Microsoft Entra ID single sign-on across every dealer subdomainmyEZData for iOS and Android, one app with a face for every roleBank reconciliation: linked accounts matched to POS cash and armored pickupsBeacon time clock with geofenced auto clock-in, live on the floorSupplier partner API: catalog, orders, ship notices and receivables every 20 minutesAsk your data in plain English and get a validated query backInbound device shipments tracked by UPS and pushed to the crew on arrivalMicrosoft Entra ID single sign-on across every dealer subdomainmyEZData for iOS and Android, one app with a face for every roleBank reconciliation: linked accounts matched to POS cash and armored pickupsBeacon time clock with geofenced auto clock-in, live on the floorSupplier partner API: catalog, orders, ship notices and receivables every 20 minutesAsk your data in plain English and get a validated query backInbound device shipments tracked by UPS and pushed to the crew on arrival
Request a demo

Platform/09/AI, with guard rails

09 AI, with guard rails

Claude, wired in. Not bolted on.

The interesting part is not the model, it is the guard rails around it. A question becomes a structured query specification the platform validates and scopes before anything runs. An assistant answers only from what the user can see. A review pass reads the performance email before it goes out. Each is a module switched on per dealer group with its own token accounting, and nothing that determines pay is decided by a model.

Request a demo On the overview
Ask
Natural language to a validated query specification, never raw SQL
Assist
In-product assistant scoped to the user, with a report tool
Review
Performance email review, payroll narrative, carrier weekly summary
Control
Per dealer group, off by default, token usage tracked, kill switch

/ Works with

    01 Ask your data

    Ask in plain English.
    Get a validated query back.

    Ask your data

    Natural language to a validated query

    Try one

    Sample responses. The permission model still applies to whatever the model proposes.

    02 Guard rails

    The model proposes.
    The platform decides.

    1. 01The question

      Typed in the studio, the assistant or the app, with the dataset schema the user may see attached.

    2. 02The proposal

      Claude returns a query specification: dataset, metrics, groupings, filters, order. Never SQL.

    3. 03Validation

      Unknown columns, disallowed datasets and impossible filters reject the specification before anything runs.

    4. 04Scope

      The caller's store scope and role are applied on the server. Nobody asks their way past what they are allowed to see.

    5. 05Execution

      The platform's own query layer runs the validated specification against the scoped dataset and returns the table.

    / 01

    Scoped per dealer group

    • Landlord config
    • Off by default
    • Kill switch

    AI is a module the landlord switches on per dealer group, with that group's own configuration. Groups that do not want it never see a button. An app-wide switch turns it off everywhere.

    / 02

    Usage tracked

    • Token accounting
    • Per tenant
    • Budgets

    Every call is metered against the dealer group on both the group's usage page and the landlord's, so spend is attributable and budgets can be set, not discovered on a bill.

    / 03

    Where it does not belong

    • No pay decisions
    • No writes
    • Review, not authority

    Nothing that determines pay is decided by a model. AI reads, summarises, proposes and reviews; the engine calculates, the person approves.

    03 Where it is used

    Six places it
    earns its keep.

    AI features

    In production

    FeatureWhat it doesWho uses it
    Ask your dataA question over the studio datasets becomes a validated specification and a result table, with the specification shown.Managers, finance, owners
    AssistantA chat widget on every admin page and a full page, scoped to what the user can access, with a report tool that opens the right report.Everyone with the module
    Performance email reviewReads the day's performance email against the numbers and flags a narrative that does not match before it is sent.Email center
    Payroll narrativeA period summary in prose a manager will read, generated from the locked run with configurable prompts.Payroll
    Carrier weeklyA weekly summary of the carrier reconciliation: deltas, disputes, patterns by store.Finance
    PDF extractionThe carrier's payment calendar PDF is read into expected deposits for financial reconciliation.Finance

    Each feature is its own switch and its own prompt configuration per dealer group.

    04 Related

    / 07

    Reports and Data Studio

    Carrier scorecards, margin, traffic conversion, report builder, saved reports.

     / 06

    Ticketing and communications

    Two-lane tickets with SLAs, chat, WhatsApp cards, email center, news feed.

     / 12

    Security and tenancy

    Entra ID sign-in, a database per dealer group, layered permissions, immutable records.

     / all

    Every platform page

    The twelve deep dives, one per capability, with the overview above them.

    → Start

    Bring three questions
    you ask every week.

    We put them to the studio on your data and show the specification the model proposes, the validation, and the result, with the scope applied.

    Request a demo