1.1 Key Discovery Findings
These are synthesized from AES's assessment replies and follow-on validation; they explain how the current-state evidence shapes the proposal.
- Peer-validated pattern: peer firms already operate internal knowledge assistants and proposal accelerators; AES is positioned to build the airport-engineering version through disciplined, governed delivery.
- Foundation gap: AES has a strong Microsoft foundation, but information hygiene, findability, version control, permissions and scanned PDFs need work before scale.
- Residency reality: KSA, SSI (Sensitive Security Information), NDA and general corporate content need classification-based routing rather than one blanket hosting answer.
- Agent identities are viable: Microsoft agent identity capabilities make AES's goal of named digital workers achievable, with fallback patterns where features are still maturing.
- Calibrated promise: the proposal commits to measurable workflow gains, better findability and stronger auditability — every promise verifiable at a gate. Speculative claims such as win-rate uplift are deliberately excluded.
1.2 Organisational Context
Organisational Baseline
| Area | Current state | Project implications |
|---|---|---|
| Scope | Group-wide AI Brain across AES Global / International, with offices in Dubai, Riyadh, Jeddah, Atlanta, Madrid, Amman, Lebanon and Tunisia. | The platform is built once, then activated locally by entity, compliance pack, champions and rollout wave. |
| Business | Airport consultancy across design, master planning, aerodrome certification, PMC / PM-CM, PMO, ORAT (Operational Readiness, Activation and Transition), airline support, condition assessment, investment planning and PPP. | The Brain needs airport-engineering context, proposal knowledge, project documentation, policies and regulatory references, not generic office search. |
| Decision stakeholders | MD sponsor / approver, Executive Director, AI workstream lead, AI engineer, support and innovation roles are already identified. | Governance, pilot testing, adoption and handover can be co-developed with AES from day one. |
| Users | AES-stated range of approximately 50-150 staff and around 100 planned AI users. | Pricing, pilots and adoption metrics are sized around the stated user base, with validation during mobilisation. |
Knowledge Landscape
| Area | Current state | Project implications |
|---|---|---|
| Fragmentation | OneDrive, SharePoint, Teams, Outlook, Procore, paper files and individual know-how. No single source of truth; findability is poor. Baselines are measured formally in Phase 0. | The first priority is permission-safe findability and version-aware retrieval before agents rely on the corpus. |
| Volume and mix | TBs+ of drawings / CAD, BIM models, specs, contracts, proposals, reports, policies, schedules, O&M manuals, scanned PDFs, minutes, letters and studies. | Different corpus types need different ingestion patterns: OCR for scanned PDFs, metadata for CAD / BIM, and citation-first retrieval for documents. |
| Governance corpus | HR policies, Code of Conduct, org chart, RACI, SOPs, ISO QMS, project database and Delegation of Authority matrix exist but are partial. | The raw material for the governance layer is present, but needs digitisation, ownership and completion before agent-mediated actions scale. |
| Baseline pain | Proposal drafting is slow; project documents can take days to locate; decisions bottleneck on a few individuals; documents are trapped with people. | Success metrics focus on proposal drafting time, document findability, cited answers and reduced coordination friction. |
1.3 Technology Baseline
M365 is the natural data plane and user surface, while Procore, Jisr, QMS, CAD / BIM and licensed references need specific integration patterns.
| Domain | Current state | Project implications |
|---|---|---|
| Microsoft 365 | Single global tenant, mixed plans, Copilot licensed widely, Teams, SharePoint, Exchange, OneDrive and Office active. | The natural data plane and user surface; Copilot licence value is activated rather than replaced. |
| Identity | Entra ID, MFA, one standard across entities, and in-house administration. | Ready for governed agent identities, conditional access and lifecycle management. |
| Engineering tools | AutoCAD, Revit and Navisworks are desktop, file-based tools with no direct API integration assumed. | CAD / BIM content is made searchable through a file and metadata pipeline, not overstated as live design automation. |
| Project systems | Procore is in use. | Strong API surface: service accounts, webhooks and bulk export create a solid integration path. |
| HR / Payroll | Jisr is the KSA SaaS HR platform. | Customer-enabled Open APIs, including a KSA-local endpoint, can support an HR Assistant once modules and residency scope are validated. |
| QMS | Proprietary and ISO-certified. | QMS remains the system of record; integration path is confirmed in discovery through read-only views, exports or approved access. |
| Reference data | Compass International cost data, SBC, GACA regulations, ICAO and IATA references. | GACA and SBC can become refreshable public corpora; ICAO, IATA and Compass require licensed handling and legal-register decisions. |
| AI today | No production agents; individual ChatGPT, Claude, Gemini subscriptions and custom GPT experimentation. | AES's current experimentation demonstrates clear demand for AI; the proposal converts that momentum into governed enterprise capability with controls, audit and ownership. |
Baseline reading: AES's existing Microsoft and identity environment provides a strong foundation for implementation; the proposal adds the governance, data readiness, retrieval and agent-control layers required to convert that foundation into a controlled enterprise AI capability.
1.4 AES Requirements Summary
This table keeps AES's stated requirements visible and shows exactly where each one is answered in the proposal.
| Requirement area | What AES asked for / raised | Where the proposal responds |
|---|---|---|
| Agent identities | Agents with standing identities, including own email and Teams presence where viable. | Agent Roadmap explains the named agent pilots; Governance explains scoped permissions, owners, lifecycle controls and fallback patterns. |
| DOA enforcement | Flag exceptions, do not block ordinary work. Client submissions and executive decisions stay human-only. | DOA Compliance & Human Decision Boundaries defines what agents can prepare and what AES people must approve. |
| KSA data handling | KSA data must stay in-country where client, government or contractual requirements demand it; routine cross-border flows must be legalized. | Local Regulatory & Copyright Posture covers country-specific obligations; Data Readiness & Governance covers classification, labels and routing. |
| Nine required controls | RBAC (role-based access control), human approvals, DOA enforcement, AI disclosure, retention / deletion, encryption, DLP (data loss prevention) / labels, full audit trail, and retrieval-grounding on AES data. | Meaningful Control Framework maps each required control to the AI Brain mechanism and the control effect for AES. |
| Top concerns | Hallucination, loss of control / auditability, data leakage, immature policies, regulatory exposure, cross-entity politics, staff trust and copyright. | Governance & Risks maps each concern to named mitigations and explains the underlying controls. |
| Commercial route | Urgent timeline, full build ambition, one-off build plus support, formal RFP and MD approval. | Investment Model shows the three options; Implementation Roadmap shows how delivery remains gated after AES chooses the preferred route. |