Beyond LLMs: Why RAG is the next big thing in GenAI

RAG connects GenAI to real data, making it smarter, safer, and more useful for South African enterprises.
Beyond LLMs: Why RAG is the next big thing in GenAI
From LLMs to RAG

South African enterprises have embraced the GenAI wave but adoption is only half the battle. That’s where RAG (Retrieval-Augmented Generation) comes in.

It’s one thing to deploy large language models (LLMs). It’s another to make them useful, trustworthy, and aligned to your business outcomes.

The next step? Grounding your AI systems in real-world, real-time data.

“The real magic doesn’t just come from having an LLM in your tech stack,” says Tony Bartlett, Director of Data Centre Compute at Dell Technologies South Africa.

“It comes from making GenAI generate tangible ROI.”

What is RAG?

Traditional LLMs are trained on static datasets, which means they’re always at risk of returning outdated, inaccurate, or overly general answers.

Retrieval-Augmented Generation (RAG) changes this by enabling AI systems to “retrieve” and draw on live or domain-specific data before generating a response.

Think of it as giving your AI access to your company’s own knowledge base, but without needing to retrain the entire model.

Why does RAG matter?

Bartlett explains that RAG enhances large language models by feeding them additional, real-time data without requiring retraining. This improves both the efficiency and scalability of GenAI systems, helping businesses adapt as their needs evolve.

Because RAG can pull from multiple internal and external sources, it allows AI tools to generate responses that are more current, context-aware, and aligned with enterprise-specific requirements.

According to Precedence Research, the global RAG market is forecast to grow from $1.2 billion in 2024 to over $67 billion by 2034.

The rise of agentic AI

But RAG isn’t the endpoint. The next evolution is Agentic RAG: a model that not only answers your questions but acts on them.

Imagine AI that can schedule meetings, resolve support tickets, or manage supply chain tasks without constant human input.

Bartlett explains: “Agentic systems […] are proactive and collaborative, operating in increasingly human-like ways.”

He says this “evolution opens doors to greater organisational agility, real-time decision-making, and massive efficiency gains.”

But with great power comes new challenges. As Agentic AI systems become more autonomous, questions around ethics, governance, and transparency become critical.

Strategic moves for CXOs

The GenAI adoption rate in South Africa is soaring.

According to the South African GenAI Roadmap 2025, adoption jumped from 45% to 66% in one year. But adoption alone isn’t enough.

Here’s how local leaders can evolve beyond the hype:

1. Fix your data foundation

You can’t feed junk into your AI and expect magic out. Invest in clean, structured, well-governed data architectures.

2. Establish AI governance now

As systems get more autonomous, human oversight matters more, not less.

Build in audit trails, ethical guidelines, and accountability from day one.

3. Train your teams

You’ll need people who understand AI’s capabilities and limits.

Think AI translators, prompt engineers, and strategy leads who can see the big picture.

4. Start small, scale smart

Don’t launch a moonshot on day one.

Focus on use cases with real ROI: customer queries, internal knowledge access, onboarding documentation, etc.

5. Tie AI to real business outcomes

Whether it’s saving time, improving accuracy, or unlocking new products, measure what matters.

If it doesn’t move the needle, it’s not strategy.

Don’t stop at LLMs

Bartlett says, “For the South African CXO, the mandate is clear: don’t stop at LLMs. Lead your organisation into the next chapter of enterprise AI; one where intelligence is not just generated, but grounded, goal-driven, and trusted.”

As GenAI matures, the shift will be about better business outcomes, smarter automation, faster decisions, and deeper insights.

Use it to build a business that doesn’t stop at efficiency but goes beyond: more intelligent, adaptable, and ready for the future.


Editor’s Note: This article is based on insights originally shared by Tony Bartlett, Director of Data Centre Compute at Dell Technologies South Africa. His commentary has been edited for clarity and relevance to our readers.

Sharing is caring! 

Featured reads: