Forge your enterprise with AI

Ask, act and automate across every system you run, on any model. Smart Forge is an award-winning cognitive layer, not another application: it makes your existing systems AI-ready without replacing them.

CortexTry it: press send
SalesforceJD EdwardsBlue YonderAzure DevOps4 systems connected
Ask across your systemsAnswers come back as tables and charts, built with approved tools.
OperationalModel: Auto
Why teams use it

What changes when your systems can answer

Smart Forge is for the everyday problems that sit between the systems you already run. Four we hear most often. See how we would work together

  • Answers take days, because the data sits in five systems.

    Anyone can ask in plain language and get an answer from live data across those systems, then keep it as a dashboard tile or a report.

    Cortex
  • Routine checks depend on someone remembering to do them.

    Agents run on a schedule or in the background, act on what they find, such as opening a ticket or putting an order on hold, and record what they did.

    Reflex
  • One job means moving between screens in different systems.

    One screen built from tools on every system, with the actions your team needs, and no development project.

    ForgeView
  • Older systems cannot use AI, and the AI add-on costs too much.

    Any system with a database, a web API or its own command language gets tools AI can use, without an upgrade or a replacement.

    Forge Studio
Integrations

Connect the systems your team uses

Warehouse, ERP, databases, tickets, messaging and AI models. Each connection runs through tools built and approved in Forge Studio. Read how tools connect your systems

    System not listed? We build new connectors to your requirements.

    The paradigm shift

    A cognitive layer, not another application

    Today every assistant, report and app is wired to each system on its own. Smart Forge puts one layer in between, where orders, stock, customers and invoices from all your systems connect live. Everything else works from that layer. Read how the layer works

    • One live layerAll your systems, from orders to stock to customers, connected in one place.
    • Keeps best-of-breedAdds intelligence without replacing your systems.
    • Capabilities, not appsEach system publishes what it can do, managed in one place.
    • Persistent contextContext carries across systems at run time.
    Kinds of question

    From answers to decisions

    The questions a business asks range from a single fact to a decision. One layer serves all of them, with the same tools on the same systems. The further up, the more systems a question draws on. Read how the layer answers them

    1. Transactional“What is the status of order 48290?”JD Edwards
    2. Contextual“Why is this order on hold?”JD EdwardsSalesforce
    3. Predictive“Which orders will miss their date this week?”JD EdwardsBlue Yonder
    4. Economic“What revenue is at risk, and where?”JD EdwardsSalesforceSnowflake
    5. Causal“Which decision caused this delay?”Blue YonderJD EdwardsSnowflake
    6. Prescriptive“What should we do, and in what order?”JD EdwardsBlue YonderSalesforceSnowflake
    Reflex

    Agents that work without being asked

    Describe the job, and Reflex builds the agent. Deploy it in one of two modes. Both are running below on their own. Read about deployed agents

    Scheduled agent · Operational

    Designed by AI, runs without it

    Every Monday it checks ERP accounts against the HR list and opens an IT ticket for leavers.

    Leavers access check v2Scheduled · every Monday 07:00
    Valid
    Run history appears here. Press Run.
    Sample data

    Try “None missing” to see the condition take the other branch.

    Background agent · Reasoning

    Deployed agents, always on duty

    Every hour the Credit watch agent checks overdue invoices against open deals, decides who gets a reminder and who goes on credit hold, and carries it out.

    Multi-agent orchestration

    One question, several specialist agents, one answer

    Scroll to follow one question through the agents. See every step an agent takes

      Multi-agent orchestration2 agents · 2 connections · led by the data
      Dashboards and reports

      Keep the answer. Pin it to a dashboard or turn it into a report

      A dashboard tile reruns the saved code on a schedule, with no AI tokens. For a report, pick a template and colours, approve the plan, and the Report agent builds the PDF.

      Screens on demand

      Screens built from your tools, ready when you need them

      Need a screen to review live data and act on it? ForgeView's screen builder assembles it from the tools already created for your systems, with filters, a grid, detail screens, actions and export. Simple screens need no code, and every action shows what it will change before it runs. Read about ForgeView

      Controls

      Enterprise-grade controls

      Guardrails, approvals and an audit trail on every action, from sign-in to every write. Read about guardrails

      Order Analyst wants to change live data

      Guardrail · writes to ERP-PROD need approvalrelease_credit_hold(orders = 12)
      Waiting for approvalApproveEditReject

      Recorded with who asked, what ran and who approved

      Guardrails and approvals

      Every agent is read-only by default and carries its own guardrails. When one wants to change something, the chat pauses and a person approves, rejects or edits.

      Audit trail

      Every action runs as the user, so each call and change is tied to the person who asked.

      Deploy in your own cloud or on-premise

      Runs as containers on Kubernetes: in our cloud, in your own Azure, AWS or Google Cloud account, or on-premise in your own data center.

      AzureAWSGoogle Cloud
      • Environment guardrails. Try it on test data before it touches live operations.

      • PII guardrails. Emails, phone numbers and names are masked before the model sees them.

      • Data stays home. In Operational mode the model sees metadata only.

      • Change control. Every change is reviewed in Git, like any software.

      Smart model router

      The right model for every questionComing soon

      Set a cost target and the router works to stay under it: hard questions get a flagship model, simple ones a fast, cheap one. Use Anthropic, OpenAI, Google and others, or open models on your own servers.

      Model names are examples of what you can connect. Logos are trademarks of their owners.

      Frequently asked questions

      01What is Smart Forge?

      A cognitive layer that sits on top of the systems you already run, such as your ERP, warehouse system, CRM and databases. Your teams ask questions in plain language and get answers from live data, and the same layer runs agents that do routine work across those systems.

      02Which systems can it connect to?

      Any system with a database, a web API or its own command language, legacy or modern. That includes an AS/400, a modern system with no AI of its own, or one whose AI add-on costs too much. If yours is not on the list, we build the connector. See how tools connect your systems.

      03Do we have to replace or upgrade our systems?

      No. Smart Forge connects to your systems as they are and adds intelligence on top. Nothing is replaced, and your data stays where it is.

      04What can our teams do with it?

      Ask questions in Cortex and get tables and charts back, keep an answer as a dashboard tile or a report, open screens built from your tools in ForgeView, and deploy agents with Reflex that run on a schedule or in the background and act on what they find: open a ticket, send a reminder, put an order on hold or update a record, then record what they did.

      05Which AI model does it use?

      The one you choose. A hosted provider such as OpenAI, Anthropic or Google, a model in your own cloud account, or an open-weight model on your own servers. Your tools, agents and guardrails stay the same whichever you pick. Read about bringing your own model.

      06Does the AI see our data?

      It depends on the mode. In Operational Mode the model sees only your question and the descriptions of your tools, never your records. In Reasoning Mode it reads results to work things out, after guardrails have hidden the values you choose. Read about Operational Mode.

      07Where does it run?

      As containers on Kubernetes, in our cloud, in your own Azure, AWS or Google Cloud account, or on-premise in your own data center.

      08How do we get started?

      Bring the questions your teams ask every day. We connect to your systems and show you Smart Forge answering them. Talk to our sales team.

      See it on your own systems

      Bring the questions your teams ask every day.

      Get started