Agentic AI is gaining traction in South Africa and is quickly proving to be indispensable wherever it is deployed.
This is according to Spatialedge, an AI and data science consulting specialist and partner of Digicloud Africa, Google's chosen reseller enablement partner in Africa. Spatialedge, which boasts a client base including leading South African enterprises, says its focus on agentic AI has stepped up considerably this year.
Spatialedge Co-Founder and Executive Director Pierre le Roux says: ”AI models have matured a great deal during the past year, and as soon as a person experiences working with agents for the first time, it feels magical. This year, a lot more organisations have started asking how to use it to help them deliver work better. This year, we’re seeing a lot of interest locally and internationally from clients who want to streamline older processes using agents or LLMs, because it has suddenly become more feasible to do it.”
The company works closely with clients to go beyond the GenAI hype and put the technology to work. Spatialedge is building agentic AI solutions tailored to real business needs, such as a medical agent that streamlines the administration for doctor-patient consults and automatically files claims; a legal assistant that helps compile case files and enables the client to react fast and win settlements; and a meeting agent that extracts insights and nudges teams (such as insurance brokers) to follow sales playbooks.
“What we’ve learned is that success hinges on getting the foundations right: clean, well-structured data, the right cloud infrastructure and a clear path to production. That’s what separates enterprise-grade GenAI from the everyday tools people might use in their browsers or on their mobile phones,” Le Roux says.
“For example, if you were a wealth management firm and wanted to know which clients you should ‘fire’, your model might assess factors like the cost of keeping that client, the amount of interaction and attention the client requires, or the performance of this client's portfolio. The infrastructure required to have all of this data easily accessible is important. For example, if all the data is loaded into BigQuery, we can click one button and our self-service analytics agent is available for people to use.”
Underpinning agentic AI with the right tools
Le Roux says: “We lean heavily on Google Cloud and Google BigQuery – Google Cloud's fully managed and completely serverless enterprise data warehouse, as well as Google’s Agent Development Kit (ADK) – a flexible and modular framework for developing and deploying AI agents, Firebase – Google's Mobile and Web App Development Platform, and Vertex AI to build and deploy ML models faster, with pre-trained APIs within a unified AI platform.”
He adds: “We’re using Gemini Pro 2.5 throughout the whole business, with NotebookLM for aggregating information, and we prefer deploying the agents within the Google Model Garden. We prefer Google solutions because of their high levels of security and the fact that Google has spent a lot of time thinking about the user experience and the developer experience. For instance, Firebase is a phenomenal tool when it comes to rapidly developing prototypes. When we need to show a prospective client how a solution might work in practice, we can generate a demo within a few days and have it deployed on Firebase so the client can experience what the end result is going to be.”
Fast-forward to ubiquitous agentic AI
Le Roux sees agentic AI becoming ubiquitous and more personal in future.
“When agentic AI can access any data in your organisation, it can work on that data and query models, giving you predictive insights and intelligence so you can do what-if analysis, and you can experiment, and you can plan, and you can do all these things rapidly in one day... what’s next?
"Organisations are now capturing so much more data with IOT, drones, transcriptions of meetings and engagements with customers, and that data is like a gold mine of information around how the organisation works, the health of the organisation and critical issues that might be appearing in the organisation. If you then add all operational data, as well as meeting transcripts to the mix, then there is a future where the agents can jump in and resolve problems or pull in support from people as needed."
Furthermore, he says: “In future, people will have their personal and business agents. They will want to interact with their AI agents wherever they are. So instead of going to Gemini to ask the agent something directly, they will want to interact with it from Slack, e-mail or WhatsApp. They will interact with the same agent from any app or platform. Maybe even talk to the agent over the phone. It will have their context and history, know their meeting schedules, understand their organisation and the work they are busy with, so it can provide exactly the answers each person needs, in a way they prefer.”
“While making food, you might get an idea to solve a problem at work; you can immediately ask your agent to start solving that for you, so that by the time you open your computer tomorrow morning, it is 80% of the way there.
“As another example, at the end of the day, you will climb into the car and as you drive home, have a 15-minute podcast generated by your agent, where it discusses your performance for the day and gives you recommendations on how you can improve, so the next day, you're performing better.”
Share