DataInbox
    AI Companions

    AI is a capability. Your business remains in control.

    An AI Companion is an external intelligence capability that DataInbox invokes under company-defined conditions. Use a direct provider, an AI router, a managed multi-model platform, or private inference without moving business logic, consent, or authority into the provider.
    • Included with every plan
    • Per-Inbox or shared AI policy
    • Sovereign and private options
    • Standardised task and result contract

    The Companion principle

    DataInbox decides. Companion performs.

    Connect the service provider you choose to the Inboxes you authorise. Each Inbox determines what context may be shared, what operation is allowed, which limits and approvals apply, and what evidence must return. The provider supplies the capability. Your conditions retain control.

    1. 01

      Business message

    2. 02

      Inbox context and policy

    3. 03

      Companion capability

    4. 04

      Result back to Inbox

    One pattern, many capabilities

    One model, clearly divided by availability.

    AI is included in every plan. Payment and Identity are available specialist Companions. Delivery and Voice are coming soon.

    AI Companion

    Direct models, managed platforms, routers, and private inference

    Choose approved AI for a specific Inbox or through shared company policy. The governed AI Companion layer is included with every plan.

    Included in every plan

    Available now

    Specialist Companions

    €100 each

    Payment Companion

    Your approved payment service provider

    Execute a permitted payment from a webchat conversation, through an authorised agent, or in response to a defined business action. Assign the provider to any authorised Inbox. Inbox policy controls who or what may initiate the payment, its purpose and amount, and the required limits and approvals. The outcome returns as a governed business message.

    Identity Companion

    Your approved identity, access, account resolution, or lifecycle provider

    Make the identity provider you choose available to any authorised Inbox. Verify identity, resolve an account, or control access to a personal Inbox while purpose, authority, consent, context, outcome, retention, and evidence remain governed.

    Coming soon

    Next Companion capabilities

    Coming soon

    Delivery Companion

    Fulfilment, logistics, shipping, and delivery providers

    Make an approved delivery provider available under Inbox-specific order, recipient, status, exception, and evidence conditions.

    Coming soon

    Voice Companion

    Voice and conversation providers

    Govern voice interactions, channel consent, context, transcripts, actions, and evidence inside the relevant Inbox.

    Four valid AI Companion patterns

    Choose the capability boundary that fits.

    DataInbox does not force model selection into one layer. The company decides where routing belongs for each operating context.

    Direct AI Companion

    DataInbox selects one eligible provider endpoint for the task and receives its standardised result.

    AI Router Companion

    DataInbox selects an orchestration service that can choose an underlying model inside an approved boundary. euqai is one example of this category, not the category itself.

    Managed Multi-Model Companion

    DataInbox invokes a cloud platform that exposes several approved model families through one managed control surface.

    Private AI Companion

    DataInbox invokes a local, dedicated, or privately hosted inference endpoint when data must stay inside a controlled environment.

    Selection remains business policy

    Data and sovereignty

    Data classification, consent, residency, endpoint, and permitted fields.

    Capability and output

    Task type, output schema, quality threshold, evidence, and review requirement.

    Cost and capacity

    Budget ceiling, rate limit, throughput, and high-volume handling.

    Reliability

    Availability, timeout, retry, fallback order, and human escalation.

    Configuration proof

    The boundary is configuration, not a slide.

    This sanitised customer-service pattern demonstrates how message context, lookup requirements, AI instructions, and result handling can remain explicit. Sensitive prompt logic and customer details are not published.

    Sanitised policy configuration

    Real configuration pattern, customer details removed

    Validated pattern
    taskcustomer_order_status
    companion.typeai
    context.latest_message$message.content
    context.order_number$message.orderNumber
    context.postcode$message.postcode
    policy.existing_orders_onlytrue
    policy.lookup_requiredtrue
    output.languagenl
    Prompt instructions, credentials, provider keys, customer identity, and internal endpoints are intentionally omitted.

    Runtime decision trace

    1. 01Message context

      Latest customer message and permitted known fields

    2. 02Deterministic gate

      Order number and postcode policy must pass

    3. 03Required lookup

      No status claim without a successful business-system result

    4. 04AI Companion

      Drafts a short answer from approved result fields only

    Governed result: the Companion output returns as a structured business message. Existing rules decide whether it may be sent, reviewed, retried, or rejected.

    AI Companion directory

    Provider pages without provider lock-in.

    These pages support discovery and entity indexing. They describe stable architectural fit, not a promise about a specific model version. Connector availability depends on configuration, credentials, contract, region, and deployment.

    OpenAI

    Direct AI Companion

    General-purpose reasoning, multimodal understanding, generation, and tool-enabled tasks.

    View Companion profile

    Anthropic

    Direct AI Companion

    Reasoning, long-context analysis, generation, and tool-enabled tasks.

    View Companion profile

    Google Cloud

    Managed AI Companion

    Managed multimodal, generative, and model-serving capabilities in Google Cloud.

    View Companion profile

    AWS

    Managed multi-model Companion

    Managed access to multiple model families through an AWS control surface.

    View Companion profile

    Mistral AI

    Direct or privately deployed AI Companion

    European generative AI capabilities with managed and deployment-flexible options.

    View Companion profile

    Meta

    Open-weight AI Companion

    Open-weight language-model capabilities deployed through infrastructure selected by the organisation.

    View Companion profile

    Cohere

    Direct enterprise AI Companion

    Enterprise generation, retrieval, embedding, and reranking capabilities.

    View Companion profile

    AI21 Labs

    Direct AI Companion

    Language-model capabilities for analysis, generation, and long-context workloads.

    View Companion profile

    Hugging Face

    Model ecosystem Companion

    Access to open-model inference and deployment tooling across a broad model ecosystem.

    View Companion profile

    Stability AI

    Media AI Companion

    Image and media generation or transformation for approved creative workflows.

    View Companion profile

    Perplexity

    Web-grounded AI Companion

    Web-grounded search and answer capabilities for tasks that permit external research.

    View Companion profile

    Groq

    Inference AI Companion

    Hosted inference for supported model families where response speed is an important condition.

    View Companion profile

    Define capability first. Select providers second.

    Start with one business message, the AI task it may invoke, the data it may share, and the result contract it must satisfy.