Simple Explanation (ELI5)

Imagine having a toolbox full of strange gadgets: small businesses are trying out a few of them, but they don’t yet know which tool is truly useful for each job, so they experiment without having built the perfect set.

Key Advantages & Current Limits

🟢 Pros: high experimentation shows openness to innovation; 🟢 Pros: over 70% of firms plan to adopt AI in the coming years; 🔴 Cons: lack of technical skills and inadequate infrastructure limit AI effectiveness; 🔴 Cons: the gap between strategic perception and real capability may make investments less fruitful.

Future Outlook & Takeaways

In the next 3-5 years, targeted training programs and public incentives will enable micro‑ and small enterprises to move from isolated experiments to a structured use of AI across the value chain. This will boost productivity, create skilled jobs, and narrow the digital gap between the Andes and coastal regions.

Small businesses and AI: why are they experimenting but not yet...

Article

100%

A research team examined 781 Ecuadorian micro and small enterprises (MSEs) to see how they use artificial intelligence (AI). Using an entropy‑based model, they measured the variety of AI applications across the value chain. The findings reveal that most firms have only partially adopted AI, with high functional entropy in mining and services – a clear sign of ongoing experimentation.

Micro‑enterprises mainly rely on simple tools like chatbots, while small firms experiment with data analytics and automation. Yet a clear gap exists between the strategic perception of AI and its actual operational setup: many firms view AI as strategic but lack the internal capacity to fully exploit it.

Key barriers are limited technical skills, high costs, inadequate infrastructure, and cultural resistance. Still, over 70 % of non‑adopters intend to adopt AI in the future. Regionally, the Andes and the coast are labeled “innovative,” but they show different digital maturity profiles.

The analysis found no significant link between functional AI diversity and perceived strategic value, suggesting that qualitative factors – such as company culture and leadership – play a bigger role. The entropy model can serve as a diagnostic tool for public policies, helping to design tailored support for MSEs in emerging economies.

Bonus

100%

The entropy approach used in the study draws on Shannon's information theory, quantifying the uncertainty of a probability distribution. Here, each value‑chain activity (production, logistics, marketing, etc.) is treated as a “cell” where AI can be present or absent. Entropy H is calculated as H = - Σ p_i log₂(p_i), where p_i represents the fraction of firms using AI in that specific activity. A high entropy value indicates a more uniform adoption distribution, typical of experimental phases where firms test various solutions without yet converging on an optimal model.

The model was enriched with dynamic capability indicators (sensing, seizing, transforming) and strategic perception of AI, measured via Likert scales. Correlation analysis (Pearson r) between functional entropy and dynamic capability scores showed no significant relationships (r ≈ 0.12, p > 0.05), suggesting that mere application diversity is insufficient to generate competitive advantage without proper organizational alignment.

Methodologically, entropy allows the creation of comparable benchmarks across sectors and regions, highlighting innovation “hot‑spots.” However, the cross‑sectional nature of the dataset limits causal inference; longitudinal studies or controlled experiments could explore how learning dynamics and support policies affect entropy evolution over time. Moreover, integrating qualitative data (interviews, focus groups) could deepen understanding of cultural and leadership factors that emerge as critical drivers for transitioning from an experimental to a consolidation phase of AI in MSEs.