Intelligence, coordination and verification for TSES
Know what is verified. Decide what to approve. See the receipt.
Ohevai Brain reads the work recorded in ATM, the AI workforce in Ohevai Core, and the evidence in your source and build systems. It plans realistic daily work, proposes exact changes for human approval, and reports only what it can show.
Intended address brain.ohevai.com. Brain does not replace ATM or Ohevai.com; it is their independent intelligence and governed action layer.
What Brain does today
A daily operating loop for managers and staff
Brief → inspect capacity → draft plan → approve → assign in ATM → observe progress → review evidence → verified result → report. Every step keeps its source, version and freshness visible.
Designed and prototyped
Role-aware planning
Trainers, developers, QA, project managers and DevOps see the result expected of them, the acceptance criteria and the evidence that counts for their role. Unknown availability is shown as unknown, never as free time.
Independent verification
Reported completion and verified acceptance are different states. Each criterion shows expected versus actual behaviour, the artifact, build and environment it is bound to, and the accountable reviewer. Missing evidence blocks acceptance.
Accountable decisions
Brain recommends; people approve. A proposal shows the exact before and after, the objects and versions it touches, who may approve it and when it expires. Execution returns a per-item receipt, and a timeout is reported as unknown, not as success.
The connected ecosystem
Each product owns its records. Brain references them.
Brain keeps stable source identifiers and versions. It never duplicates another product's database, and a change to an authoritative record happens in that product through a supported, approved command.
| Product | Owns | Brain's relationship |
|---|---|---|
| Ohevai Brain | Recommendations, orchestration, independent verification, approved-action coordination, source-grounded memory, receipts | This product |
| ATM | Human-work tenant records: staff mappings, assignments, daily work, evidence and review, formal evaluations | Reads projections; proposes assignment changes for approval; never edits evaluations |
| Ohevai Core / Ohevai.com | AI staff and hubs, specialist service invocation, model and tool policy, AI usage | Pairs a workspace to a hub; invokes only invocation-ready specialists within mandate |
| College | Learning delivery and outcomes | Maps learning to applied work; shows unverified API state explicitly |
| GitHub / CI | Source, changes, builds and test artifacts | Reads artifacts to bind evidence to commits, builds and environments |
| Production providers | Deployment and operational receipts and signals | Reads receipts; a release is executed by the provider under human authority |
The Living Brain
Six functional regions. Every pulse is an observed event.
The Living Brain is an operational map of authorised source events, grouped by product function. Selecting a region or an event leads to the underlying work, evidence or decision. It runs in three visibly different modes: Live, Replay and Simulation.
Regions are product functions, not anatomy. Animation shows observed activity and its source; it does not claim cognition, certainty or verified work. A keyboard-accessible region and event list and a static alternative are always available.
- Coordination
Plans, blockers, proposals - Workforce
Capacity, assignments, availability - Engineering
Source changes, builds, releases - Learning
Skills and College outcomes - Memory
Records, decisions, provenance - Governance
Approvals, policies, audit
The long-term vision · Enterprise App Factory
Intended capability — not implementedCreate and operate complete enterprise applications through governed roles
In time, Brain's agents will act as solutions architect, full-stack developer, QA, project manager and DevOps engineer, each within an explicit mandate, source scope and budget. Every stage has a gate, every gate needs evidence, and consequential steps need a named human authority.
- 01DiscoveryProblem, users, sponsor, classification
- 02ArchitectureTopology, contracts, ADRs, isolated cell
- 03DesignStories, criteria, journeys, frames
- 04BuildAuthorised workspace, bounded runs, diffs
- 05TestQA, security, AI assurance evidence
- 06ReleaseManifest, human approval, receipt
- 07OperateSLOs, incidents, recovery, improvement
The design preview simulates these stages so the workflow can be reviewed. The delivery engine itself requires backend, tool and provider implementation; no button in the preview deploys anything.
Governance
Distinctions the product refuses to blur
Source context
Every figure carries its period, denominator, source and freshness. Visibility of a source is not the same as its freshness or its ability to execute.
Human authority
A recommendation is not an approved action, and an approval is not a successful execution. Authority has a name, a scope and an expiry.
Traceability
Decisions, runs and receipts are appended, never rewritten. A rejection after completed work creates linked remediation; history stands.
Product independence
Brain's shadow insights are advisory. Formal evaluations stay in ATM. No leaderboard, no activity score, no automatic employment or pay decision.
Available in the design preview Prototyped
- Role-aware navigation and six home variants
- Command Centre, Today brief, grounded Ask with citations
- Capacity planning with known/unknown availability
- Criterion-level verification with evidence binding
- Exact-change approvals with per-item receipts
- Living Brain in Live, Replay and Simulation modes
Planned — requires implementation Planned
- Authenticated sign-in and SSO handoff
- Live ATM, Ohevai Core, College and CI pairing
- MCP tool gateway with provider-side authorisation
- Agent execution within mandates and budgets
- Enterprise App Factory delivery engine
- Scheduled report delivery