Private: Building an AI Platform Where Privacy Comes First

Private is easier to show than to describe. Here's a short introduction to what it does and why it's built the way it is.

Why we built it

AI adoption has accelerated faster than most organisations' comfort with where their data actually goes. Sensitive business documents, legal agreements, intellectual property, customer information, and internal communications routinely get processed by third-party platforms, which is a hard sell for any business with real confidentiality or compliance obligations. Private exists to remove that trade-off: an AI platform built so organisations can use document intelligence and search without handing sensitive information to an external provider.

The problem with most AI adoption

A few issues kept coming up with the organisations we spoke to before building this:

• Data privacy risk: most AI tools require sensitive information to leave the organisation's environment to reach an external provider.
• Fragmented tooling: document management, knowledge retrieval, communication, AI assistance, and data analysis usually live in separate tools that don't talk to each other.
• Compliance pressure: legal, financial, healthcare, and government work all come with strict rules about where information is stored and processed.
• Lack of ownership: a lot of businesses are, reasonably, uncomfortable building critical workflows around a system they don't control.

The market needed AI capability without the privacy trade-off, and nobody we looked at was solving for that directly.

The principle behind the build

Private was designed around one rule: your data should remain yours. Rather than routing sensitive information out to external services, the platform is built to bring intelligence to the data instead, while still operating within private environments, processing sensitive information securely, delivering enterprise-grade intelligence, integrating with existing workflows, and keeping the organisation in control throughout.

What's actually in the platform

Private is a modular intelligence ecosystem, not a single chatbot bolted onto a search bar. It combines a few distinct pieces:

• AI assistant: a conversational interface for interacting naturally with documents and business systems.
• Document intelligence: processing for contracts, policies, legal agreements, reports, PDFs, and office documents.
• Knowledge retrieval: Retrieval-Augmented Generation (RAG) grounds responses in an organisation's own data rather than relying purely on general AI knowledge.
• Local AI infrastructure: organisations can deploy models within their own environment, cutting dependence on external cloud services.

Deep research: connecting answers across many documents

A single document lookup is the easy case. The harder, more valuable problem is a question whose answer is scattered across dozens or hundreds of files, the kind of question that normally means hours of manual cross-referencing. Private's deep research mode handles that: it runs a multi-step search across an entire document set, pulls out the relevant figures and clauses from each source, and assembles them into one grounded answer with every figure traceable back to where it came from.

A finance team is a good example of why this matters. Asking "how has our gross margin trended across the last eight quarterly reports, and what drove the biggest swing" isn't a single-document question, it requires reading every report, pulling the relevant line items out of each one, and reasoning about the comparison. Deep research does that pass automatically and returns an answer with each number linked back to the specific report and page it came from, instead of a person manually opening eight PDFs.

Document intelligence for professional services

The most immediate use case is document intelligence. Professionals lose real hours every week hunting through documentation for one specific clause or date. Private indexes documents, understands their structure, extracts metadata, identifies key clauses, runs semantic search, and generates contextual answers, so a static folder of contracts or reports becomes something you can actually ask questions of in plain language and get a sourced answer back.

Privacy as architecture, not a feature

Privacy isn't a setting you toggle on in Private, it's the starting assumption the rest of the system is built around: local processing where possible, controlled data access, secure document storage, flexible deployment options, and the organisation retaining ownership of its own information throughout. That's what lets a business adopt AI tooling without quietly compromising the confidentiality requirements it already has to meet.

Where this is useful beyond document search

The platform is built to function as a central intelligence layer, not just a document Q&A tool. A few of the use cases that shaped the design:

• Legal services: contract review, clause identification, compliance analysis, case research support.
• Corporate knowledge management: internal knowledge retrieval, policy search, procedure assistance, employee onboarding.
• Operations: information discovery, process documentation, decision support, reporting assistance.
• Executive productivity: research support, strategic analysis, summarisation, knowledge management.

The technical pieces

Under the hood, Private brings together large language models, local AI infrastructure, semantic search, vector databases, Retrieval-Augmented Generation, document intelligence pipelines, and metadata extraction, all working together so responses are grounded in an organisation's actual information rather than generic training data. That's the core difference from a general-purpose AI assistant: Private understands the specific information ecosystem it's deployed into.

What this changes for an organisation

In practice, the benefits come down to a few things: sensitive information stays under organisational control, employees spend less time searching and more time acting on what they find, previously buried information becomes instantly searchable, several disconnected workflows consolidate into one platform, and the organisation can keep adopting AI without losing governance or compliance posture along the way.

Why this is a different approach

Most AI platforms are built around public data and cloud-based processing by default. Private inverts that: data ownership, privacy, local intelligence, and organisational control are the starting design constraints, not an afterthought layered on top. Instead of asking an organisation to trust an external system with its sensitive information, Private lets it deploy intelligence inside an environment it already controls.

Where it stands

Private is currently being relaunched, and we're not labelling it as live while that's in progress. The architecture and the use cases above reflect the platform as built: a privacy-first way for organisations to get real value out of AI on their own documents and knowledge, without sending anything sensitive somewhere else to do it.