Grounding in real knowledge
Knowledge bases act as a source of truth for answers, analysis, generation, and validation so teams can work from documents, evidence, and prior artifacts they already trust.
// Technical Overview
How Hey Zoran works: grounded AI workflows with reviewable outcomes.
Hey Zoran is an agentic workflow platform that adds structured memory, evidence, research, validation, and decision support on top of LLM workflows. It is designed for organizations, not isolated prompts, and it keeps work grounded, reviewable, and auditable.
Built for organizations that need evidence-backed work, the platform brings decisions, research, validation, generation, and knowledge-backed collaboration into one working system.
// Overview
Grounded AI workflows for teams that need clarity, continuity, and control.
Across the platform, teams start in a dedicated workflow, bring the right context into the run, and keep the result as a reusable artifact with history and provenance.
What Hey Zoran is
The platform helps teams work with AI using their own documents, reusable workflows, live web research, and structured collaboration. Every workflow is designed to keep evidence visible, preserve context, and produce outcomes teams can review, reuse, and trust.
How the platform works in general
- 1
Start work in a specific surface such as Chat, Document Search, Web Research, Validate Content, Content Studio, Prompt Recipe, or Decisions.
- 2
Provide context such as an assistant, one or more knowledge bases, knowledge graphs, or a decision record.
- 3
Retrieve relevant internal or external context, run the appropriate AI workflow in the background, and track progress as a structured run.
- 4
Store the result as an artifact with history, provenance, and reusable context instead of letting it disappear into a single chat bubble.
- 5
Review, refine, reuse, or publish the outcome depending on the workflow.
// Core themes
Evidence, continuity, and reviewability shape every workflow.
These principles guide how teams search, research, validate, generate, and make decisions in Hey Zoran.
Structured memory and continuity
Hey Zoran preserves working context, summaries, artifacts, and event history so useful context accumulates over time instead of resetting on every interaction.
Long-running workflows with visible progress
Many workflows are asynchronous runs that retrieve, analyze, generate, and synthesize over time, making the platform a fit for work that is too important to live inside one message.
Auditability and human review
Runs, analyses, events, and outputs are stored with traceable history so teams get confidence, transparency, and better control over what the system produced.
// Collective Decision Space
A structured, auditable workspace for evidence-driven team decisions.
The decision space turns a messy mix of comments, documents, and opinions into a first-class workflow with evidence, branches, analysis passes, and an explicit accepted outcome.
What it includes
- A clear decision record.
- Linked evidence and source material.
- Branches for alternative lines of thinking.
- Structured contributions from collaborators.
- AI-generated analysis that is reviewable pass by pass.
- An audit trail of how the decision evolved.
- A clear record of what was actually accepted.
How a decision starts
Each decision begins with a decision record. The team defines the decision, adds context, and links the relevant sources so the decision becomes a persistent working environment for the work ahead.
Knowledge basesKnowledge graphsReferenced decisionsPrompt recipe runsGeneration runsWeb research runs
Branch-based deliberation
A branch is a first-class line of exploration. Teams use branches to compare alternatives, test paths, and capture structured contributions instead of flattening everything into a single comment thread.
OptionsAssumptionsRisksCostsImplicationsObjectivesValuesEvidenceProposalsComments
The world model concept
The decision space builds an evolving model of the evidence around a decision. It pulls together accepted evidence, claims, contradictions, relationships, coverage, and linked source context so analysis stays grounded in what the team knows, what conflicts, and what is still missing.
Structured analysis workflow
The assistant analyzes a decision in passes, not in one opaque leap.
This keeps analysis grounded in evidence, tradeoffs, contradictions, and gaps before a team commits to an outcome.
1
Source audit
Identify which evidence is available and how strong or relevant it is.
2
Claim proposal
Surface what the available evidence appears to say.
3
Relation proposal
Map how claims, options, and criteria connect.
4
Factor modeling
Highlight the variables or decision drivers that matter most.
5
Scenario analysis
Model what different paths might lead to.
6
Contradiction audit
Show where evidence conflicts or reasoning breaks down.
7
Gap audit
Point out what is still missing or unresolved.
8
Recommendation
Produce a grounded recommendation or next-step view based on the analysis.
Human review is central
The most important rule is simple: the assistant proposes, humans decide. AI-generated decision analysis is stored as reviewable output, and the system preserves both what the AI suggested and what the humans approved.
- AI-generated analysis is stored as reviewable output that people can accept, reject, or edit.
- The reviewed version stays separate from the original generated version so the record preserves both the proposal and the human-approved outcome.
- Participation, change flows, and important mutations are traceable over time.
- The accepted branch and accepted option remain explicit so downstream work knows what was actually decided.
- Completed decisions can publish a clean summary without exposing the entire internal working process.
Research inside the decision workflow
The decision space is not isolated. Teams can gather more evidence, generate supporting material, validate assumptions, and enrich the record without leaving the decision context.
Chat with sourcesContent generationPrompt recipe runsValidationDocument searchWeb research
Teams can bring internal evidence, external research, and a clearly accepted outcome together in one decision record.
// Workflow surfaces
Each workspace is tuned for a different kind of evidence-backed work.
Each workspace is designed for a distinct job, from source-backed search and validation to external research, structured generation, and reusable runs.
Chat with sources
Assistant-backed conversation grounded in organizational context.
Why it matters
Use it to ask questions against internal knowledge, explore documents conversationally, brainstorm with a configured assistant, and keep history as an ongoing thread.
How it works
Users select an assistant, optionally add source context, and the platform retrieves relevant information to support the exchange.
Document Search
A retrieval-first workspace for finding relevant information inside knowledge bases.
Why it matters
It is built for teams that want direct, evidence-oriented access to relevant material, excerpts, and supporting information quickly.
How it works
Users search across one or more knowledge bases and receive results grounded in the selected document collections.
Validate Content / Align
A workspace for checking whether a piece of content, reasoning, or output aligns with selected context.
Why it matters
It helps teams move from merely generated content to content that can be defended against a source of truth.
How it works
Users provide content plus the relevant assistant, knowledge bases, or knowledge graphs, and the platform returns an alignment-oriented evaluation.
Web Research
An organization-scoped workflow for structured external research when the needed context is not already in internal knowledge bases.
Why it matters
It stores the query, summary, normalized source list, optional metadata, and links back into decisions when needed.
How it works
A research run gathers live external information and turns it into a reusable artifact instead of a temporary browser session.
Content Studio / Generate
A template-driven generation workspace built for structured creation instead of casual prompting.
Why it matters
It helps teams produce repeatable deliverables with sections, templates, and organized runs that are easier to standardize and review.
How it works
Users start from a template, define the run, and the platform generates structured outputs section by section.
Prompt Recipe Runs
Reusable AI workflows built around recipes rather than one-off prompts.
Why it matters
Prompt Recipe is process-driven: users choose a repeatable method and run it against selected knowledge bases.
How it works
A user chooses a recipe, selects context, launches a run, and the platform stores the result as a reusable run artifact.
// Knowledge and memory
The knowledge layer is what keeps the workflows grounded and reusable.
Knowledge bases, knowledge graphs, and preserved artifacts are what let the system carry real organizational context across time instead of starting from scratch on every run.
Knowledge bases
Knowledge bases are collections of documents used for retrieval and grounding. They are the backbone of many workflows because they turn organizational documents into usable context for AI work.
Knowledge graphs
Knowledge graphs add structured context for workflows that need relationship-aware information beyond plain document storage. They support the broader idea that Hey Zoran can work with both document-based and structured context.
Memory, traceability, and reusable artifacts
The platform preserves run history, analysis artifacts, event trails, structured outputs, and reusable context so work can be revisited, reviewed, linked into future workflows, or reused inside a decision.
What this means
Hey Zoran is a workspace for grounded AI work with traceable, reviewable outputs.
It does not auto-decide for the team; it structures evidence, research, validation, and outputs so humans stay in control.
It is designed to keep context visible and reusable across time instead of losing work in one-off prompts.
See it in action
Bring structured, evidence-backed AI work into your organization.
Talk with the team about decisions, research, validation, and structured generation workflows built around your documents, operating context, and review requirements.