Project Preparation

Strategic roadmap · 2025–2028

Public Infrastructure Investment Coach

A phased strategy for embedding AI across all 8 elements of the public investment project cycle — augmenting human judgement, strengthening appraisal quality, and building institutional capacity in government budget agencies.

  • Version 1.0 · March 2025
  • IMF PIMA aligned
  • HM Treasury Green Book
  • Human-in-the-loop first

Strategic vision

“Every public infrastructure investment decision backed by the best available evidence, analysis and guidance — delivered at the speed government needs.”

The strategic opportunity

40%

Efficiency gap

Public investment efficiency gap in low-income countries vs. advanced economies (IMF estimate)

⅔

Closable via PIM reform

Share of the efficiency gap that better infrastructure governance can close

8

Project cycle elements

Guidance, Appraisal, Review, Selection, Implementation, Adjustment, Operation, Evaluation

2×

Growth dividend

The most efficient public investors get twice the economic growth impact per dollar invested

Three-phase implementation roadmap

Strategic Focus

Phase I

  • Validate AI coach core capabilities with 3–5 government pilot agencies
  • Focus on Project Appraisal & Guidance elements
  • Build trust with human-in-the-loop design

Phase II

  • Extend to Selection, Review, and Implementation elements
  • API integrations with national PIMS platforms
  • Multi-sector deployment across 15+ agencies

Phase III

  • Full 8-element PIM cycle coverage
  • Regional/multilateral institution partnerships
  • Sovereign AI capacity building offering

AI Capabilities Delivered

Phase I

  • Feasibility study quality checker
  • CBA/CEA methodology guide & calculator
  • International standards knowledge base (PIMA, Green Book, PCR)
  • Project complexity classifier

Phase II

  • Agentic appraisal workflow automation
  • Risk & sensitivity analysis engine
  • Climate-informed investment screening
  • Portfolio ranking & selection support

Phase III

  • Ex-post evaluation & lessons capture
  • Predictive implementation risk model
  • Cross-country benchmark analytics
  • Autonomous shadow pricing updates

Governance & Oversight

Phase I

  • Ethics & AI policy framework published
  • Human review required on all AI outputs
  • Audit trail & explainability logs

Phase II

  • Government advisory board established
  • Independent quality assurance review
  • Tiered AI autonomy levels by decision risk

Phase III

  • ISO/IEC AI governance certification
  • Published transparency reports (annual)
  • Regulatory sandbox partnerships

Capacity Building

Phase I

  • Field guide for budget officials (v1)
  • AI literacy workshops (MoF, line ministries)
  • Champion network: 20 trained officials

Phase II

  • Certified PIM.ai Practitioner programme
  • University curriculum partnerships
  • 200+ officials trained across 8 countries

Phase III

  • Open-source knowledge commons
  • South-South learning network
  • 1,000+ practitioners in PIM.ai community

Data Infrastructure

Phase I

  • International standards corpus (IMF, WB, OECD)
  • National parameters database (shadow prices, discount rates)
  • Secure government data protocols

Phase II

  • Live project pipeline data integrations
  • Climate & geospatial data layer
  • Cross-border data sharing frameworks

Phase III

  • Federated learning across government agencies
  • Real-time project monitoring feeds
  • Privacy-enhancing technology (PETs) deployment

AI applications across the 8 PIM project cycle elements

01

Guidance

AI knowledge base synthesising IMF PIMA, Green Book, World Bank PCR and national policies into context-aware guidance for project teams.

Phase I

02

Project Appraisal

Automated CBA/CEA scaffolding, shadow price lookups, feasibility study completeness checks, and climate-adjusted economic analysis.

Phase I

03

Independent Review

AI-assisted quality assurance checklists, optimism bias detection, methodology compliance scoring against international standards.

Phase II

04

Selection

Portfolio ranking using weighted MCA + CBA, fiscal space modelling, strategic alignment scoring, and pipeline prioritisation dashboards.

Phase II

05

Implementation

Progress monitoring against baselines, early warning alerts for cost overruns and schedule slippage, disbursement pattern analysis.

Phase II

06

Adjustment

Scenario modelling for project redesign options, re-appraisal automation when scope changes exceed thresholds, risk-adjusted cost projections.

Phase III

07

Operation

Asset lifecycle cost optimisation, preventive maintenance scheduling, performance benchmarking against design-stage projections.

Phase III

08

Evaluation

Counterfactual impact evaluation, predicted vs. actual benefit tracking, institutional lessons extraction and feedback into future project guidance.

Phase III

Success metrics by 2028

30%
Reduction in feasibility study preparation time
50+
Government agencies using PIM.ai actively
1,000+
Budget officials trained & certified
8/8
Project cycle elements with AI capability
95%
Appraisals meeting IMF PIMA quality standards

Guiding design principles

01

Human Authority, AI Augmentation

Every AI output is an input to human decision-making, never a replacement. Budget officials retain full authority. AI earns trust by being transparent, auditable, and consistently explainable — not by claiming certainty.

02

Multi-Tier Knowledge Architecture

Systematically integrates international good practice (IMF, World Bank, OECD), national policy frameworks, and project-specific data. No single tier dominates — the coach bridges all three coherently.

03

Progressive Complexity Scaling

Simple tools for simple projects; sophisticated analytics for complex mega-projects. The system adapts its depth of analysis to project risk, scale, and data availability — never over-engineering small decisions.

04

Climate-Integrated by Default

Climate risk and transition considerations are embedded throughout — not added as an afterthought. Every appraisal includes climate scenario analysis. Shadow carbon pricing is applied automatically where appropriate.

05

Sovereign Capacity, Not Dependence

The goal is to build lasting government capability — training officials, transferring knowledge, and strengthening institutions — not to create perpetual reliance on an external AI platform.

06

Transparent & Auditable by Design

Every recommendation includes its reasoning chain. Every data source is cited. Every assumption is visible and challengeable. PIM.ai is built to survive parliamentary scrutiny and supreme audit institution review.

Governance & risk management

AI Risk Classification (inspired by EU AI Act)

  • Low risk: Guidance queries, information retrieval, standards lookup — AI responds autonomously

  • Medium risk: Appraisal scaffolding, template population, indicator calculation — AI drafts, official reviews

  • High risk: Project ranking, portfolio selection recommendations — AI provides analysis, committee decides

  • Critical: Final investment approval, budget appropriation — humans only, AI provides supporting evidence dossier

Institutional Safeguards

  • Independent AI Ethics & Quality Board with external members (academia, civil society, MDB representatives)

  • Annual transparency report published covering AI decision volumes, override rates, and quality outcomes

  • Mandatory human-in-the-loop checkpoints aligned to PIMA gate review stages

  • Full audit trail for every AI-assisted analysis stored for 10 years

  • Bias monitoring across sectors, project sizes, and country income groups

  • Right-to-explanation: any official can request plain-language rationale for any AI recommendation

Inspired by
  • Singapore NAIS 2.0 Whole-of-government AI strategy with phased enablers
  • State of Georgia AI Roadmap Three-horizons model & innovation lab approach
  • EY Government AI Framework 5-step scaling from pilots to institutional embedding
  • IMF PIMA Framework 15-institution assessment & reform sequencing
  • EU AI Act Risk Tiers Impact-based governance levels

PIM.ai · Public Infrastructure Investment Preparation Coach · Strategic Roadmap v1.0 · March 2025 · For government budget officials