The core problem
Your team is already using AI on client work. They’re doing it on personal accounts, with client documents, under terms nobody reviewed. At month end, a single provider invoice arrives with no client codes attached - and lands in overhead, where practices that barely touch AI quietly subsidize the ones that live on it. You can’t say which clients benefited, which files were used, or whose agreement covered the work. That’s also the honest answer to the AI section of every client security questionnaire you sign. Parlé was built to close that gap: put a client code on every dollar of AI your firm spends, keep the files inside your own tenant, and give you a record you can actually show someone.Six pillars of Parlé
Visibility
Access tied to your identity provider, files stored in your own SharePoint
or Google Drive tenant, and model calls running under your own API keys —
never a personal consumer account.
Cost Allocation
Every project carries your own project code - a matter number, cost center
or expense code - straight through to the export. Spend and token usage
break down by project, person, provider, or model over any date range.
Control
Give each project a monthly cap, a lifetime cap, or both, in USD, GBP, EUR
or CAD. Pre-flight enforcement blocks a send before it takes the project
past either one, an 80% warning shows before a cap is hit, and anyone who is
blocked can request a higher limit right from the notice — with a short note
for the Org Admin who approves it.
Shared Context
Shared project threads every team member can see, with real-time presence,
keyword and semantic search, and the ability to promote a personal chat into
a shared project.
Defensible Work
The full exchange - question, reply, and model - stays in the thread. Real
tool calls against the project’s own files ground the work in your
documents, not in a reply that appeared from nowhere.
Governance
Single sign-on via Microsoft Entra ID or Google Workspace, files that never
leave your tenant, an append-only audit log, and role-based access at the
org and project level.
How it fits together
A Parlé project maps to a client engagement. It has a name, a project code you assign (your matter number or expense code), spending caps (monthly, lifetime, or both) in the currency you choose, and a set of members drawn from your organization. Inside a project, the team runs AI-assisted conversations using whichever models you’ve connected - Anthropic, OpenAI, Google Gemini, or Azure AI Foundry on your own API keys. Every token consumed is recorded against the project’s budget in a single canonical ledger that drives both the real-time enforcement and the export report. Files live in a SharePoint site or Google Drive folder inside your own tenant. Parlé never copies them to its own storage. When a team member asks the AI to read a contract or query a spreadsheet, the model calls the file through a real tool call and cites what it found — so there’s something to review, not just a paragraph that appeared. Admins control which models are available at the org level, which models each project may use, and who has access to what. When someone leaves the firm, revoking their identity-provider account removes their access from every project they were on.What’s next
Quick start
Create your organization, connect a model provider, and run your first
AI-assisted conversation in under ten minutes.
Plans & Billing
Compare Team, Business, Business Elite, and Enterprise plans to find the
right tier for your firm’s size and governance requirements.