About KB Sandbox

Bring AI into the enterprise—without losing the knowledge, judgment and control that make the enterprise work.

People are already using tools such as ChatGPT, Claude, Gemini and NotebookLM to research, understand documents and complete work faster. That individual productivity is valuable—but an enterprise needs more than a collection of private AI conversations and independently assembled notebooks.

KB Sandbox helps organizations turn rapid, individual AI adoption into shared organizational capability. It brings approved knowledge, employees, AI models, specialist agents, evidence and human authority together within governed Projects.

KB Sandbox's goal is to enable organizations to discover where AI can genuinely make their people more productive—and establish the safest, most effective way to do it.

The adoption gap

From individual productivity to organizational capability

AI adoption did not begin with an enterprise transformation programme. It began with people discovering that an AI assistant could make ordinary tasks easier.

Instead of searching through several recipes and deciding which instructions to follow, someone can ask an assistant how to cook beef stroganoff using the ingredients already in the kitchen. The same person can use AI to plan a trip, understand a difficult document, summarize a long email or draft a response.

At work, employees naturally apply the same approach: they summarize reports, research customers, interpret policies, prepare proposals and work through unfamiliar technical problems. This is valuable, but individual adoption does not automatically become organizational capability.

Is the assistant using current, approved organizational knowledge?

Can useful work be shared instead of repeatedly recreated?

May this employee—and this AI provider—use the information?

Can another person inspect the sources behind the result?

Who approves a recommendation or action when consequences matter?

Can a successful practice become repeatable across the organization?

Personal AI assistance → Shared organizational knowledge → Governed Projects → Repeatable Methods and approved agents → Measurable organizational capability

Two experiences, one platform

Useful for employees. Governable by the enterprise.

For everyday work

Ember Workspace

Ask questions, find approved information, prepare documents, continue Project work and use approved agents—all through one conversational interface grounded in the knowledge you are permitted to use.

✓ Ask Ember
✓ View my Projects
✓ Continue recent work

Enter the workspace →

For improving how the organization uses AI

AI Workbench

Curators, consultants and administrators build trusted knowledge, investigate opportunities, compare approaches, establish Methods, evaluate agents and retain human authority over important decisions.

✓ Build trusted knowledge
✓ Apply Workbench Methods
✓ Evaluate models and agents

Explore public knowledge →

People using shared knowledge, evidence and governed AI in their everyday workplaces — One shared platform supports different kinds of work while keeping knowledge, evidence and approval connected.

Designed around organizational roles

People remain part of the system

Member

Uses Ember for everyday work.

Consultant

Discovers and designs valuable AI applications.

Curator

A department head or trusted assistant who knows the team and helps staff become more productive with approved knowledge and AI.

Administrator

Operates and secures the organization’s platform.

The Curator is central to adoption. Usually a department head or a trusted assistant, the Curator understands the team, its responsibilities and the work people are trying to complete. Their purpose is not simply to manage documents—it is to help staff become more productive by turning departmental knowledge and practical AI use into a trusted everyday capability.

Platform roles determine what someone may do. Project membership, knowledge permissions and assigned approval authority determine where they may do it and which decisions they may make.

A simple adoption journey

Discover → Try → Apply

Discover

Find where AI could make employees more productive. Begin with real activities, recurring questions, delays, duplicated effort and knowledge people struggle to find.

Try

Explore practical ways to help. Use approved evidence, compare models and approaches, decide whether an agent is needed, and determine how AI should connect with existing systems and workflows.

Apply

Evaluate what worked, establish the Method and guardrails, obtain human approval, and make the capability available through Ember or the appropriate business workflow.

Observe → Learn → Improve

Build on what the organization already uses

AI should improve existing work—not force everything to start again

Organizations already have document stores, CRM and HR systems, service desks, finance applications and established workflows. KB Sandbox helps teams understand where those tools should supply knowledge, provide live information or support a governed action. The Workbench can then guide the design and evaluation of the appropriate connector, API integration, MCP server or specialist agent.

Connect knowledge

Bring approved information from existing document stores and business applications into the knowledge available to the right people.

Connect live systems

Use APIs, connectors and webhooks when Ember or another approved capability needs current information from an existing system.

Add governed actions

Introduce MCP servers and specialist agents when AI should do more than answer—such as prepare a record, start a workflow or request an approval.

Existing tools and workflows → governed connections → approved AI assistance and action

Start with real work

Help people with activities they already perform

KB Sandbox does not begin by declaring an enterprise inefficient. It begins by finding where approved knowledge and well-governed AI can remove friction, improve judgment and preserve useful learning.

A salesperson prepares for a customer meeting.

A proposal team assembles an evidence-backed solution.

A call-center employee answers a technical question.

An employee understands an approved company policy.

A consultant designs and evaluates an MCP server.

A student develops a controlled, testable AI agent.

A team collaborating around documents and a shared AI knowledge workbench — AI adoption becomes organizational capability when people can work from shared evidence, learn together and improve the way work gets done.

Governance that enables adoption

The safe path must also be the useful path.

Employees should not have to choose between productivity and governance. Ember gives them a sanctioned, useful interface grounded in the knowledge available to their Project.

The organization can govern which knowledge, people, models and agents participate; evaluate important results; preserve evidence; and keep consequential decisions under explicit human authority.

Why “Sandbox”?

A safe place to learn what deserves to become normal work

New AI uses should begin within clear boundaries. A sandbox is somewhere people can try an idea, compare alternatives, challenge assumptions and learn without confusing a convincing answer with a trusted one—or a successful demonstration with a production-ready capability.

The sandbox is not the final destination. Successful approaches can become approved knowledge, repeatable Methods and governed agents used through Ember and existing business workflows. Evidence and feedback from everyday use then return to the Workbench so the organization keeps learning.

Explore further

Go deeper—learn about the latest AI trends here

These links open material published for public viewing. Some organizational knowledge is protected; sign in first to see the additional Wiki and Project knowledge available to you.