Kenya’s developers, engineers, students, and policy minds packed into Safariccom’s Decode 4.0 conference this week, gathering under a theme that doubled as a declaration: Made of Kenya.
Now in its fourth edition, the annual technology festival has grown from an internal engineering showcase into something closer to a national conversation about where Kenya’s digital economy is headed and who gets to build it.
A 25-Year Platform, Pointed at 2030
James Maitai the Group Chief Technology and Information Officer (CTIO) at Safaricom PLC opened the morning by connecting Decode to the company’s 25th anniversary, framing the event not as a celebration of the past but as a commitment to the future.
“What started as a platform to showcase engineering and build social talent has grown into something more transformative,” he told the audience. “Not just in Kenya but beyond.”
The company’s stated ambition is to become Africa’s leading purpose-driven technology company by 2030. Maitai sketched out the three pillars carrying that ambition: intelligent infrastructure, fintech, and the creative economy. On fintech, he pointed to last year’s upgrade of the M-PESA platform to what the company calls Fintech 2.0, describing it as “one of the most complex and significant updates to our platform in the last 10 years,” one that introduced autonomous fraud detection, credit scoring, and self-healing capabilities.
On creativity and youth, he cited Kenya’s median age of 21 and argued that 5G connectivity, now covering 30 percent of the population, was already enabling a new generation of developers, gamers, and content creators to compete globally.
The Kitchen Metaphor That Stopped the Room
Principal Secretary for Broadcasting and Telecommunications Mr. Stephen Motari Isaboke, delivered a memorable address, and he arrived with an image that cut through every slide deck and product demo.
At Mobile World Congress in Barcelona earlier this year, Isaboki met a partner who had promised 30 percent efficiency gains from an AI deployment within 12 months. When he followed up a year later, the response was simple: “We were too early. The meal was in the kitchen. It was not ready to be served.”
That image, of ingredients assembled but the kitchen unprepared, became his framework for Kenya’s AI moment.
“The ingredients are extraordinary,” Isaboki said. “Huge volumes of data flowing through our networks. Breakthrough models that can see patterns, understand language, and generate content. A young, ambitious population that is curious and entrepreneurial. But the kitchen still matters.”
Preparing that kitchen, he argued, means training people to work alongside AI rather than fear it, building clean and well-governed data architecture, establishing accountability frameworks, and embedding AI into how organisations actually function rather than running it as an isolated pilot.
He pushed back gently on the framing of AI as a fourth industrial revolution. “It is not the fourth industrial revolution,” he said. “It is the engine that accelerates all the revolutions before it.”
Agents, Not Chatbots
The technical sessions moved quickly from theory to demonstration. Marvin Gessa, who has spent two years building agentic systems that handle real payments and real customer decisions, drew a sharp line between AI tools that answer questions and AI agents that take action.
“The difference between an AI that you can talk to back and forth and an AI that can do an action for you,” he told the audience, “is what I am going to take you through.”
His demo of Zuri Enterprise, Safariccom’s internal AI companion, showed agents managing emails, querying backend systems, generating sales proposals, and preparing meeting briefs without the user switching between applications. A second demo showed a field technician using a multimodal agent for live guidance during a 4G base station installation.
The afternoon session in the Audion breakout room took that concept down to the scale of a small business. Presenters walked delegates through building a WhatsApp-connected AI agent for a fictional electronics company, using a no-code tool called N8N and a Google Sheet as the inventory database. The agent could answer return policy questions, check stock availability, and recommend products, all without a single line of code written by the business owner.
Security Is Not Intelligence
The day’s final technical session delivered an essential corrective. Two Safariccom engineers demonstrated live how an AI-powered ecommerce chatbot could be manipulated into changing a product’s price to zero through a simple prompt. A second demo showed how malicious instructions embedded in a downloaded coding skill file could execute a script on a developer’s machine without any warning or approval prompt.
“Just because something can reason in record time does not mean it is secure,” said Joel, one of the presenters. “As you make these agents, red team them before someone else does.”
The lesson ran directly against the day’s prevailing optimism, and that was the point.
The Number on the Wall
Between sessions, entrepreneur Tony Ndu’ngu offered the day’s most direct provocation to the students in the room. Write a number on your bedroom wall, he said. The amount you want AI to earn you by the end of the year. Then ask the tools you already have access to how you close that gap.
“The opportunity is there,” he said. “The only thing that is left is for you to decide.”
That tension, between an extraordinary set of tools and the discipline required to use them well, ran beneath everything said at Decode 4.0. The kitchen is stocked. The chefs are talented. Whether the meal reaches the table depends on what happens next.
