A live, production AI platform on Microsoft Azure that turns a company's own documents into persona-targeted B2B sales content, where every claim traces to a source and every asset ships EU AI Act Article 50 compliant.
3-tier
Governance on every asset
0
Swappable LLM providers
0
Stores per tenant
A live, production AI platform on Microsoft Azure that turns a company's own documents into persona-targeted B2B sales content, where every claim traces to a source and every asset ships EU AI Act Article 50 compliant.
Companies that sell through partners need sales content that is accurate, persona-specific, and legally defensible, three requirements that a generic AI writing tool cannot meet at once. TechAelia's AI engineering team built the retrieval, orchestration, and compliance layers underneath Cosellus, a business-to-business SaaS platform that generates emails, landing pages, one-pagers, and solution briefs grounded in a customer's own uploaded documents rather than an LLM's general training data. The platform runs on a three-database knowledge architecture (PostgreSQL, ChromaDB, Neo4j), orchestrates specialized agents through LangGraph, and calls the strongest available model per task across OpenAI, Azure OpenAI, Anthropic, Google Gemini, and Perplexity. It is live today with paying customers and real production traffic.
What is AI content grounding? Grounding means an AI system generates text using only retrieved, verifiable source material, then maps every claim in the output back to the exact passage it came from, instead of letting the model draw on unverified general knowledge.
What is retrieval-augmented generation for sales content? It is a generation method where an AI system retrieves real source passages from a company's own documents before writing, then constrains its output to what those passages support, so every sentence can be traced back to where it came from.
Client
B2B SaaS (channel/co-sell go-to-market)
Timeline
Ongoing production engineering engagement
Team
AI/ML and platform engineering
The starting point
Generic AI content fails the exact test enterprise buyers apply first: can you prove it?
Enterprise sales content has three hard requirements a standard language-model wrapper cannot satisfy.
It must be accurate, since an unsupported claim gets rejected by a buyer's legal team. It must be specific to the reader, since a message written for a CFO does not land with an engineer. And increasingly, especially for any vendor selling into Europe, it must be defensible: what is the AI generated content disclosure requirement under EU law? Article 50 of the EU AI Act requires that AI-generated content be marked in a machine-readable, detectable format, with enforcement beginning 2 August 2026 and civil penalties reaching into the tens of millions of euros for non-compliance.
Cosellus needed an architecture where content quality and content governance were solved by the same system, not bolted together as an afterthought. That single design decision became the core engineering challenge.
- Sales content had to cite a real source for every factual claim, not just sound plausible
- Persona targeting had to be a first-class reasoning step, not a prompt-engineering trick
- Every generated asset had to carry proof of its own AI origin, tamper-evident and machine-readable
- Human approval had to sit in the loop before anything shipped, with an audit trail that survives even after the source data is deleted
How we moved from a hard constraint to a system running in production.
The challenge
Enterprise sales content has three hard requirements a standard language-model wrapper cannot satisfy.
It must be accurate, since an unsupported claim gets rejected by a buyer's legal team. It must be specific to the reader, since a message written for a CFO does not land with an engineer. And increasingly, especially for any vendor selling into Europe, it must be defensible: what is the AI generated content disclosure requirement under EU law? Article 50 of the EU AI Act requires that AI-generated content be marked in a machine-readable, detectable format, with enforcement beginning 2 August 2026 and civil penalties reaching into the tens of millions of euros for non-compliance.
Cosellus needed an architecture where content quality and content governance were solved by the same system, not bolted together as an afterthought. That single design decision became the core engineering challenge.
Our approach
TechAelia's engineers designed the knowledge layer first, then layered governance and orchestration on top.
A tri-store system splits relational data (PostgreSQL) from semantic search (ChromaDB vector embeddings) from relationship mapping (Neo4j graph), with a fail-closed tenant filter so one customer's data can never leak into another's results, even on a malformed query. On top of that sits a retrieval-augmented generation pipeline: an intent classifier extracts the target persona, vector search pulls the most relevant source passages, the model writes strictly from that retrieved evidence, and a grounding score flags any output that leans too far on ungrounded reasoning.
Content generation itself runs through a structured, multi-agent pipeline orchestrated with LangGraph rather than a single model call: an intelligence layer analyzes uploaded documents for strategic readiness, a plan-before-execute flow surfaces the intended output for approval, and checkpoints walk the user through calls-to-action, persona-specific value propositions, and a final pre-delivery review. The model layer itself was engineered to be swappable by configuration across OpenAI, Azure OpenAI, Anthropic, Google Gemini, and Perplexity, plus local open-source models for embeddings and reranking, so the platform stays current as new models ship without a rewrite.
Our implementation
The team built EU AI Act Article 50 compliance directly into the content-delivery pipeline.
Every delivered HTML asset carries an embedded, tamper-evident manifest that software can detect as AI-generated, satisfying the Article 50(2) machine-readable marking requirement, and the marking service fails closed: if it cannot produce the manifest, delivery is blocked rather than shipping unmarked content. An append-only, tamper-proof audit ledger records who approved each asset, when, and a cryptographic fingerprint of the content, and that record survives even after the source campaign is deleted for GDPR purposes.
On the infrastructure side, the team owned deployment and release safety for a live, revenue-generating platform: idempotent database migrations so a deploy can never wedge production, off-site backups for the vector and graph stores, secrets management, and access control across a Docker-containerized stack running on Microsoft Azure with a Python/FastAPI backend and a React/TypeScript frontend.
Product views and systems running on live data.
Document intelligence dashboard showing strategic readiness analysis across uploaded knowledge assets, flagging persona coverage gaps before content generation begins.
Grounding and source-trace view showing each generated sentence linked back to its originating document passage, with a visible confidence score.
Compliance and approval workspace showing the human-review checkpoint, the embedded AI-Act manifest status, and the audit ledger entry for a delivered asset.
Turning a legal obligation into a competitive advantage was the deliberate product strategy behind this work.
Last updated
3-tier
Governance enforced on every asset
100%
Assets carry Article 50(2) manifest
5
Model providers, swap by config
3
Database stores unified per tenant
1
Audit ledger record per delivered asset
Aug '26
Regulatory deadline met ahead of schedule
Enterprise buyers in regulated markets increasingly ask "is your AI governed?" before signing, and unmarked, untraceable AI content stalls in legal review regardless of how well it is written. By engineering grounding, disclosure, and audit evidence into the same pipeline that produces the content, Cosellus can hand a buyer a documented, traceable system instead of a promise.
The platform's design also proved resilient to the pace of AI model releases. Because the model layer is swappable by configuration rather than hardcoded, the team validated new provider models against the pipeline without touching the orchestration logic, keeping the platform current without a rebuild cycle.
“Cosellus is built as a strategic consultant that happens to produce content. It reads your documents, reasons about what matters for the target persona, cites its sources, and refuses to make claims it cannot support.”
Tools chosen for scale, auditability, and clean handoff.
AI & Orchestration
Data & Knowledge
Application Layer
Infrastructure
Compliance & Governance Engineering
Delivery & Release Safety
Building an AI product that has to survive contact with enterprise legal review, not just a demo? TechAelia's AI engineering team designs the grounding, orchestration, and compliance architecture underneath production AI platforms.
Contact UsOther programs from our portfolio.
Partner with TechAelia to architect custom enterprise software, autonomous AI workflows, and high-performance native mobile applications. We build scalable digital products engineered to outpace the competition.
