Drachma raises $1M to plug AI agents into corporate data systems
What's the deal? Drachma, a New York-based startup founded by Brazilian software engineer Leonardo Felipe Nerone, has closed a $1 million pre-seed round. The company builds custom autonomous AI agents designed to integrate directly into corporate data systems, APIs, and operational workflows — tackling what it sees as the biggest barrier to enterprise AI adoption.
Rather than training models from scratch, Drachma layers a proprietary integration architecture over existing large language models. Nerone calls the approach "context engineering": embedding each client's business rules, permission controls, and internal tools so AI agents can operate reliably inside real enterprise environments.
Why now? The market is awash with generic AI tools, but adoption in critical business processes remains low because most solutions lack deep connections to a company's actual data and systems. "The problem isn't a lack of AI tools. The problem is that they aren't connected to the real context of companies," Nerone said.
One early use case lets both technical and non-technical users query massive structured datasets on platforms like Databricks through natural-language conversation — replacing complex query-writing with plain speech.
What could go wrong? Enterprise AI reliability is a known headache. LLM hallucinations and data security lapses can erode trust fast, and Drachma's value hinges on preventing both. The startup also uses an unconventional pricing model: fees are tied to measurable client impact — cost reduction, efficiency gains, or productivity improvements — rather than standard SaaS licensing. That approach could complicate revenue predictability as it scales.
The signal: Drachma's bet reflects a broader shift in enterprise tech away from passive software toward systems where AI actively orchestrates automation and bridges corporate data silos. The $1 million will fund a reusable integration infrastructure meant to speed up bespoke deployments without sacrificing customisation.
If the thesis holds, startups that solve the "last mile" of enterprise AI integration — not the models themselves, but the messy plumbing that connects them to real operations — could capture outsized value as companies move past demos and demand production-grade results.
Read more: brazilianfinance.com
