THEGENERAL COUNSEL
GC OF 2035 · AUTHORITY & EVIDENCE ENGINE

From business event
to permitted action.

Thabo Mokoena, South Africa General Counsel, connects legal books and source material, project evidence and supervised AI agents. The tools help find and organise; a human remains accountable for judgment and approval.

Thabo Mokoena at work with legal books and AI assistant panels
Thabo Mokoena · fictional South Africa General Counsel
Prototype using synthetic information. This page demonstrates the proposed workflow. It is not connected to legal sources, client records, a live DOA, IFC/BIM, OC4IDS, AI agents or project systems. Do not enter confidential information.
Open obligations123 need a human decision
Evidence gaps042 could affect payment
Authority requests021 agent request paused
Learning signals03One new scenario ready
SAMPLE PROJECT · SOUTH AFRICA

Moya Health Infrastructure · Community Hospital

Illustrative NEC-style project record · all names, amounts, dates and rules are fictional.

ConstructionCorporate approvalsDemo data
The GC’s queue

Three items need a decision.

Sorted by consequence and time pressure. A person remains accountable for legal and commercial decisions.

!
Site instruction may change scopeNotice evidence missing · package B-14 · owner: Project Manager
Act today
↗
AI agent requested a variation approvalAbove its configured authority · request paused for GC approval
Human gate
▤
Board action needs an evidence linkDraft resolution recorded · source pack not attached
Evidence gap
Where this could connect

One control layer across legal and operational systems.

Document intelligenceOpenContracts

Candidate for document parsing, annotation and evidence links.

Construction dataIFC + OC4IDS

Standards and tools to structure project and infrastructure information.

Workflow engineTemporal

Candidate to run approvals, reminders and long-running workflows.

AI qualityPromptfoo + Langfuse

Candidate tools for red-team tests, evaluation and traces.

These are candidate components, not installed integrations. Licences, dependencies, security and data residency need review before adoption.

Candidate open-source foundation

Reuse proven components; build GC’s African control logic.

CandidatePotential roleGC-specific work still requiredAdoption note
OpenContractsDocument parsing, annotation, evidence linksJurisdiction-aware source graph and review rulesMIT-licensed project; verify dependencies and data handling
DocassembleGuided intake and document assemblyGC templates, approval gates and local legal contentMIT-licensed project; evaluate a bounded workflow
TemporalDurable deadlines, approvals and retriesContract and DOA workflow definitionsMIT-licensed; adopt when workflows need durable execution
Promptfoo + LangfuseAI testing, red teaming, traces and prompt evaluationGC playbook tests, human review and incident responseMIT core/project licensing shown by project sources; verify feature scope
IfcOpenShell + OC4IDSConstruction model data and infrastructure project structureMap evidence to project, contract, event and model identifiersIfcOpenShell LGPL; OC4IDS is a data standard, not a turnkey app
PM4PyAnalyse process logs for bottlenecks and control deviationsDefine meaningful legal/process indicatorsAGPL-3.0 open-source version; isolate and review before SaaS adoption

License labels are an initial screen, not a clearance opinion. Review exact versions, bundled dependencies, hosting and intended use before adopting code.