The Suite
Private Conversational AI

Ask your data anything. Get an answer, not a data breach.

The AI Assistant lets anyone query your institution’s data in plain language. A locally hosted model writes a safe, read-only query, a security layer neutralises anything dangerous, and you get an answer in words — all without a single byte leaving your network.

What it does

A copilot that runs on your hardware, not someone else’s cloud.

Generic AI assistants are powerful and unusable for sensitive institutions at the same time — because every question ships your data to a third party. ReconSync’s AI Assistant removes that trade-off entirely.

Ask a question against a specific data source. A locally hosted open-weights model translates your prompt into SQL, runs it as a read-only query against your live database, and synthesises a clear text answer. When the model gets the SQL wrong, the system catches the error and auto-corrects in a loop until the query is valid.

A query guard intercepts everything the model produces, neutralising any mutating statement — DROP, UPDATE, DELETE — before it can touch your data. The AI can read what your role permits, and nothing more.

What happens when you ask

You ask in plain language
Against a chosen data source
Local model writes SQL
Open weights, on your hardware
Query guard screens it
Mutations neutralised
Auto-correct on error
Self-healing query loop
Answer in words
Synthesised, sourced, local
Capabilities

Conversational access, enterprise guardrails

Natural-language power, wrapped in the safety and governance regulated data demands.

Natural-language queries

Anyone on your team can ask questions of their own data and get accurate, sourced answers — no SQL, no analyst, no data leaving your walls.

Local open-weights models

Runs natively hosted models on your infrastructure for text-to-SQL and document reading — never an external API, never an internet dependency.

SQL query guard

Every statement the model generates is intercepted and screened; data-mutating operations are neutralised before execution against live pools.

Self-correcting queries

When generated SQL fails, the system reads the error and retries in a loop — auto-fixing syntax until it returns an accurate result.

Role-aware reading

Access is sliced by role and department tags, so the assistant can only read what a given user is actually permitted to see.

Gets sharper over time

Corrected interactions feed the self-learning Tuner, so the assistant continually improves on your institution’s specific data and terminology.

In practice

Built for the way regulated institutions work

Frontline operations

Let non-technical staff answer their own data questions instantly, instead of filing tickets with a reporting team.

Finance & leadership

Ask for balances, trends and breakdowns conversationally and get sourced answers without opening a BI tool.

Compliance & audit

Every AI query is logged immutably, so conversational access never compromises your audit position.

Any regulated institution

Bring generative AI inside the perimeter where cloud assistants are simply not permitted.

Runs entirely inside your network

AI Assistant is part of one offline platform. Every model, query, and document stays on your hardware — no external API calls, no third-party providers, and no internet dependency, even for AI.

Explore the suite

Works hand in hand with

See AI Assistant on your own data.

We'll walk you through AI Assistant in a private session — on your infrastructure or ours. No data leaves your control.

Request a Private Demo