Smart Digital Procurement Platform for Kazakhstan

DPA 2502

Ministry of Education and Science of the Republic of Kazakhstan

Smart Digital<br>Procurement Platform<br>in Kazakhstan

Transforming Public Procurement Through AI &amp; Digital Innovation

Course:

Digital Public Administration (DPA 2502)

Prepared by:

Maksim Moshko, Sabina, Altynay A.

Instructor:

[Instructor Name]

Astana, 2026

02

Problem Statement

BACKGROUND

The Problem with Current Procurement

Current Platform: goszakup.gov.kz

Relies heavily on manual verification — causing systematic latency

Vulnerable to restrictive "tender engineering" practices

Post-facto (reactive) auditing only — risks go undetected in real time

Project Objectives

Eliminate subjective human gatekeeping through algorithmic validation

Implement real-time ML screening for fraud and collusion detection

Build mathematical and probabilistic models for system stability & scalability

03

Analytical Foundation

GLOBAL CONTEXT

From E-Government 2.0<br>to Cognitive Governance 3.0

Guided by the 'Once-Only' principle — citizens and businesses submit data once; the state handles the rest.

Global Benchmarks

🇪🇪 Estonia (X-Road)

Decentralized data exchange layer

Absolute data integrity via blockchain

Automated inter-agency trust

🇸🇬 Singapore (GeBIZ)

Centralized Singpass identity infrastructure

Predictive market analysis

Automated risk profiling

🇰🇿 Kazakhstan Context:

High UN EGDI index — but back-end data silos prevent full automation. This platform bridges that gap.

04

Solution Design

PLATFORM OVERVIEW

Smart Digital Procurement Platform

An AI-driven, blockchain-anchored, citizen-centric procurement ecosystem

🏛 Procuring Entities

(Ministry & akimat staff)

Automated tender templates

Market pricing recommendations

🏢 Private Suppliers

(Micro, SME & large enterprises)

Frictionless registration

Instant qualification verification

🔍 Oversight Bodies

(Anti-Corruption Agency & auditors)

Immutable audit logs

Live analytics dashboard

Core Functions

AI Document Checking

Semantic Anti-Collusion Screening

Decentralized Ledger Logs

05

User Flow

PROCESS ARCHITECTURE

End-to-End Procurement Flow

Stage 1 — Initiation

Automated semantic scan of technical specs — blocks restrictive criteria before publication

Semantic AI Scanner

Stage 2 — Submission

OCR/NLP processing of supplier credentials against state tax & legal databases in real-time

OCR & Data Validation APIs

Stage 3 — Evaluation

Algorithmic scoring + network fraud & IP analysis — objective, bias-free ranking

Predictive Scoring Models

Stage 4 — Award

Immutable smart-contract execution — full transaction transparency on distributed ledger

Blockchain Distributed Ledger

All steps generate immutable audit logs visible to oversight bodies in real time.

06

KPI FRAMEWORK

PERFORMANCE METRICS

Key Digital Transformation Indicators

📊 Usage Metrics

Platform Adoption Rate

Target: 100% within 5 years

Active Supplier Density

Tracks market competitiveness

Automated Processing Ratio

Measures system autonomy

⚡ Efficiency Metrics

Mean Cycle Time to Award

Reduction: 24 days &rarr; 5&ndash;12 days

Administrative Transaction Cost

Target: &darr;30%

Price Variance Deviation

Monitors market distortion

♿ Accessibility Metrics

SME Participation Index

Ensures inclusive procurement

Public Data Request Resolution Time

Transparency measure

Systemic Risk Flag Resolution Rate

Audit response speed

07

Forecasting

STRATEGIC HORIZON

25-Year Platform Maturity Roadmap

3-Year Horizon: Stabilization

75% adoption in republic-level ministries

Cycle time drops to 12 days

$45M USD in fiscal savings

5-Year Horizon: Universal Integration

100% adoption across all regional akimats

90% automated processing ratio

Collusion detection improved by 70%

10-Year Horizon: GovTech Maturity

Shift to predictive economic instrument

Joint procurement recommendations

Cumulative savings: $320M USD

25-Year Horizon: Cognitive Governance

Fully autonomous infrastructure

Self-generating micro-tenders via smart contracts

99.9% operational uptime

08

Optimization

EFFICIENCY GAINS

From Reactive Control to Proactive Algorithmic Governance

40%

Reduction in Administrative Latency

via real-time API integrations replacing manual data lookups

30%

Reduction in Transaction Costs

average cost per tender dramatically lowered through automation

Proactive Audit System

Inline algorithmic shield replaces post-facto cameral control

Impact Comparison

Metric

Before (Manual)

After (AI Platform)

Tender Cycle Time

24 days

5–12 days

Audit Trigger

Post-facto

Real-time

Error Detection

Reactive

Predictive

09

Resources

IMPLEMENTATION RESOURCES

Resource Allocation Matrix

Senior ML Engineers, Blockchain Architects, GovTech Security Specialists

Algorithmic screening, immutable logging, zero-knowledge security

Core Development Budget, Continuous Maintenance Fund, Public Training Budget

Backend software production, API integration, regional onboarding

Distributed Cloud Clusters, Cross-Agency API Gateway, TLS 1.3 / AES-256 Encryption

High-availability hosting, data interoperability, data protection

10

RISK ANALYSIS

PROBABILISTIC RISK MODELING

Bayesian Risk Assessment Framework

R₁

Systemic Technical Failure

15%

HIGH

R₂

Cyber Attack

8%

CRITICAL

R₃

Low Institutional Adoption

25%

MEDIUM

R₄

Algorithmic False Positives

10%

MODERATE

Bayesian Inference Method

Allows public managers to dynamically calculate the true probability of a threat given an automated alert.

P(Alert) = Σ[P(Rᵢ) × P(Alert|Rᵢ)] + [P(No Risk) × P(Alert|No Risk)] = 0.4205

11

RESULTS

POSTERIOR PROBABILITY RESULTS

Prior vs. Posterior Risk Probabilities

Technical Failure (R₁)

↑ +18.89 pp — Highest posterior risk

15%

33.89%

Cyber Attack (R₂)

↑ +9.12 pp — Critical infrastructure threat

8%

17.12%

Low Adoption (R₃)

↓ −1.22 pp — Relatively stable

25%

23.78%

Algorithmic Bias (R₄)

↑ +10.21 pp — Needs monitoring

10%

20.21%

An automated alert is statistically most likely to indicate infrastructure failure or cyber breach

(R₁ + R₂ = 51.01%)

Resource priority must shift toward technical redundancy and cybersecurity.

Demand Model

12

BASS DIFFUSION MODEL

Digital Service Demand Forecasting

Total addressable pool: 35,000 public entities + 180,000 private suppliers

📈 Optimistic Scenario: +18% Annual Growth

Mandated integration of semi-state holdings (e.g., Samruk-Kazyna)

45,000 concurrent peak requests/hour

Requires automated cloud sharding & Kubernetes orchestration

📊 Baseline Scenario: +8% Annual Growth

Stable growth, seasonal budget cycles handled smoothly

20,000 concurrent peak requests/hour

Standard cloud scaling sufficient

📉 Pessimistic Scenario: +2% Annual Growth

Institutional stagnation in rural areas

Requires regional IT support expansion

Simplified interfaces needed for low-digitization regions

13

Conclusion

SUMMARY

Conclusion & Feasibility

Value Proposition

Maximizes public value, ensures absolute transparency through immutability, and generates a 12–15% reduction in state procurement expenditure.

Feasibility in Kazakhstan

Highly viable due to existing eGov infrastructure and digital signature (EDS) ecosystem — foundational layer already in place.

Success Prerequisites

Requires strong political will, deep back-end database integration, legislative updates, and proactive civil service training.

Smart Digital Procurement is not a future vision — it is a present-day imperative for Kazakhstan's GovTech maturity.

Digital Public Administration | DPA 2502 | Astana, 2026