GeneralMind
- Platform Engineering, AI / Autonomous Agents
Client
GeneralMind
Location
Berlin, Germany
Engagement
Platform Engineering and AI / Autonomous Agents
Technologies & Tools Used
From document chaos to autonomous operations.
GeneralMind builds fully autonomous AI employees that handle document-heavy back-office work.
The platform reads purchase orders, invoices, and sales orders arriving through email, portals, or EDI, extracts and validates the data using AI, and writes it back into ERP systems including SAP, JD Edwards, and Dynamics.
It runs in production across enterprise tenants in food and beverage, distribution, healthcare, and logistics, including Hartmann, Sennheiser, Bayer, Oatly, Kloeckner, Lekkerland, Müller and Bofrost, from one shared codebase.
Building an auditable, multi-tenant AI platform for enterprise operations.
Enterprises receive thousands of purchase orders, invoices, and sales orders every month as PDFs, EDI files, scans, and portal exports.
Many still rely on manual entry to transfer that data into SAP, JD Edwards, and Dynamics.
GeneralMind aimed to replace this process with an AI agent that could read any format, extract data with a defensible confidence level, reconcile it against live ERP master data, and escalate to a human only when uncertain.
Spark Eighteen helped GeneralMind turn manual, document-heavy back-office work into an auditable, multi-tenant AI platform that enterprise finance and procurement teams can trust.
The Challenge
GeneralMind needed to serve multiple enterprise clients from one codebase, make every AI-generated value verifiable, and ensure long-running AI workflows could survive restarts and deployments without repeating costly work.
Multi-Tenancy at Scale
Trust in AI Output
Durable AI Pipelines
The Solution
Spark Eighteen partnered with GeneralMind’s engineering team across the multi-tenant product surface.
We built feature-flag-driven customer resolution that allows one Next.js codebase to serve more than 17 enterprise tenants.
The team created a shared design system, icon library, and configuration-driven data-table framework.
We also delivered AI-assisted review experiences, including citation highlights over source documents, field-level confidence scoring, an explainable audit trail, and an AI email composer.
These capabilities were built on top of a durable, model-agnostic AI orchestration and data platform.
The Results
GeneralMind now serves a growing portfolio of enterprise clients from a single codebase.
Finance and procurement teams can verify AI extractions at a glance.
New tenants, document types, and AI capabilities can be introduced without re-architecting the underlying platform.
1 codebase
3 layers
1 system
Business Outcomes
Multi-tenant at scale - One codebase serves every enterprise tenant, with no per-client forks to maintain.
Resilient pipelines - Temporal orchestration resumes costly AI work after restarts instead of re-running it.
Consistent delivery - A shared design system and data-table framework ship tenant features fast.
Trustworthy AI - Confidence scores and source citations let reviewers verify every extracted field.
No vendor lock-in - Automatic fallback across Claude, GPT, and Gemini keeps documents processing.
Explainable operations - The audit trail surfaces the AI's confidence and reasoning, not just changes.
Spark Eighteen embedded with our engineering team and took real ownership of our multi-tenant product surface. They built the architecture that lets us serve every enterprise client from a single codebase, and the review experiences – citations, confidence scoring, and audit trails – that let our users trust what the AI produces. A well-structured, professional, and genuinely valuable partnership.
