How rygg works

From scattered knowledge to answers you can trace

rygg turns what your organization knows about its users into a maintained graph. Every statement has a source, a status and an owner. People and agents work with the same answers, and people decide what counts as knowledge.

OntologyL1 · licensedContextL2 · yours, maintainedEvidenceL3 · yours, maintainedAI-found pattern
Three layers on one backbone. Cyan marks what AI contributed.
01 · The cycle

Five steps, repeated with every new question

rygg is not built once and handed over. It runs in a cycle, and each round leaves the graph more complete and more reliable.

  1. 01Gather

    Knowledge comes in, in whatever form it exists

    Field research, usability tests, interviews, existing reports, wikis, spreadsheets, requirement documents. Nothing has to be prepared or restructured before it enters rygg.

    Research & Design Engineers, your teams
  2. 02Structure

    Statements are mapped to the managed ontology

    Each statement is broken down into nodes and relations that the ontology defines. AI does the first pass. Every node keeps a link to the source it came from.

    AI, checked by Research & Design Engineers
  3. 03Verify

    People decide what becomes part of the graph

    AI contributions stay suggestions until a person has approved them. Contradictions with existing statements are flagged and resolved, not overwritten.

    Research & Design Engineers, domain owners in your teams
  4. 04Maintain

    The graph is released, not just edited

    Changes are bundled into releases with a change log. Outdated statements are marked. Every release is validated against HCRD standards and comes with a coverage report.

    leefs, accountable for correctness
  5. 05Use

    People and agents work with the same answers

    Teams query the graph through an interface built for their questions. Agents query it through an API. Questions that the graph cannot answer yet become research questions, and the cycle starts again.

    Your organization, your agents
02 · Provenance and approval

You can see who contributed what

Every node shows whether it comes from a person, from an AI suggestion that still waits for approval, or whether a person has verified it. AI suggestions always list the human sources they rest on, across the three layers: ontology, context and evidence. Try it: approve the suggestion below.

REQ-114 · Braking distance on wet roadsAgent · awaiting approvalCompleted
AI SUGGESTION · TC-31Create test case “Braking distance on wet roads”.
PROVENANCESafety requirementL1 OntologyREQ-114, wet roadL2 ContextFinding P-22L3 Evidence
TC-31 verified · M. Berg. The node is now part of the graph.

  • HumanCreated or owned by a person
  • AI workingAn agent is reading or writing in the graph
  • AI suggestionBacked by provenance, not yet approved
  • VerifiedA human has confirmed it. The node becomes part of the graph.

The rygg symbol marks verified nodes: the human shoulder frames the AI core.

Suggested by agentday 1 · coverage agentLinked to sourcesREQ-114FMEAP-22day 1Reviewed · M. Bergday 2 · edited titleVerifiedday 2 · now part of the graphUsed in agent answer · REQ-120Used in agent answer · REQ-140
The life of one node, from suggestion to use. Sample data.
03 · Evidence status

Evidence and assumption are kept apart

Each statement carries a status, a date and a source. That is how the graph can answer how much of what you believe is evidence, and which assumptions matter most to check next.

Observed
Seen in the field: observation, interview, diary study.
Tested
Confirmed or contradicted in a usability test or evaluation.
Derived
Concluded from observed or tested statements, with the reasoning recorded.
Assumed
Not yet backed by evidence. Visible as an assumption, with an owner.
04 · From context of use to metric

Every metric traces back to an observation

Every requirement traces back to an observed need, every solution to a requirement, every evaluation to a decision and every metric to a finding. Every node is a Markdown file in the Open Knowledge Format (OKF) v0.2, readable by people and agents.

  1. 01Context of useWho works on what, under which conditions, and where does the work break down?
    • Conditions
    • User groups
    • Tasks
    • Current scenarios
  2. 02User requirementsWhat do the user groups need?
    • Needs
    • User requirements
  3. 03SolutionWhich elements meet the requirements, and which decisions shape them?
    • Design solutions
    • Design decisions
  4. 04EvaluationHow is it tested, and what was confirmed or contradicted?
    • Evaluation measures
    • Evaluation findings
  5. 05MetricsHow does the organization see effectiveness, efficiency and satisfaction?
    • Metrics
05 · Roles

Who does what

The graph is a shared responsibility with clear lines. Your teams own your content, we are accountable for the graph being correct, and agents never approve their own work.

Your teams

  • Ask questions and bring in existing knowledge
  • Own company-specific content
  • Approve statements in their domain
  • Decide on the basis of the graph

Research & Design Engineers from leefs

  • Maintain the ontology and the method network
  • Close gaps with research in the field
  • Verify AI contributions and resolve contradictions
  • Release, validate and report

AI agents

  • Read the graph instead of guessing
  • Suggest new nodes and relations with provenance
  • Never approve their own suggestions
  • Report questions they could not answer
06 · Access and ownership

Hosted by us, readable by people and agents, exportable by you

We run rygg as a hosted service. Your company-specific content and research results are yours. The managed ontology, the method network and our tools stay with us and are licensed.

Query interface
For teams: search, browse and follow a statement back to its source.
API for agents
For agents and tools, for example as an MCP server. Every answer includes provenance.
Reports
Coverage report and change log per release. Metrics review in the Context Layer package.
Open format
Nodes are Markdown files in the Open Knowledge Format (OKF) v0.2. Your content can be exported at any time.
07 · Questions

Frequently asked

Is rygg a search over our documents?

No. A document search returns text passages. rygg returns statements that are typed, linked, sourced and verified, and it shows what is evidence and what is assumption. Documents are one of the inputs.

What do we need to start?

One use case and the knowledge you already have, in whatever form. If there is no usable research yet, we start with a context of use analysis.

How long until the graph is useful?

The pilot takes four to eight weeks and ends with a working graph for one slice, a coverage report and a decision.

Who owns the content?

Your company-specific content and research results are yours and can be exported in an open format. The managed ontology, the method network and our tools stay with us and are licensed.

Could we build this ourselves?

The software, yes. What takes time is a domain model that stays correct as products, users and markets change, and someone who is accountable for it. That is what we offer.

Does rygg replace research?

No. It makes research reusable and shows where research is missing. New questions still need to be answered in the field.

Where does AI come in?

AI speeds up structuring, comparing and suggesting. People approve what enters the graph. On every node you can see whether a person or an AI contributed it.