AIDoo · Reason & Act
Status: Coming SoonBuild AI that can reason, use tools and act.
Create RAG applications, AI agents and intelligent workflows connected to your data, processes and deployed models.
- What it is
- AIDoo is designed as the place where language models meet the rest of your platform: governed data to retrieve from, FlowDoo workflows and MLDoo models to call as tools, and a record of every step an agent took.
- Who it is for
- AI engineers and developers building assistants and agents that have to work with real company data and take real actions.
- Status
- Coming Soon. In development. Capabilities described are the design intent.
The problem
An assistant that cannot reach your data or your systems is a demo.
Agents are built in isolation, then wired to data and tools with custom code that nobody governs.
Answers arrive without sources, so nobody can check them.
An agent that can act is an agent that can act wrongly — and most stacks have no approval step and no trace.
Core capabilities
What AIDoo is designed to do.
AIDoo is not yet released. These are the capabilities it is being designed around.
AI gateway
One governed entry point to the language models your organization uses.
Model routing
Send each task to a suitable model by cost, capability or policy.
RAG
Ground answers in your own documents and data instead of the model's memory.
Vector and hybrid retrieval
Combine semantic and keyword search to find the right context.
Citations
Show which sources an answer was built from so a reader can check it.
Agents
Compose agents that plan, call tools and report what they did.
Tools
Give agents well-defined actions with typed inputs and outputs.
Prompt management
Version prompts and see which version produced which result.
Evaluation
Test prompts and agents against cases before they reach users.
Guardrails
Apply policies to what goes into a model and what comes out.
Human approval
Pause an agent before a consequential action until a person agrees.
Model calling
Let an agent call a deployed machine learning model like any other tool.
FlowDoo execution
Turn a FlowDoo workflow into an action an agent can take.
MLDoo integration
Reason over a prediction together with the explanation behind it.
How it works
How it is intended to work.
Step 1: Connect knowledge
Point AIDoo at the documents and datasets an assistant is allowed to use.
Step 2: Define tools
Expose FlowDoo workflows, MLDoo models and APIs as actions with clear inputs and outputs.
Step 3: Compose the agent
Set its instructions, its tools, its guardrails and where a person must approve.
Step 4: Evaluate, release, observe
Test against cases, release it, and review the trace of what it retrieved, called and did.
Example workflow
What it looks like in practice.
The design intent: an answer you can check, and an action someone approved.
- Question
- Retrieve context
- Reason
- Call a model
- Human approval
- FlowDoo action
Part of DooHorizon
How AIDoo connects to the platform.
AIDoo is not a standalone chatbot layer. It is the intelligence execution layer of the DooHorizon platform.
FlowDoo
FlowDoo workflows become the actions an agent can take — and the automation that runs after it decides.
About FlowDooMLDoo
Deployed MLDoo models become tools; their explanations become context an agent can reason over.
About MLDooVizzDoo
Governed datasets and dashboards become sources an assistant can answer from and cite.
About VizzDooGovernanceDoo
Status: Coming SoonAgents, prompts and tool permissions are designed to be governed assets with owners, policies and an audit trail.
About GovernanceDoo
Security and governance
Governed like the rest of the platform.
Designed with enterprise security and governance requirements in mind. No certification is claimed that has not been obtained.
How DooHorizon approaches security- Designed so an agent can only use the data and tools it has been granted.
- Designed to keep a trace of each run: the prompt, the context retrieved, the tools called and the action taken.
- Human approval is a first-class step, not an afterthought.
- Designed so the choice of model provider — and what is sent to it — is visible and configurable.
Use cases
What you can build with AIDoo.
Intelligent support
Answer from your knowledge base with citations, and open or update a case through a workflow.
Operations copilots
Let staff ask for a report, a reconciliation or a status in plain language and get a governed result.
Decision support
Combine a model's prediction, its explanation and policy context into a recommendation a person approves.
Document-heavy processes
Read incoming documents, extract what matters and route the case.
The concept
What AIDoo could look like.
- Prompt“Which enterprise accounts are at risk this quarter, and what should we do?”
- Context3 sources retrieved and cited: renewal policy, account notes, support history.
- Tool · MLDoo modelchurn-risk scored 212 accounts; top drivers returned with each score.
- Human approvalProposed outreach list reviewed and approved by an account director.
- Action · FlowDooWorkflow “Open retention cases” started for 9 accounts.
FAQ
AIDoo: common questions.
Does AIDoo support RAG and agents?
AIDoo is intended to provide RAG, intelligent agents and AI execution capabilities, subject to the current product release.
Is AIDoo available today?
Not yet. AIDoo is in development; what this page describes is the design intent. Join the roadmap list to be told when early access opens.
Which language models will AIDoo work with?
AIDoo is designed as a gateway, so that organizations can choose and change model providers. The supported list will be published with the release.
Is AIDoo just a chatbot?
No. It is designed as an execution layer: retrieval, tools, approvals and traces around the model, connected to the rest of DooHorizon.
Follow AIDoo from concept to release.
Start with one use case. Expand across the ecosystem.