Most tech events follow a dull blueprint: lukewarm coffee, long presentations, and zero hands-on execution. Cloud Mastery 2026 Second Edition broke that pattern with agentic AI. Pawa IT, a Premier Google Cloud Partner in Kenya, hosted it at The Social House, Nairobi. We set one strict rule: no passive watching, only building.
On July 16, 2026, developers, senior architects, and IT leads arrived with empty sandboxes. By the end of the day, they had built a production-grade agentic AI system on Google Cloud. They used Vertex AI Agent Builder, Cloud Run, and BigQuery ML.
Here is how we built a multi-agent AI system live in one day.
Why Cloud Mastery 2026 Was a Hands-On Agentic AI Workshop
A few weeks before Cloud Mastery 2026, our team held a routine planning session for the Google Cloud workshop in Nairobi. The draft agenda was safe, with standard cloud topics and a standard format. However, our CTO, Oscar Limoke, rejected it. “We are not doing another basic cloud seminar. If we bring Cloud Mastery back, we have to give people something they can actually deploy.”
That one line became the brief. Attendees would start at 9 AM with nothing. They would leave by 4 PM with a live multi-agent AI system running in their own cloud environment.
To get there, our engineers and cloud architects spent late nights testing container builds and configuring Workload Identity Federation. They also grounded every AI agent in real Kenyan market data. As a result, nothing broke when the room logged in.
Meet SokoAI: A Multi-Agent AI System Built for Kenya
The day opened with a keynote from Kenneth Olang, Platform Customer Engineer at Google. He made the strategic case for agentic AI. He explained why autonomous reasoning systems are moving software past simple chatbots.
Next, Oscar Limoke introduced the centerpiece of the day: SokoAI. It is an AI agent built for local market realities, not a generic demo app. Oscar ran a live demo of the multi-agent AI system. SokoAI handled four real jobs at once:
- Soko Shop for retail and grocery search.
- Soko Parts for compatibility queries, such as “I have a Mazda Demio 2018, what battery do I need?”
- Soko Wealth for context-aware personal finance advice.
- Soko Support for FAQs, store locations, and logistics.
How We Built Agentic AI on Google Cloud, Layer by Layer
The hands-on work started immediately. Each layer built on the one before it.
1. Cloud Run and Workload Identity Federation: A Secure Foundation
Evans Mwongera, Cloud Architect at Pawa IT, led the containerization of Next.js and NestJS microservices on Google Cloud Run. Then attendees configured Workload Identity Federation. This removed stored service account keys and solved the classic “keys under the mat” security risk. By lunch, most engineers had a secure, live app on Cloud Run, backed by Cloud SQL.
2. Vertex AI Agent Builder: Wiring the Agent Brain
After lunch, Eddie Ngugi, then Head of Engineering, guided attendees through Vertex AI Agent Builder. First, we built specialist sub-agents. Next, we grounded them in BigQuery product data stores. Finally, we wired streaming AI responses into the Next.js frontend through API routes.
3. BigQuery ML K-means: Adding Data Intelligence
John Higi, Senior Data Engineer, closed the loop with BigQuery ML. Using pure SQL, attendees ran K-means clustering to segment users and predict conversion rates. Then they fed the model outputs straight back into SokoAI. As a result, its recommendations got smarter in real time.
Debugging Live Deployments at Cloud Mastery 2026
Live deployments bring quirks, port errors, environment mismatches, and container build fixes. Cloud Mastery 2026 was no exception. Therefore, our engineering team provided direct, 1-on-1 debugging at the tables.
To keep the momentum going, we also gave every attendee a full week of free Google Cloud tokens. We provisioned them directly to their project sandboxes, so everyone could keep building at home.
Agentic AI on Google Cloud: Frequently Asked Questions (FAQ)
What is agentic AI?
Agentic AI refers to autonomous AI systems that can reason, plan, and use tools or APIs. They complete complex multi-step tasks and go far beyond standard conversational chatbots.
What is Vertex AI Agent Builder?
Vertex AI Agent Builder is Google Cloud’s low-code/no-code environment. Developers use it to create, ground, and deploy enterprise-grade AI agents with natural language playbooks and custom data stores.
Can you build a multi-agent systems on Google Cloud?
Yes. With Vertex AI Agent Builder, Cloud Run, and BigQuery ML, you can build and deploy a full multi-agent AI system. That is exactly what our Cloud Mastery 2026 attendees did with SokoAI.
Is Pawa IT a Google Cloud Partner in Kenya?
Yes. Pawa IT is a Premier Google Cloud Partner in Kenya. We host hands-on technical workshops in Nairobi, like Cloud Mastery. Each one is built around real, deployable projects instead of passive sessions. We also provide enterprise cloud consulting across Africa.
Ready to Deploy Agentic AI in Your Enterprise?
🚀 Explore Our Product: Check out Konversations, our production-ready AI agent platform powering enterprise customer engagement.
📩 Get in Touch: Book a personalized Google Cloud & Agentic AI Consultation with our architecture team today.
