DIH PM Module — Business Requirements

The Case for Change

Executive oversight can't scale on spreadsheets and goodwill

The CEO Office governs a growing portfolio of initiatives across multiple OpCos with no single control surface. Governance, risk and cost depend on manual diligence — and the decisions that steer the business are never captured.

TODAY

5

structural gaps an AI-native platform is built to close — each carrying delay, risk, and lost institutional memory.

Fragmented oversightManual governance Reactive riskDisconnected cost No decision logSpreadsheet-to-deck Siloed OpCosLost knowledge

26

CEO-Office digitalization requirements captured verbatim — none dropped, all traceable.

1

control surface needed across every department, OpCo and initiative — replacing scattered trackers.

4

strategic drivers: executive control, governance & compliance, predictive risk, institutional memory.

0

decision-grade AI actions taken without a human checkpoint — governance is the design constraint.

The objective is control and leverage, not more tools. One governed, AI-native core gives the CEO Office real-time visibility, automates the governance that today depends on individual diligence, and turns every decision into reusable institutional memory — while delivery teams keep the engine they already know.

The Approach

Build on what works — add the layer that's missing

Three inputs shaped the design: the CEO-Office requirements, the proven Augment delivery engine, and the best patterns from the leading PM platforms. The decision — one dual-mode core, governed AI, delivered in six months.

01

Executive Initiatives (charter-driven, OpCo-scoped) and Delivery Projects (site-based, plan-driven) share one governance, AI, reporting and audit core. The mode never forks the control plane — the CEO governs both from the same cockpit.

02

Keep the proven delivery engine — plans, Gantt, multi-site, evidence, audit — and add the missing executive layer: SLA & escalation, DOA approvals, an immutable decision register, KPI ↔ performance, and HRMS-linked cost.

03

Governed agents in every domain, grounded in the DIH AI OS, with an accept / edit / reject checkpoint on every decision-grade action. Everything runs inside the Microsoft 365 / Azure / Claude Enterprise tenant.

The trade-off, managed Building on Augment means DIH owns the build and integration effort a commercial vendor would otherwise carry. That is managed through the phased plan, the P0 governance foundation, and the risk controls in this document — in exchange for a platform shaped to DIH's exact governance priorities and grounded in its own knowledge graph.

The dual-mode object model

One configurable model carries every record type, viewed at executive and delivery altitudes — and the two modes share one governance, AI, reporting and audit core.

Portfolio

Program

Initiative / Project

Workstream

Task

Work objects nest and are scoped by Entity (OpCo / country / function) and Location.

GOVERNANCE

Decision (immutable) · Approval (DOA) · Risk / Issue / Change · SLA / Red-Flag

RESOURCING & FINANCE

Allocation · Resource · TimeEntry · Cost / Budget

PERFORMANCE & KNOWLEDGE

KPI · Outcome / Benefit · Document / Evidence · Comment / Meeting

MASTER DATA

Entity (OpCo / Country / Function) · Location / Site · Template · Tag

The Starting Point

The Augment delivery engine already covers the middle of the lifecycle

Augment is a mature multi-site delivery engine — strong in planning, scheduling, execution, evidence and audit; thin at the executive-governance ends; and not yet AI-native. Knowing exactly what exists prevents rebuilding it and frames every gap.

LIVE

Create-from-template, the setup wizard, locations and task assignment, plus faceted listing and saved views — the start of every delivery project.

LIVE

Milestone / list / board views, a status state machine (incl. a Pending-Approval gate), multi-assignee.

LIVE

WBS tree, dependencies, critical path, baselines, and forecast-vs-actual — a genuine scheduling strength.

LIVE

Tenant-wide site master with clone / bulk — the multi-site delivery model mapping to the physical estate.

ENHANCE

Progress, milestone DAG, critical path, overdue / upcoming. Rebuilt as a health composite with cost & AI signals.

ENHANCE

Typed exception workflows with state machines — generalized to a full issue & change-control register.

ENHANCE

Structured evidence keyed to task × location — folded into the unified documentation & knowledge archive.

ENHANCE

Immutable system-action audit and export — extended to full decision & audit lineage.

NEW

SLA, escalation, DOA approvals, decision register, predictive AI and agents — the missing executive layer.

The model: a configurable object model — Portfolio → Program → Initiative / Project → Workstream → Task — scoped by Entity (OpCo / country / function) and Location, with a task × location evidence store and pervasive audit. A strong delivery foundation; what it needs next is the executive-governance and AI layer.

P0 turns a delivery tool into an executive control system. Before new features: the dual-mode core, the governance & control plane (SLA, escalation, DOA approvals, red-flags, decision register, validation), the CEO cockpit, and the first governed agents on an AI control-center baseline.

AI-Native by Design

A governed agent workforce — every decision-grade action human-reviewed

Intelligence is specified into every domain and delivered through nine governed agents plus an ask-the-portfolio copilot — each bounded by an autonomy level and a human checkpoint, grounded in the DIH AI Operating System.

PLAN

AI-drafted charters, suggested KPIs & milestones, and effort & risk estimation.

AUTOMATE

Triggers, escalations, approval routing and document generation — policy-bound.

PREDICT

SLA-breach, slippage, cost-burn and milestone forecasts — with confidence & drivers.

OPTIMIZE

Over / under-allocation detection and within-policy rebalancing proposals.

CONVERSE

The ask-the-portfolio copilot — grounded, cited answers and natural-language views.

REMEMBER

Permission-aware Q&A over records, evidence and precedent; decisions made reusable.

DETECT

Anomaly & bottleneck detection and predictive risk scoring across the portfolio.

DECIDE

Scenario & what-if, precedent surfacing, and an immutable decision register.

Grounded, governed, and tenant-resident. Agents are grounded in the DIH AI OS / Relationship Graph; every decision-grade answer cites its sources; execution stays inside the tenant with no training on DIH data; and the evaluation gate is a release criterion enforced by the AI Steward.

The hard cap No agent — at any level — takes a decision-grade action (creating an approval, committing cost, altering a decision) without the named human checkpoint. Autonomy runs Advisory → Suggest → Draft → Act-with-override → Route → Act-within-policy, set and enforced per agent by the AI Steward.

Five-layer architecture

Retrieval, reasoning and action are deliberately separated — and governance wraps every layer. That separation is what keeps the AI auditable and explainable.

Experience

Role-aware cockpit, workspaces, dashboards, and the mobile field client.

Action / Orchestration

Workflow engine, agent orchestration, approvals and notifications.

Reasoning

AI services — drafting, prediction, classification, RAG — via Claude Enterprise & MCP.

Data / Retrieval

Object model, evidence store, knowledge index, and integration adapters.

Governance

Identity · RBAC + ABAC · audit lineage · evaluations & observability · guardrails — wrapping all four layers.

Process Architecture

How work flows — and how governance holds

One governed lifecycle carries both modes, with an agent at every stage. Beneath it, two control-plane state machines keep approvals and SLAs moving — and every transition is recorded in the immutable decision and audit log.

End-to-end lifecycle

Originate to close — the same governed lifecycle for executive initiatives and site-based delivery projects.

OriginateIntake Agent

ApproveApproval-Routing

Plan

ResourceResourcing Agent

Execute

MonitorRisk & SLA Agent

ReportStatus Agent

Close

Approval — Delegation of Authority

Draftsubmit

Pending Approvalapprove ≤ threshold

Approvedactivate · recorded

Active

Exceptions from Pending Approval: escalate → Higher Authority · reject → Returned (reason logged).

SLA & escalation

On TrackAI predicts breach

At RiskSLA missed / red-flag

Breachedunresolved in SLA

EscalatedCEO smart alert

Prediction is the point — the Risk & SLA-Breach Agent flags At Risk before a breach, so escalation is pre-emptive.

The Roadmap

Three increments over six months, milestone by milestone

Scope is sequenced so CEO-mandate-critical capabilities land first — P0 Govern & See → P1 Resource & Predict → P2 Remember & Extend. Click any milestone to see what it delivers; every increment runs discovery → sign-off → iterative build → UAT → OpCo pilot → release.

Why phased CEO-mandate-critical scope lands first. Independent gates let value arrive early, keep each increment piloted and signed off before the next begins, and hold scope to the approved envelope through a prioritization rubric (CEO-mandate × value × AI-differentiation × feasibility) and a full traceability matrix.

The Capability Model

Seven capability pillars across twenty configurable modules

DIH's requirements organize into seven capability pillars. This is the scope structure — each pillar mapped to the increment that delivers it, all on the P0 governed core.

1

Executive Governance & Control

SLA & escalation, DOA approvals, red-flags, decision register, validation, control dashboard.

P0

Key requirements

SLA & escalation — bind SLAs to assignments, escalation ladders, predictive breach, CEO smart alerts
Approvals (DOA) — multi-level + Delegation-of-Authority routing by threshold, at item / document / project level
Red-flag rules — configurable thresholds and escalation protocols embedded in the workflow
Decision register — immutable auto-log of CEO decisions, approvals and inputs
Validation cycle — every material update passes a completeness + AI check before it stands
Control dashboard — firm-wide oversight across all departments, OpCos and projects

Current build: none — the new governance & control plane. No equivalent exists in Augment today; delivered first in P0.

2

Strategy, Portfolio & Intake

Structured intake & charter, portfolio & program rollup, prioritization, cross-entity scoping.

P0 · P1

Key requirements

Structured intake — objectives, KPIs, owner, timeline, country and OpCo; AI-drafted charter
Auto codes — deterministic project & contact codes, linked across assets
Portfolio & program — rollup, heatmaps, cross-portfolio analytics, balancing
Prioritization — rubric-based scoring and the portfolio lifecycle (intake → review → rebalance)
Cross-entity scoping — OpCo / country / function dimension drives rollup and access

Current build: partial — templated creation exists in Augment; charter, KPIs, codes and portfolio rollup are new.

3

Planning & Delivery

WBS & templates, Gantt / critical path / baselines, tasks, multi-site execution, evidence.

P0

Key requirements

Planning — Project-Plan templates, plan-driven WBS, RACI, communication plan
Scheduling — Gantt, critical path, predecessor DAG, baselines, forecast-vs-actual
Execution & tasks — milestone / list / board views, status state machine, multi-assignee
Multi-site delivery — Task = Plan × Location across the physical estate
Documentation & evidence — task × location evidence store and comms archive

Current build: strong — this is the preserved Augment engine, enhanced with an AI schedule-risk overlay (P0).

4

Resource, Capacity & Cost

HRMS time → cost / resource center, hour tracking, capacity, computed financials, calendars.

P1

Key requirements

Resource & capacity — staffing, utilization, capacity-vs-demand, AI rebalancing
HRMS time — time and hours feed into the project cost & resource center
Hour tracking — by individual / team against assignments
Cost & financials — budgets, burn, and computed / derived financial fields
Calendars & leave — resource calendars and vacations per project timeline (HRMS-fed)

Current build: none — a new domain delivered in P1 on the Core foundation.

5

Performance & KPIs

KPIs bound to Performance Mgmt, monthly pack, assessment scorecard, variance, metric taxonomy.

P1

Key requirements

KPI framework — measures bound at charter, linked to the Performance Management System
Monthly review — auto-compiled, human-reviewed monthly KPI pack
Assessment — project assessment / evaluation scorecard per project
Variance — baseline original-dates reference and variance analysis
Metric taxonomy — portfolio, governance, delivery, resource, performance and AI metric families

Current build: none — KPI ↔ performance linkage and the metric taxonomy are new (P1).

6

Intelligence & Automation

Governed agents, predictive analytics, copilot, no-code agent builder, AI control center.

P0 → P2

Key requirements

Governed agents — nine agents + copilot, each with an autonomy level and HITL boundary
Predictive analytics — SLA-breach, slippage, cost-burn and attention scoring, with confidence
Ask-the-portfolio copilot — conversational, cited Q&A across the portfolio
No-code builder — admin-composed agents & automations, governed by the AI Steward
AI control center — agent catalogue, evaluations, observability, eval-gate release criterion

Current build: none — the AI-native layer; baseline agents in P0, depth and the no-code builder through P2.

7

Knowledge, Reporting & Cockpit

CEO cockpit, slice-and-dice, report packs, auto-briefings, knowledge & decision memory.

P0 · P1 · P2

Key requirements

CEO cockpit — live health (timeline, cost, resource, bottlenecks), drill-down, act-in-place
Slice-and-dice — saved executive views by entity, region, function, phase and status
Reporting — configurable report packs; auto-drafted narratives and executive briefings
Knowledge layer — Q&A, precedent library, benefits realization, lessons captured
Smart alerts — preference-aware, AI-prioritized notifications by topic / delay / interest

Current build: partial — basic dashboards and audit exist in Augment; the cockpit, briefings and knowledge memory are new.

Built once, reused everywhere Cross-cutting services — the configurability engine, RBAC + ABAC entity scoping, the approval & SLA engines, audit & decision lineage, the notification engine, and the AI control center — are built in P0 and reused by every domain above.

Current Build vs. Target

Where the Augment engine stands against the mandate

Every capability area traces to a pillar, a priority, its status in the Augment build today, and the increment that delivers it. A read on coverage.

Available today Partial — foundation exists Not yet built

Planning & scheduling (Gantt) Pillar 3 · Must · P0

P0

Tasks & execution Pillar 3 · Must · P0

P0

Multi-site delivery & evidence Pillar 3 · Must · P0

P0

Audit & activity log Pillar 1 · Must · P0

P0

Structured intake & charter Pillar 2 · Must · P0

P0

CEO cockpit & slice-and-dice Pillar 7 · Must · P0

P0

Governance & control plane (SLA · DOA · decisions) Pillar 1 · Must · P0

P0

Portfolio analytics & reporting Pillar 7 · Should · P1

P1

Resource, capacity & cost Pillar 4 · Must · P1

P1

KPI ↔ performance Pillar 5 · Must · P1

P1

Predictive AI & agents Pillar 6 · Must · P0–P2

P0–P2

Knowledge & decision memory Pillar 7 · Could · P2

P2

Bars indicate the share of each area's full requirement set already implemented today, read from the traceability matrix (BRD Appendix A). The 26 CEO-Office requirements remain individually traceable to the source documents by requirement reference (R-01…R-26).

Integrations & Data

The systems the platform connects to

API-first and event-driven — the platform consumes systems of record rather than rebuild them, each connection sequenced to the increment that first needs it. Everything runs inside the DIH tenant.

Integration	Purpose	First needed
Microsoft 365 (Entra ID, SharePoint, Teams, Outlook)	Identity & access, documents, collaboration, email and calendar	P0
Azure AI Search	Permission-aware retrieval index for knowledge & Q&A	P0
Claude Enterprise (via Model Context Protocol)	The reasoning layer over live platform data — tenant-resident	P0
Delegation of Authority (DOA)	Approval thresholds and routing rules	P0
DIH AI OS / Relationship Graph	Grounding and institutional context for every agent	P1
Performance Management System	KPI definitions and the monthly performance linkage	P1
HRMS	Time, cost, resource master, calendars and leave	P1
E-signature (e.g. DocuSign)	Approval and acceptance e-signature	P1–P2

Risks & Mitigations

The risks we are managing, and how

Each risk from the BRD's RAID log carries a concrete mitigation built into the plan.

Risk	Mitigation
Scope expansion across two modes	Strict P0 boundary; configurability defers customization; a full traceability matrix; phase-gate sign-off before each increment.
AI trust & adoption	Human-in-the-loop on every decision-grade action; explainability and visible accept / edit / reject loops; evaluation gates as a release criterion.
Integration dependencies (DOA · HRMS · Performance Mgmt)	Confirm interfaces early; integration spikes at design time; phase the integrations into P1 with manual fallbacks.
MNPI / data exposure	Tenant-resident AI with no training on DIH data; need-to-know gating; immutable audit from day one; a security review before every release.
Change management	Preserve the familiar Augment delivery patterns; role-based onboarding; OpCo pilots before each broad release.
Six-month timeline pressure	Ruthless MVP scoping — cut scope, never extend the increment; CEO-mandate-critical capability lands first.
Pending business inputs (DOA matrix · SLA catalogue · KPI defs · org hierarchy)	Close in discovery as a hard M0 exit criterion, so improvement KPIs and governance rules remain provable.

In short One governed core, the proven Augment engine, and a six-month plan with real gates. P0 delivers executive control and a live cockpit; P1 adds resource, cost and prediction; P2 builds institutional memory — every increment piloted and signed off by DIH as it lands.

See It Live

Explore the working demo

A live walkthrough of the DIH PM Module — the governed cockpit, the delivery engine, and the agent workforce in action.

Open the demo →