Platform · Management intelligence
Help your leadership team see, decide and learn faster
The Enterprise Brain is designed as a reasoning layer over everything One3 runs. It turns events into signals, explains why things changed, proposes options, and tracks whether decisions worked.
Why the Brain is the core of One3
Every other One3 app answers what can the company do? or who can do it? The Enterprise Brain answers the questions a management team is paid to answer: what is happening, why, what should we do, did it work, and are we getting better at deciding.
The Brain is not a separate product with its own data. It is a layer of reasoning services over the kernel's Objects, Actions, Events and Policies, the Context Graph, the semantic layer and the event log. Keeping it inside the platform keeps context, trust and execution in one place.
Honesty principle. The Brain is designed to say "I don't know", "this is correlation, not causation" and "run an experiment first". A management AI that sounds confident and is wrong loses trust faster than no AI at all.
From dashboards to decisions
| Question | Traditional software | Enterprise Brain |
|---|---|---|
| What is happening? | Dashboards | Signal Hub turns events into ranked, explained signals |
| Why is it happening? | An analyst spends a week | Root-cause engine with stated confidence |
| What should we do? | Meetings and opinions | Decision Engine: options, expected value, risk, simulation |
| Do it | People open six apps | Executor runs Actions through people and agents under governance |
| Did it work? | Rarely checked | Outcome tracking against an expectation recorded before acting |
| Are we deciding better? | Never measured | Calibration Engine scores humans and agents |
Sense → Understand → Decide → Act → Learn → Govern
The Brain runs one cognitive loop over the whole business.
- 1
Sense
The Signal Hub reads the event log, connector streams such as bank, ads, marketplaces and ERP, and human inputs from meetings and notes. It ranks what changed and explains why it matters.
- 2
Understand
The World Model combines the Context Graph, the semantic layer and metric tree, and organisational memory into a queryable picture of the business and what depends on what.
- 3
Decide
The Goal Stack and Strategic Planner feed the Decision Engine. It frames options, runs simulations on the Business Twin, and is challenged by a Critic agent before anything is recommended.
- 4
Act
The Governor decides whether an Action runs within autonomy, needs approval, or is forbidden. The Executor turns approved plans into projects, tasks, automations and agent runs.
- 5
Learn
The Outcome Tracker compares results with the pre-registered expectation. The Calibration Engine and Decision Journal feed validated learnings back into memory.
- 6
Govern
Board policy, a Company Constitution, a delegation-of-authority matrix, autonomy budgets, the autonomy ladder and Action risk tiers bound every step.
Autonomous loops at four clock speeds
| Loop | Clock | Example work | Autonomy ceiling |
|---|---|---|---|
| Operational | Seconds to minutes | Route a lead, reorder stock, reply to a ticket, pause a failing ad | Up to A4 for R1–R3 Actions within the autonomy budget |
| Tactical | Daily to weekly | Shift budget between campaigns, rebalance workload, adjust prices within a band | Up to A3 within the department budget |
| Strategic | Monthly to quarterly | Re-allocate budget across initiatives, start or stop initiatives, hiring plans | A2 maximum: agents draft, people decide |
| Transformational | Yearly or event-driven | Strategy change, new markets, business model, org design | A1 maximum: agents analyse, people decide |
Faster loops get narrower authority
Each loop is a versioned Pack artefact with a contract: what it observes, how it decides, which Actions it may take, its autonomy ceiling and budget, guard limits such as a maximum change per run, rollback rules, stop conditions and a human reviewer. Every loop run writes a Decision Journal entry, even when the decision is "do nothing".
Decision intelligence
The Decision Engine is planned for W3 (alpha) and W4 (general availability).
Root-cause engine
Metric-tree decomposition, segment drill-down with multiple-comparison control, graph traversal, timeline alignment and causal checks, ending in ranked hypotheses.
Evidence labels
Every hypothesis is labelled observed correlation, quasi-experimental evidence or experimentally confirmed. The Brain may not say "caused by" below quasi-experimental.
Options, not one answer
Options always include "do nothing". Each carries a P10/P50/P90 impact range, cost, time to effect, reversibility and risks.
Critic agent
Checks for missing options, optimistic estimates, goal conflicts and second-order effects. The Critic never executes anything.
Experiment first
When evidence is weak and the decision is reversible and testable, the Brain proposes an experiment: A/B, holdouts, geo tests, switchbacks or stage-gated pilots.
Decision Journal and calibration
Brier scores, reliability curves, forecast error and interval coverage show how good forecasts and decisions are, per human, agent, model and method.
Business Twin maturity levels
| Level | Name | Capability |
|---|---|---|
| L0 | Descriptive | Unified metrics and entity views with data-quality scores |
| L1 | Diagnostic | Root-cause engine explains variance with evidence levels |
| L2 | Predictive | Forecasts with intervals for core KPIs, backed by published backtests |
| L3 | Prescriptive | Scenario comparison and recommendations with expected value |
| L4 | Interventional | Simulated interventions validated against experiment results |
| L5 | Continuous autonomous twin | Runs tactical loops within autonomy and re-calibrates, with governance approval |
What the Twin will not claim
A model that does not beat a seasonal-naive baseline is not used. Every simulation lists its assumptions. The Twin will not claim precise long-range forecasts in novel conditions (it shows scenarios instead), causal effects without experimental or quasi-experimental support, or predictions about individual employees' behaviour for employment decisions.
Cockpits and executive surfaces
Every surfaced number is designed to show freshness, a data-quality score and an evidence level. Every recommendation shows what would change it.
Morning Brief (W2)
Top signals, overnight changes, decisions awaiting you and today's risks, in app, email, WhatsApp or voice in Indian languages.
Weekly Business Review (W2–W3)
Metrics against plan, variance explanations with evidence levels, decisions and outcomes, and a pre-read deck.
CEO Cockpit (W3)
Goal tree health, KPI forecasts, cash runway, initiative portfolio, autonomy budget use and open strategic decisions.
Department cockpits (W2–W3)
Metric trees, running loops, experiments, agent scorecards, budget, approvals and staffing for each department head.
Board Portal (W3)
Board pack, strategy performance, risk register and a quarterly AI oversight report sourced from the audit log.
Meeting-to-Decision (W3)
Meeting transcripts become proposed decisions, owners, deadlines and assumptions, confirmed in one tap.
A Chief-of-Staff system, not an "AI CEO"
One3 does not sell an AI that runs the company. Fiduciary duty and accountability sit with directors and officers, and no current model deserves unbounded authority.
Instead, the plan is a team of executive agents that prepare, analyse, challenge, track and execute within delegated authority: a Chief of Staff agent, a Strategy agent, and CFO-office, CRO-office, COO-office and CHRO-office agents, plus a Critic. Each has clear limits. The Finance agent never moves money or signs filings. The People agent never takes individual employment decisions. The full executive team is planned for W5.
The Brain is also method-agnostic. OKR, Balanced Scorecard, Hoshin Kanri, a generic scorecard-and-rocks method, Lean, DMAIC, Theory of Constraints, increment planning, zero-based budgeting and scenario planning are planned as method Packs mapped onto one management cascade.
The Brain in plan numbers
- loop clock speeds
- 4
- Business Twin maturity levels, L0 to L5
- 6
- management method Packs
- 10
- executive agents, including the Critic
- 7
- target autonomous actions outside budget or Constitution
- 0
Questions about the Enterprise Brain
Will the Brain make decisions on its own?
Only within the limits you set. Operational loops can act within autonomy budgets. Strategic loops are capped at A2, where agents draft and people decide, and transformational loops at A1.
When will it be available?
It is built in waves. The Goal Stack, Morning Brief and Decision Journal v0 are planned for W2. The Decision Engine alpha, Critic and cockpits are planned for W3. Forecasting and simulation follow in W4, and Business Twin L3–L4 in W5.
How do I know a recommendation is trustworthy?
Each one shows its evidence level, data freshness, data quality and what would change it. The Critic agent reviews significant recommendations, and calibration scores show how accurate past forecasts were.
Can it score individual employees?
Calibration of individual people is private to that person by default. Aggregate views for managers depend on tenant policy and employment law, so verify with counsel per market. The Twin will not predict individual employee behaviour for employment decisions.
Is every autonomous decision auditable?
Yes, that is the design. Decisions, approvals, overrides and autonomous Actions go into a hash-chained audit log, and any autonomous decision can be replayed with the context and model versions used at the time.
Bring decision intelligence to your leadership team
Talk to us about the Enterprise Brain roadmap and how it fits your management rhythm.