-experimental designs (difference-in-differences, synthetic controls). - Dynamic general-equilibrium models of task reallocation incorporating AI as an endogenous technological frontier—study long-run impacts on wages, sectoral composition, and firm-size distribution. - Investigate bargaining and rent distribution between startups and AI platform providers (contracts, pricing), and implications for surplus allocation. - Longitudinal studies on deskilling vs. upskilling trajectories and how organizational routines mediate labor-market outcomes. - Empirically test the "AI-augmented orchestration" framework: which coordination routines maximize value capture and which increase fragility (vendor lock-in, single-point failures).

Concluding note This paper provides qualitative evidence that GenAI fundamentally alters how startups organize work and create value. For AI economics, it signals the need to incorporate algorithmic actors into models of production, labor demand, firm boundaries, and market structure, and to develop empirical measures that capture both productivity gains and newly emerging dependencies.

Assessment

Paper Typedescriptive Evidence Strengthlow — Findings are based on 17 qualitative expert interviews and descriptive analysis; there is no causal identification, no counterfactual or comparative design, and results rely on self-reported accounts and interpretation rather than observable changes in economic outcomes. Methods Rigormedium — Appropriate qualitative method for exploratory theory-building (expert interviews and thematic coding) but limited by a small, non-random sample (n=17), no stated triangulation with quantitative data, and likely selection and recall biases; rigor would be higher with clearer sampling strategy, coding protocol, and external validation. SampleSeventeen expert interviews with employees at startups (roles unspecified in summary); purposive/qualitative sample of startup practitioners providing rich, context-specific accounts of GenAI use and organizational change; geographic, sectoral, and firm-size coverage not specified. Themesorg_design human_ai_collab innovation adoption GeneralizabilitySmall, non-representative sample (n=17) limits statistical generalization, Likely selection bias toward early adopters or firms positively disposed to GenAI, Unclear geographic and sectoral coverage — may not generalize across regions or industries, Cross-sectional, self-reported data — limited evidence on longer-term, causal effects, Findings may not apply to larger firms or more regulated industries

Claims (9)

Claim	Direction	Confidence	Outcome	Details
Generative AI (GenAI) is influencing how startups form, operate, and create value. Organizational Efficiency	positive	high	how startups form, operate, and create value (organizational formation and value-creation processes)	n=17 0.18
Process-related insights into how GenAI transforms startups are limited. Governance And Regulation	null_result	high	availability of process-related insights in literature	0.09
Using a qualitative approach with 17 expert interviews from employees at startups. Other	null_result	high	study methodology and sample	n=17 0.3
GenAI drives structural recomposition across four domains: shifting roles, AI-embedded workflows, evolving capability expectations, and leaner work architectures. Task Allocation	mixed	high	structural recomposition across roles, workflows, capability expectations, and work architectures	n=17 0.18
Startups integrate GenAI not as a peripheral tool but as a structural collaborator. Organizational Efficiency	positive	high	degree of integration of GenAI within organizational structure	n=17 0.18
GenAI enables small teams to expand capacity while creating new dependencies and coordination logics. Team Performance	mixed	high	team capacity expansion and emergence of dependencies/coordination requirements	n=17 0.18
The study links GenAI-driven organizational changes to four value dimensions: operational, structural, innovation, and market value. Firm Productivity	positive	high	mapping of organizational changes to operational/structural/innovation/market value dimensions	n=17 0.18
The paper conceptualizes 'AI-augmented orchestration', where human and algorithmic actors jointly configure work and value creation. Organizational Efficiency	positive	high	mode of organizing/work configuration (human-algorithm collaboration)	n=17 0.09
Findings extend digital transformation theory by showing that GenAI moves organizing from human-driven adaptation toward technology-embedded reconfiguration. Organizational Efficiency	positive	high	nature of organizational change (human-driven adaptation vs technology-embedded reconfiguration)	n=17 0.09