AI comes of age

You can’t have a conversation these days without artificial intelligence (AI) taking centre stage.

There are many good reasons for this: AI has come of age, and is ready to start making a massive impact on how we work, live, and play.

But how should companies be approaching their AI ambitions, and how can the channel ensure it’s got a meaningful role to play in this dynamic new world?

Kathy Gibson reports

 Artificial intelligence (AI) is the wave of the future, with IT teams under pressure from business leaders to quickly create and deploy AI solutions.

Generative AI (GenAI) tools like ChatGPT and CoPilot have made AI accessible – and understandable – to most users, further fuelling the push to make AI a key ingredient in any new application.

But should every application be AI-led? And does AI always look like what the marketing hype has led users to expect?

Once they work out how AI can help them either save money, boost productivity, or increase competitiveness, businesses will start putting real investment into the technology.

“At the end of the day, AI is all about the data,” says Jacques Visagie, GM: AI and services at Pinnacle. “What data do you have stored, and why? What is the value of it? Is it relevant? If there is no value associated with the data, and no return on investment (ROI), there can be no use case.

“This is why so few customers have actually gone into production; and why so many businesses are struggling to justify an AI investment.”

But these business cases are rapidly becoming reality, he adds. Nvidia research indicates that 2025 will be the year that AI shifts from concept to reality. “AI will go into mainstream production in many enterprises this year,” Visagie says.
At the same time, agentic AI is quickly becoming the dominant trend. “AI agents are going to be everywhere, helping to automate complex tasks, make decisions, and adjust as needed,” Visagie explains. “We’re gping to see machines talking to machines and new methodologies emerging.”

As more businesses adopt AI, the way we see infrastructure is going to change. Visagie explains that the data centre with AI services will come to the fore. Meanwhile, new data centre designs will have to take into account radical increases in the amount of power and cooling required for AI computing.

As edge devices become smarter, AI will move out of the data centre to the edge. This will include IoT devices as well as PCs running technologies like CoPilot that can perform AI without being connected to the data centre.

AI is set to ramp up the discussions around data sovereignty that began with the cloud hyperscalers. “Sovereign AI will accelerate adoption,” Visagie believes “Sovereign AI has the potential to catalyse economic growth and enhance national competitiveness. But governance and policy will be key.”

Just about every business around the world is talking about AI, but there is still a lot of hype around the topic.

President Ntuli, MD of HPE South Africa, believes that many businesses are still battling with how to get a true return on investment (ROI) out of AI.

This doesn’t mean companies aren’t making those investments – around the world generative AI (GenAI) spending is expected to increase by 60% over the next three years.

“AI is poised to surpass the impact of any previous technology, transforming every business and industry while addressing some of the most significant challenges of our time,” Ntuli says. “However, while basking in the possibilities, it can be easy to overlook the fact that AI is merely a tool to achieve extraordinary objectives, not the objective in itself.”

Although 75% of executives worldwide, including those in South Africa, consider AI a strategic priority, 60% fail to establish or monitor financial KPIs related to AI value creation.

“This gap arises because AI projects are often viewed in isolation, rather than as integral components of the organisation’s broader business objectives,” Ntuli says. “Consequently, many AI initiatives struggle to gain traction as businesses find it challenging to assess their return on investment (ROI).”

Another problem is that, often, AI is implemented from a technological standpoint, but the IT director or CIO is unable to clearly articulate the actual benefits to the business.

Key to a sustainable AI implementation is strong data infrastructure, capabilities, and governance, Ntuli explains. “AI initiatives rely on high-quality, often company-specific data. A hybrid by design approach, integrating edge-to-cloud, reduces latency and enhances data control.

“It is essential to align your AI strategy and existing processes like data lifecycle management, security, and operational planning with each other.

“Without this foundation, AI projects often fail to move beyond pilot stages. A strategic approach ensures an AI application’s potential is fully realised, driving meaningful business outcomes and innovation.”

AI also needs to be demystified in terms of its perceived complexity and cost, he adds. “Organisations can begin with small, impactful projects and scale as they achieve success, but it is crucial to start with a clear strategy linked to tangible business outcomes.

“These first steps are always the hardest, especially for those lacking the know-how and in-house talent. This is where vendors and the channel can support, providing expertise and supporting organisations throughout their journey.”

Anton Herbst, CEO of Tarsus on Demand, believes that many people’s perception of AI has been shaped by the rapid emergence of generative AI (GenAI) tools like ChatGPT and CoPilot.

“I think GenAI has value, but AI is a bigger discussion,” he says.

Today, companies are looking at AI from two angles, he adds: the first is to help them optimise their business; and the second is using it to build out a new business based on AI, with new business models that use AI to assimilate and analyse data.

In the first use case, Herbst points to the value that can be achieved by using AI tools to run the business better and gain deeper understanding of customers.

“But companies need to realise it isn’t about simply using AI to replace people, but rather aiming to make their people more productive.”

In the second instance, where new digital businesses based on AI are springing up, Herbst cautions that these need to have a very clear understanding of what problem they are solving.

“In the same way we have always built businesses, they need to identify the customer together with opportunities and challenges, and then look at how they can use AI to address those. And bear in mind, they might need to combine several kinds of AI to do that.

“If new businesses are not taking that approach, then they are just generic and offer no new value.”

And, for any business, data is key to how successful it will be. “Without good, clean data no business will be able to thrive,” Herbst says. “This has always been the case – but, with AI, businesses can get into trouble a lot faster.”

The first flush of hype is fading, and Herbest believes more sanity is coming into the AI discussion.

“The pace is not slowing, but we are now hearing more proper business discussions about GenAI and agentic AI,” he says. “We are also seeing a renaissance for RPA, and the opportunity to combine it with agentic AI.”

Around the world, real business solutions are being actively deployed with AI, and many of them are starting to reap real business value, says Herbst.

Barry Buck, chief technology officer of Saucecode and author of the locally-built Roboteur platform, agrees that there is still a lot of confusion in the market.

“There’s always a lot of new terminology with any new technology and this can tend to cloud things a bit,” he says.

The biggest confusion is usually around the concept of automation – which is certainly not a new application by any means, but which can be greatly enhanced with AI.

“AI can automate, but automation isn’t necessarily AI,” Buck points out. “Automation is a set of tools that do something – open files, find data, extract data, structure it etc. These are all actions that can be made against software, and can be either specific to a task or generic.

“But the bottom line is that they are tools: they cannot do anything on their own.”

The next step is robotic process automation (RPA), where the developer – understanding the business process that needs to be automated – creates a workflow that uses automation tools to perform actions in software that a human would normally do in order to achieve the business process or goal.

“So, although this is more intelligent use of the tools, once the workflow is built it is quite static,” Buck explains. “RPA doesn’t perform decision-making on the fly: most of the decisions are actually taken ahead of time when the developer programs the logic statements. These define the process flow and allow for static decision-making.”

Automation, he concludes, is where software tools like bots use the software in a way that people would normally do, to perform a specific workflow.

AI, on the other hand, can go beyond the pre-defined workflows.

“A term that is getting thrown around a lot is agentic AI,” Buck explains. “Your typical automation project has a defined workflow that won’t change while its running: it’s predictable, testable, and generally secure.

“With agentic AI, the objective is typically not clearly defined, there is no pre-defined route for the process to take, and no guaranteed steps that will happen – so it’s very dynamic.

“The agent will have a set of tools, and access to context, but no pre-defined steps on how it should reach its outcome. It will perform an action, then define the prompt for the next action, continuing until it reaches the defined directive.”

It’s not quite as freewheeling as this sounds: developers would define inputs and outputs and set guardrails in place. And, unlike an automated workflow that performs clear steps, agentic testing and training will require a much larger set of use cases that encompass all of the possible scenarios that could be encountered.

“For document processing, AI is really good at extracting fields and values, even in documents that have been poorly scanned,” Buck says. “We are also using it where customers used to struggle with mapping fields, or different document types. Making sense of these has all been made a lor easier with AI models.”

Context is another area where AI is proving useful, as it is able to understand human communications and use those to trigger or make a decision about document processing or workflow.

Web automation is another area where AI is proving its mettle, doing away with the need for strict locators to interpret Web pages despite the challenges of dynamic changes.

“AI could be a game-changer,” Buck says. “Because AI understands context, once you tell it what you want to look for it can enhance workflow with low-risk benefits that offer high yields.”

The potential for these tools is immense, particularly for creative tasks or contextual search, Buck says. But we need to be cautious when using them for more complex processes. “Agents could introduce risk, so we need to be meticulous is setting guardrails and ensuring security.”

Andre Nel, head: digital client enablement at Nedbank, puts the related concepts of automation and AI into perspective.

“Automation is an overarching term for any technology that takes people’s repetitive tasks and relegates them to a form of software,” he says. “In most cases this is rule driven with minimal variance in the outcomes.

“Artificial intelligence (AI) adds the intelligence to the technology. This is where the automation gains the ability to make decisions through repetition, that were not coded as rules. AI learns (machine learning) the repetitive behaviour and imitates it without the human coding.”

 

The channel’s role

Visagie recognises that reseller partners are the key to helping customers implement AI – but many of them still don’t have a firm understanding of the technology themselves.

He points to a recent introduction to AI course, where just three of the 70 attendees passed at the end of the day. “So there is still a lot of confusion in the channel when it comes to the different types of AI.”

That said, Visagie believes that partners must be able to provide guidance to their customers; to help them with ideation, with identifying opportunities, with crafting solutions, and with implementation.

The first step, he says, is to help customers identify where AI could add value to their business – the ideation phase.

Thereafter, the ambition phase would look at coming up with AI-driven scenarios that could address the idea; followed by the use case where specific solutions are matched to the idea and ambition; then the project plan and, finally, implementation.

IT reseller partners could again find themselves disintermediated by emerging technology trends if they don’t pivot quickly.

Herbst points out that many companies, especially small and medium enterprises (SMEs) are battling to find value in AI.

There is an opportunity here for partners to become trusted advisers, and guide them on their AI journeys. “The issues around how to architect and implement the business model don’t go away,” says Herbst. “Then there are issues around the data such as privacy and security, along with governance, compliance and risk.

“These are all opportunities for partners to assist their customers,” Herbst adds. “They should consider their customers’ business and use design thinking to find relevant opportunities for them.

“The partner would help customers make sense of all the many tools and agents proliferating out in the market.”

Digital-native partners, and those that have made the transition to digital-first, are already having these discussions with their customers, Herbst says. And digital-first end users are the ones that are already finding applications for AI.

“The first thing to remember is that you can’t run AI on manual processes.”

So digitalisation and IT modernisation is still very much an opportunity for partners.

More advanced AI discussions would be in helping customers assess their AI readiness, and helping them map out the journey.

“For me, that consulting piece is vital,” says Herbst. “And we are seeing the channel role move in that direction. The cloud partners are already doing more consulting and advising than reselling.”

Change management is another opportunity for partners, he adds. The impact of AI could be profound, and needs to be properly managed. “We could be talking about up to 100-million jobs lost,” Herbst points out.

 

Where does distribution fit in?

It might look like traditional distribution’s role in the AI world is limited, but Herbst believes it is more important than ever.

“Distributors are having to play a bigger role in enablement, in taking their partners from where they are to where they need to be,” says Herbst.

“The distribution role has typically been in aggregation of stock or credit or people, but I think it’s going to shift to more of an aggregation of professional services.”

“I think distributors are going to have to step up to help partners with consulting to their customers, with assessments, and then the pooling of services. As partners build capacity and capability, distributors will play a key role in guiding them and helping them to scale. Whenever you need scale in the mid-market and SMB space, distribution needs to step in.”

Building an AI practice is not a trivial task, and many partners will struggle to make the journey alone. Just getting to grips with the thousands of tools and models available is a mammoth task. “But the distributor helps to sift through all the offerings and bring the right solutions to market.”

Pinnacle was the first South African distributor to launch a dedicated AI practice, and Visagie says it’s doing a lot of training and enablement among its resellers.

“We are helping our partners to have those conversation with their customers, and to change their own businesses as needed,” he says.

As they pivot to taking on an AI consulting role, Pinnacle assists them with AI workshops with customers where they create a journey map, outline use cases, identify stakeholders, and provide guidance on how to implement AI into their processes and, ultimately, into the process map and the business.

These conversations must be informed by AI governance, including data observability and governance structures, Visagie adds. “This is very important.”

In the IT industry, the incidence of technology companies using their own solutions – or “eating their own dog food”, as the saying goes – is not as common as one might think.

“For Pinnacle’s AI practice, our first customer is Pinnacle,” says Visagie. “We are adopting AI within the organisation and going through the same process that we are urging our partners and their customers to embark on.”

Among the lessons learnt is that there is no global AI requirement; rather, different departments all have different requirements.

“For instance, the warehouse is looking to optimisation and stock control,” says Visagie. “On the other hand, marketing wants to use external tools to create content and incorporate it into the business.”

These different ideas and ambitions need to be implemented under an umbrella governance and tools policy, and must use the products and services supplied by Pinnacle.

“We have invested in two AI servers, with a lot of GPUs, and we have invested in skills, including a data scientists and a DevOps developer,” Visagie says. “Our deadline to have cool stuff that works is September; and by November we will have customer-presentable AI that we have designed and deployed.

“ So we are graphically demonstrating our commitment and experience by doing it all internally.”

As a value-added distributor, Pinnacle believes it has a role to play in taking the complexity out AI. “We can help our resellers build AI solutions,” Visagie says. “We understand the technology and we have relationships with the OEMs, so we can be the central point, offering options that help reduce the time to implementation.

“As a distributor, we are not pushing a vendor’s agenda, but can talk about what the customer needs and guide them in the right direction.”

 

The end user perspective

Most of South Africa’s enterprise companies are using artificial intelligence (AI) to some degree, and the big banks are leading the charge.

One of these is Nedbank, according to Andre Nel, head: digital client enablement at Nedbank.

He explains that Nedbank views AI and automation as transformative tools that can significantly enhance operational efficiency, customer experience, and overall business performance.

“The bank’s journey in intelligent automation has been marked by relentless innovation, strategic execution, and a culture of continuous improvement.

“From eliminating manual inefficiencies to embedding automation into core processes, Nedbank has transformed its operations, delivering measurable financial impact and creating seamless, customer-centric banking experiences.

“The problem statement that is required to be solved will include all or a best-fit combination of the various tools,” he explains. “The goal always remains financial benefit and customer experience.”

Nedbank employs a mix of solutions to address various challenges through AI and automation:

Robotic process automation (RPA): Nedbank’s journey in intelligent automation has been marked by relentless innovation and strategic execution. The bank has embedded automation into its core processes, eliminating manual inefficiencies and delivering measurable financial impact. RPA is used to automate repetitive and mundane tasks, allowing employees to focus on more intelligent and high-value work. This not only improves efficiency, but also enhances service quality, delivery, and compliance with a customer-centric banking experience as the goal.

Intelligent automation: Combining RPA with AI technologies such as optical character recognition (OCR) and natural language processing (NLP) enables the automation of more complex processes. This approach helps in analysing, categorising, and extracting unstructured data, making it functional and improving the output of automated business processes. Again the focus is on efficiency management with customer experience in mind.

Hyper automation: For tasks with variable inputs and treatment, Nedbank adopts hyper automation, which involves advanced, next-generation automation technologies. This approach maximizes returns.

“Nedbank has been actively leveraging AI and automation to transform its operations, enhance customer experiences, and drive business performance at all levels in the organisation,” Nel says.
“The bank’s vision for AI involves using data and AI for commercial advantage, building data assets, and prioritising responsible use of AI. The focus is on accelerating innovation, enhancing customer experience, strengthening competitive advantage, and optimising operations through GenAI.”

To ensure continued innovation, while providing growth and security for its employees, Nedbank’s people are constantly being upskilled in all of the technologies, Nel adds.

This is not to say that implementing AI is a trivial exercise.

“In Nedbank, as with many other big corporates, I believe there are three major challenges that need to be taken into account.”

These are:

Change management for staff impacted either directly or indirectly. “It’s important to show them the benefits of removing mundane tasks and letting them focus on their true job activities.”

The financial break-even of the RPA or AI applications versus the business case benefit being solved.

Integration issues associated with small secondary systems and processes integrating into the corporate mainframe systems.