Skip to content
Aveni Intelligence Engine

Aveni Intelligence
powers every Wealth
and Compliance product

It combines three tiers of AI models with a compliance layer that checks every result against UK regulation.

Why three model types

The right model at the right time

Each task goes to the tier that fits its complexity, speed and accuracy needs, automatically.

Live layer Specialist classifiers · Built for speed

Tier 1 runs first

Every conversation is scanned in real time by lightweight classifiers before anything downstream is triggered.
Depth layer FinLLM · Built for regulatory reasoning

Tier 2 does the reasoning

FinLLM, Aveni’s own models, handles the specialist work: suitability, complaint handling, vulnerability, Consumer Duty analysis.
Learn more about FinLLM
Fallback layer Frontier models · Built for breadth

Tier 3 handles the rest

General-purpose models take anything outside FinLLM’s specialism: general reasoning, creative drafting, edge cases.
Tier 01

Fast, real-time compliance monitoring

Tier 1 models check every conversation as it happens, answering in about 0.2 seconds. It powers the high-volume monitoring inside Detect and Assure.

Detection Methods

Regex Embeddings RoBERTa ModernBERT
Cost Low
Latency ~0.2s
Accuracy Broad
Tier 02

FinLLM, built for UK financial services

FinLLM is Aveni’s own family of language models, trained on UK financial services. It handles the work that needs judgement: suitability, complaint handling, vulnerability, Consumer Duty.

FinLLM model suite – 6 sizes

FinLLM-24B FinLLM-14B FinLLM-8B FinLLM-3B Instruct + Compliance
Cost Low
Latency Varies
Accuracy High
Learn more about FinLLM
Tier 03

General-purpose AI

Models from OpenAI, Anthropic and Google take the work outside FinLLM’s specialism: general reasoning, drafting, edge cases.

Providers

Google Gemini Anthropic Claude OpenAI GPT
Cost High
Latency Mid
Accuracy High

Built for the reality of regulated financial services

See how the three tiers power Aveni Wealth and Aveni Compliance.