AI development & emerging products
Better context. More useful AI.
We are developing ways to connect AI with relevant business information and put AI assistance to work in practical product workflows.
From business information to supported answers
Documents, spreadsheets and operational exports hold valuable business knowledge. Our RAG Engine research explores how AI can retrieve the relevant information and use it to answer a question, with references back to the material used.
This approach is known as retrieval-augmented generation, or RAG: find relevant source material first, then give it to an AI model as context for its answer. Source references help people check an answer; they do not make AI infallible.
What we have developed
Our prototype brings together document ingestion, retrieval, structured querying and evaluation. The work includes:
- Processing text, Markdown, PDF, JSON and Excel sources into searchable information.
- Combining semantic search—matching by meaning—with keyword search.
- Generating answers with references to retrieved source passages.
- Exploring spreadsheet questions through query planning and database calculations, rather than relying on a language model to calculate from prose.
- Evaluation scenarios for retrieval, source references and structured-data answers.
This is a research and product-development prototype. We are continuing to evaluate retrieval quality, document handling and structured-data queries as we work towards a product.
A potential next step: Xaventa Context
We are exploring Xaventa Context as a future addition to the Xaventa product family: a context and evidence layer for business AI agents, built on our RAG Engine research.
The proposed direction is to give an agent a focused package of relevant information, source references and structured results for a task. The aim is to help software teams understand what information an agent used and keep that information within the appropriate customer and permission boundaries.
Our productisation ideas include secure access to business knowledge, evidence packages that work across different AI models, and governed longer-term agent memory. Access policies, source freshness, auditability and memory lifecycle controls are part of the proposed roadmap.
In development. Xaventa Context is a proposed product direction, not a launched service. Product scope, integrations and availability are still being explored.
AI assistance in Xaventa Transform
Xaventa Transform includes optional AI-assisted header mapping to help match incoming column names with fields in a target template. It is useful when different files use different terminology for the same information.
AI suggestions support the mapping process alongside reusable mapping profiles and conventional matching. Users can review and adjust mappings before validating and processing the data. Xaventa remains in closed beta; availability depends on the enabled features.
Explore a shared opportunity
Our AI work follows the same model as our wider business: products and ventures we own or part-own. We welcome conversations with organisations bringing complementary market knowledge and a clear business problem to a potential shared venture.
Products & venture opportunities