AI agents that close the books, clear the alerts, and show their work.

We design, build, evaluate, and run production agents for reconciliation, intercompany close, financial-crime investigation, and claims — for fintech, multi-entity companies, and healthcare. Fixed fee. Live in 6–12 weeks. Deployed in your VPC. Official Databricks partner and member of the Claude Partner Network.

Book a 30-min architecture walkthrough →

Built with

Databricks · Official Partner

Anthropic · Claude Partner Network

Agents We've Built

Working demos and production builds — every card is clickable. Recorded runs use demo data.

Claude

Databricks

AWS

Transaction Intelligence Agent

End-of-day reconciliation across gateway, processor, and bank settlement files — every break surfaced with its evidence before your team starts the day.

• Surfaces critical exceptions — orphan settlements, amount mismatches, duplicates

• Two-tier architecture — PII never reaches the public-tier agent

• Audit trail + human approval on every privileged action

Open the recorded run →

Claude

NetSuite

AWS

Concilo — Intercompany Close Agent

Intercompany close for multi-entity groups — GL exports from every entity matched, netted to nil, and consolidated with a hash-stamped evidence record.

• Auto-clears ~90% of intercompany matches — deterministic engine, LLM never invents a match

• Routes the ambiguous ~10% to a human queue with suggested resolutions

• Hash-stamped, audit-grade evidence record on every match and elimination

Open the recorded run →

Claude

Databricks

AWS

Claims Triage Agent

Pre-adjudicates and routes healthcare claims, prioritized by denial risk, with a certified-reviewer gate on every decision.

• Extracts structured data from messy documents

• Prioritizes by denial risk

• Built HIPAA-ready from day one

Walk through this agent in 30 min →

Claude

Databricks

AWS

Lakehouse Ops Copilot

Answers your team’s data questions over a governed Databricks lakehouse — cited sources on every answer, inside your VPC.

• Natural-language queries on trusted data

• Cites its sources on every answer

• Deploys inside your VPC

Walk through this agent in 30 min →

How we build an agent your auditors will sign off on

Four rules on every build: the deterministic engine decides, the model explains, the evidence chain proves, a human approves. Evals gate every release, and personal data stays in a vault the everyday agent cannot open. It is the pattern Anthropic describes in Building Effective Agents — hardened for regulated finance and healthcare.

Agent architecture: Human and LLM Call exchange, the LLM Call loops Action/Feedback with the Environment, and can Stop.

See your workflow as an agent — in 30 minutes.

We map one process to an agent, the evidence chain it needs, and a fixed-fee range. No slides, no pitch — and if it isn’t a fit, we say so on the call.

Book a 30-min architecture walkthrough →
© Soleil Technologies LLC · Fremont, CAHome · Assessment · Security · Fintech · Healthcare · Embedded Engineering · Privacy · Terms