Cost-Constrained Multi-Agent Orchestration for Business Intelligence in Emerging Markets

The crisis

• Small businesses in tier 2/3 cities make high-stakes decisions on intuition analytical tools cost $5K–$50K/year
• India: 63 million MSMEs, the majority with zero access to data intelligence tools
• Africa: small businesses are the economic backbone of entire countries but have no access to market intelligence
• The digital divide is not only internet access it is access to intelligence tools incumbents price at enterprise rates
• The analytical gap compounds over time: businesses without data lose to businesses with data, regardless of product quality

About this research

Business intelligence tooling operates on a pricing model that assumes enterprise budgets. The result is a structural analytical gap: organizations in emerging markets, small enterprises, and tier 2/3 city businesses make decisions on intuition that their better-resourced competitors make on data. This research investigates how domain-constrained specialist agent decomposition can achieve enterprise-grade analytical output at a cost point accessible to resource-limited organizations.

Research question

Can domain-constrained specialist agent decomposition achieve enterprise-grade business intelligence output quality at a cost structure accessible to organizations previously excluded by incumbent pricing?

Methodology

5-agent architecture with domain-scoped specialists coordinated through MCP-based orchestrator on AWS Bedrock; cost benchmarking methodology comparing per-query cost against equivalent monolithic LLM and commercial BI tool baseline; quality evaluation across analytical output dimensions (accuracy, comprehensiveness, actionability); deployment and validation across 4 geographic markets; ablation study isolating cost reduction contributions from agent scoping, context limitation, and caching layers.

Key findings

• 90% cost reduction vs monolithic LLM approach validated across 4 geographic regions. (Full findings in progress.)

References

• World Bank (2024) MSME access to digital tools in emerging markets
• IFC (2024) Digital divide in business intelligence access, emerging economies
• McKinsey Global Institute (2024) AI adoption in small and medium enterprises
• Connected to Prof. Park's supply chain carbon costs paper (Supply Chain Optimizer)

Related project