Skip to main content

Services / AI Solutions (Governed)

AI you can put in front of an auditor

We design and build governed AI into your ERPNext deployment: numbers computed deterministically, actions gated green, amber, or red, and every invocation written to an immutable audit log. The same pattern already runs in production across seven of our own applications.

Six rules, applied to every build

Deterministic-first

Numbers are computed in SQL and Python; the language model only narrates what was already calculated. AI never invents a figure, and the maths is auditable.

Governance gate

Every automated action is classified green (auto), amber (confirm), or red (review-only) before it can touch a record. Autonomy is bounded and deliberate.

Immutable audit log

Every AI invocation, prompt, and action is written to a tamper-evident log, the evidence trail that makes AI adoptable where hallucination is unacceptable.

Bring-your-own models

Run on cloud providers (OpenAI, Anthropic, OpenRouter) or fully on-premise with Ollama. Sensitive data never has to leave your infrastructure.

Domain-specialist agents

Registry-dispatched agents for finance, audit, BI, lab review, and legal, each grounded in your live ERP data, with human sign-off where it matters.

Graceful degradation

When AI is disabled or unavailable, the deterministic core keeps working. AI is an accelerator, never a single point of failure.

Proof, not promise

Already running in production

This is not a theoretical pattern. It is the same discipline we ship across our own portfolio, and the one we bring to your build.

Coale Finance

OpenAI, Anthropic, OpenRouter, Ollama, or local. Deterministic compute, then a governance gate, then an optional governed write, then LLM narration, every step logged to an immutable Agent Action Log across 30+ analytics agents.

Mkaguzi

Ollama cloud or local. An AI finding reviewer enriches and stress-tests audit findings; the auditor retains sign-off on every conclusion.

Maabara

Ollama on-premise. An AI result reviewer cross-checks lab results against reference ranges and never overrides the deterministic flag engine.

Want AI your auditors trust?

We will identify where AI earns its place, where deterministic automation is safer, and how both stay under your control.

Book an AI Consultation