The road to Agentic AI in the Public Sector

Transcript

1. The road to Agentic AI in the Public Sector Mohammed Fazalullah Presales Manager CNTXT THE FUTURE, IN CONTEXT | 2026 cntxt.com
2. The era of passive AI is over. Agentic transformation has arrived. Chat What is the latest performance data for our summer 2026 ad campaign? Think and reason Compare the click-through rates of our email campaigns from Q1 and Q2 Execute Identify our highest-value segment and draft personalized campaigns to maximize their lifetime value
3. Using different tools to execute your agentic vision will only introduce context walls that stall your momentum PERSONAL WORK CONTEXT ENTERPRISE CONTEXT This limits the potential for both human and AI agents from doing their best work
4. “In- country” means different things for different states of data Data at rest Data in use (ML processing) Data in transit Stored data: datasets, model artifacts, buckets, logs The compute that runs inference / tuning Data moving across the network Pin - able to me - central2 (Dammam) Stays in the customer - selected location regardless of which endpoint serves the request. Depends on the model Managed frontier models process outside the region; a self - hosted model on in - region GPUs processes in Dammam. TLS - encrypted; not a residency guarantee Protected by encryption in transit; routing is not covered by the residency commitment.
5. Gemini Enterprise Agent Platform Google Cloud's comprehensive and open platform for enterprises to build, scale, govern, and optimize enterprise ready agents. Transform applications and workflows into powerful agentic systems at a global scale Transition agents from experimental into fully managed, production - grade systems Optimize agent quality for business performance through full data - driven evaluations 7
6. Agent Platform is at the heart of an end end system for the agentic era Apps and solutions Gemini Enterprise app Gemini Enterprise Agent Platform 1P, 3P, Open Models Agentic capabilities across: Custom apps built by you or partners ADK, Agent Runtime, and Model Training Build Agent and Tool Registry Scale - to- Gemini Enterprise for Customer Experience Agent Gateway Govern Agent Identity, Agent Security Agent Observability, Agent Evaluation Optimize Powered by the integrated stack for agentic AI AI Hypercomputer Agentic data cloud Agentic defense 8
7. Apps & Solutions Gemini Enterprise Agent Platform + Apps & Solutions App & Solutions Built on Gemini Enterprise Agent Platform, front door for AI with centralized oversight and scaled governance Gemini Enterprise app Custom apps built by you or partners Your employees can expand productivity with an embedded agentic taskforce. Your developers can build production ready AI agents and scale across the org. Gemini Enterprise for CX - Your customers can discover, purchase, and get help across every touchpoint, without starting over. Gemini Enterprise Agent Platform The foundation with world - class models, tools, and infrastructure to build, scale, govern, and optimize agents 1P, 3P, Open Models ADK, Agent Runtime Agent Gateway Agent & Tool Registry Agent Identity, Agent Security Agent Observability, Agent Evaluation Agentic capabilities across Build Create sophisticated, long Scale - running agents Connect systems, take real - world action Govern Optimize Ensure privacy, compliance, protection Observe & improve agent performance Powered by the integrated stack for agentic AI AI Hypercomputer Scale inference and training Agentic Data Cloud Fuel a system of action Agentic Defense Secure the AI lifecycle
8. Gemini Enterprise Agent Platform Gemini Enterprise Agent Platform + Apps & Solutions App & Solutions Built on Gemini Enterprise Agent Platform, front door for AI with centralized oversight and scaled governance Gemini Enterprise app Custom apps built by you or partners Your employees can expand productivity with an embedded agentic taskforce. Your developers can build production ready AI agents and scale across the org. Gemini Enterprise for CX - Your customers can discover, purchase, and get help across every touchpoint, without starting over. Gemini Enterprise Agent Platform The foundation with world - class models, tools, and infrastructure to build, scale, govern, and optimize agents 1P, 3P, Open Models ADK, Agent Runtime Agent Gateway Agent & Tool Registry Agent Identity, Agent Security Agent Observability, Agent Evaluation Agentic capabilities across Build Create sophisticated, long Scale - running agents Connect systems, take real - world action Govern Optimize Ensure privacy, compliance, protection Observe & improve agent performance Powered by the integrated stack for agentic AI AI Hypercomputer Scale inference and training Agentic Data Cloud Fuel a system of action Agentic Defense Secure the AI lifecycle
9. Integrated stack for agentic AI Gemini Enterprise Agent Platform + Apps & Solutions App & Solutions Built on Gemini Enterprise Agent Platform, front door for AI with centralized oversight and scaled governance Gemini Enterprise app Custom apps built by you or partners Your employees can expand productivity with an embedded agentic taskforce. Your developers can build production ready AI agents and scale across the org. Gemini Enterprise for CX - Your customers can discover, purchase, and get help across every touchpoint, without starting over. Gemini Enterprise Agent Platform The foundation with world - class models, tools, and infrastructure to build, scale, govern, and optimize agents 1P, 3P, Open Models ADK, Agent Runtime Agent Gateway Agent & Tool Registry Agent Identity, Agent Security Agent Observability, Agent Evaluation Agentic capabilities across Build Create sophisticated, long Scale - running agents Connect systems, take real - world action Govern Optimize Ensure privacy, compliance, protection Observe & improve agent performance Powered by the integrated stack for agentic AI AI Hypercomputer Scale inference and training Agentic Data Cloud Fuel a system of action Agentic Defense Secure the AI lifecycle
10. Build Agent Development Kit 3P agent frameworks Tools, data, and other agents Gemini API and Model Garden Gemini models Model training Agent Garden Agent Studio 3P and open models Model inference A2A Grounding RAG MCP Search APIs and connectors A2UI AP2 and UCP Cloud Marketplace Scale Agent Runtime Agent Sessions Agent Sessions Agent Sandbox Agent Memory Bank Govern Agent Gateway Agent Identity Agent Registry Agent Anomaly Detection Model Armor Agent Policy Agent Security Agent Compliance Agent Observability Agent Optimizer Optimize Agent Evaluation Agent Sessions Agent Simulation
11. Three layers decide where data lives and who touches it 1 - Your application architecture Owned by Customer Which service & region your app calls, what data it sends, how it authenticates end where” questions are answered here, not by a platform toggle. - users, what it logs. Most “who - from - 2 - Google Cloud platform controls Region selection, CMEK / Cloud EKM, VPC Service Controls, IAM, Org Policy, audit logging. Powerful guardrails they enforce the boundaries your architecture defines. - but Configured by you on GCP 3 - Google's operational commitments Provided by Google Default encryption, the no - training contractual commitment, Access Transparency, regional infrastructure, third attestations. The baseline you build on top of. - party
12. What the Agent Platform actually runs in These services execute in Dammam; pair them with in - region GPUs to host your own models. Custom model training Online & batch inference Train your own models in Serve predictions from Dammam - region - region today In- region GPUs Available in Dammam to host your own models: Model Registry Agent Platform Pipelines Versioned model artifacts Orchestrated ML workflows NVIDIA L4 Cost - efficient inference / serving Gemini Enterprise Agent Platform TensorBoard Agent Platform Vizier Training visualisation Black - box optimisation RTX PRO 6000 Blackwell High - VRAM for larger open models + Gemini Enterprise Agent Platform Explainable AI
13. What me - central2 (Dammam) gives you today Data stored at rest in Dammam Yes - selectable & enforceable Agent Platform API / regional endpoint in Dammam Yes - me - central2 is live Custom training, inference, Registry, Pipelines in - region Yes - runs in Dammam In- region inference via self - hosted open model (L4 / RTX 6000) Yes - you host it Managed frontier models (Gemini, Claude) Inference requests to US / EU / global endpoints Block the global endpoint via Org Policy Yes - gcp.restrictEndpointUsage Sovereign boundary + locally - held keys Via Assured Workloads / Sovereign Controls
14. GCP enforces boundaries. Your architecture defines them. Google Cloud handles Your application must design Encrypt data at rest & in transit by default Choose the model & whether inference is in - region Enforce the region you select for storage Map end - users to identities & enforce app - level authZ Block cross - perimeter data movement (VPC - SC) Decide what data is sent in a prompt (minimise / mask) Evaluate IAM permissions on every API call Design tenant isolation between customers / departments Record audit logs of admin & data access Set the IAM grants, conditions & service accounts Not train foundation models on your data Handle residency for data before it reaches the platform
15. The enterprise agent architecture: A blueprint for intelligence Agent Interface (UI/UX for Conversational / RESTful, Websocket for Workflow) Network Security / DLP Prompt Mgmt. Prompt Defence) Session, Memory & Example Store Authn / Authz Security Layer (User IAM) Agent IAM RAG (Agent Service Account) (Vector Stores) Input / Interaction Layer (Reasoning Logic / Responsible AI / Layer Safety Filter Safety Interaction Logging & Trace Observability & Operability Inference Monitoring / Autoscaling Tools, Data and Interoperability LLM Drift Monitoring / Agent Evaluation Grounding / Citation Checks Function Call Endpoints User Feedback for RLHF Tool Use (Code Interpreter / MCP) Continuous Evaluation and Quality Improvement Prompt Optimization Model Fine Tuning Agentic Workflow (Agent2Agent)
16. Plan Build Integrate your agent with enterprise data, workflows, and tools Custom APIs Pre - built connectors Automation workflows Databases Apigee API hub Google Cloud integration connectors Google Cloud application integration https://github.com/goo gleapis/genai - toolbox
17. Plan Build Build and publish high Google Cloud Developer Workstation Frameworks - code agents in Google Cloud Tools deplo y Runtimes register Agent Engine Gemini Enterprise Custom API ADK Employees LangGraph MCP tools App Integration Tools Cloud Run LangChain Apigee GKE Agentic Applications Built-in tools CrewAI Models Observability AG2 Cloud Trace Other Python frameworks Cloud Logging Agent2Agent Cloud Monitoring IDE, Assistant Gemini Code Assist, Gemini CLI Antigravity Write code & Test Locally Evaluation Gemini Enterprise Agent Platform Evaluation Web Android 3rd Party Runtimes Agents Deployment Environment Consumption IOS Customers
18. Built on a foundation of enterprise readiness Whether 1st - party, 3rd - party, or open source, Google Cloud gives you the tools, services, and infrastructure to make every deployment enterprise ready Data governance, indemnity, and privacy Security and compliance support Responsible AI • You own and control your data, not Google Gemini Enterprise Agent Builder Your Google Cloud Perimeter Conversational AI Reliability and sustainability Enterprise Search Google Cloud Central Hosting Gemini Enterprise Model Builder Large Foundation Model (Frozen) • Security Controls with VPC Service Controls, CMEK, access transparency • Intellectual property indemnity for training data and generated output • Cost Controls with Dynamic Workload Scheduler and Provisioned Throughput Adapter Layers • Responsible AI tooling, enablement, and support
19. Proprietary + Confidential Data foundations critical for AI implementation Scalability and flexibility The first step is data Multimodal and unified data Data is the foundation of your AI journey Democratization and controls Real - time intelligence Security and controls
20. Proprietary + Confidential Principles for building AI applications Foundational pillars for moving beyond experimentation to achieve tangible value. Build on an AI optimized infrastructure Connect your enterprise data Unify the full AI lifecycle Deploy with governance Build with performance, efficiency, and scale Ground models in high quality, enterprise - ready data Unify MLOps to accelerate and optimize value Deploy applications with enterprise - grade trust and control
21. Proprietary + Confidential Potential use cases in KSA Public Sector Foundations Scale Transform Policy, Legal & Regulatory Frameworks Utilize GenAI to summarize complex legislative drafts, compare new regulations against existing laws, and extract key legal obligations. Deploy cross-entity Knowledge Graphs to run automated Regulatory Impact Assessments (RIA) across intergovernmental datasets. Regulatory Oversight Agent that continuously scans proposed decrees, flags cross-ministry legal conflicts, and drafts compliant harmonized policy frameworks. Citizen & Resident Services (Public Experience / CX) Multilingual (Arabic/English) AI assistants providing instant, grounded answers to public inquiries and service eligibility rules. Multi-modal processing of citizen applications—automatically parsing, validating uploaded documents, and routing edge cases. Omnichannel Citizen Concierge Agent that autonomously executes complex end-toend public transactions (e.g., permits, registrations, licensing) across entities. Government Operations & Procurement Summarize RFPs/vendor bids, extract risk terms from proposals, and analyze spending logs to identify budget optimization areas. Automated vendor compliance validation, real-time contract risk analysis, and streamlined cross-departmental workflow routing. Public Procurement Agent that autonomously evaluates tender submissions against national guidelines, conducts audit checks, and triggers procurement approvals. Infrastructure & Smart City Management Catalog and summarize municipal service requests, infrastructure incident logs, and public asset maintenance reports. Deploy Urban Digital Twins connected to IoT networks to monitor civic infrastructure performance, traffic flow, and utility distribution. Autonomous Urban Management Agent that predicts infrastructure degradation, dispatches municipal maintenance teams, and dynamically re-routes traffic/grid loads. National Strategy, ESG & Vision 2030 Alignment Synthesize multi-entity performance reports and summarize national KPI milestones into executive leadership briefs. Real-time monitoring of national sustainability/energy metrics and automated ESG/Vision 2030 compliance reporting across SOEs. Vision 2030 Strategic Agent that autonomously tracks cross-sector KPI progress, simulates economic impact scenarios, and recommends policy interventions.
22. Proprietary + Confidential Your Data & AI Transformation Journey 0. Strategize 1. Foundations 2. Scale 3. Transform What you’re doing Define a clear vision for data. Identify key sponsors and a high - impact "Day One" pilot use case (e.g., risk, compliance). Begin breaking down silos . Unify 2- 3 critical data sources into a secure data cloud for your pilot project. Drive adoption by proving the pilot's value. Ingest new, complex data sources (e.g., unstructured, real - time). Reimagine core processes (e.g., underwriting, client service) with agentic AI. Scale data - driven insights across the workforce. How we support you CNTXT to lead a strategic "Art of the Possible " workshop. Jointly develop a data first roadmap and governance plan. CNTXT or Google Partner to support data unification and build the secure "Data Foundation " and "Proof of Value " for the pilot. CNTXT or Google Partner to support scaling the data platform, rolling out new capabilities, and ensuring robust security & controls. CNTXT or Google Partner to provide agentic AI thought leadership and co - develop transformative use cases. Data Readiness Assessment Strategic workshop Data Foundation Lab Proof of Value (PoV) Monthly cadence Architecture reviews Deep dive workshops AI Professional Services How we can customize Reach out to your CNTXT account team for further discussion
23. Proprietary + Confidential CNTXT is your partner in the Kingdom from vision to value Strategic onboarding Our industry - expert teams help you define a high - impact AI roadmap with collaborative engagements CNTXT Delivery Accelerated success 80+ partners help you implement and scale AI across your business Tailoring advanced technologies to your business maximize your ROI
24. Thank you Visit us at the Google Booth and Hall-2 CNTXT Booth