Copilots, AI agents, and so on: artificial intelligence can be found in many forms and names today. The technology is changing so rapidly that it’s impossible to put a simple definition on the term, even for AI experts.
‘What exactly is AI?’: the question causes a brief moment of silence during the round table. ITdaily brings together five experts from the AI world to discuss the wonderful world of artificial intelligence. At the table are Gianni Cooreman, Presales Director at Salesforce, Joachim Ganseman, Research Consultant at Smals, Maarten Callaert, co-founder of Paperbox, Lander T’Kindt, co-founder of Donna, and Christophe Robyns, Managing Partner at Agilytic.
The panel quickly comes to the unanimous conclusion that there is no single definition for AI. Ganseman gives it a try: ‘There is no clear description for the term ‘artificial’, nor for ‘intelligence’. So how would you explain ‘artificial intelligence’? It’s a philosophical and semantic discussion that doesn’t get you anywhere in practice. According to some descriptions, even your thermostat is AI’.
Digital Employees
In the last two years, we have experienced an explosion of AI models. While last year Copilot was used for every new announcement, now the so-called ‘AI agents’ dominate the debates. ‘Many tasks you can do with AI are text- or audio-based. AI is very strong in that. I also regularly use AI tools to structure my own thoughts,’ Callaert admits without shame.
‘We have experienced a wave of generative AI, which is still in full swing,’ Cooreman observes. ‘Companies can now imagine what AI entails, but it’s difficult to express it in productivity figures. The next evolution is AI agents that perform routine tasks more autonomously. There’s a shift towards digital labour. It’s economically difficult for many companies: efficiency is top of mind for everyone’.
‘This is also very much alive among SMEs,’ Robyns chimes in. ‘Many pilot projects have been carried out on a personal level. Now companies are looking for how they can apply productivity at a company level, but many still don’t know exactly how to do that’.
Matter of trust
For AI agents to take over tasks, humans will need to gain trust in the technology. That doesn’t come automatically, T’Kindt knows, comparing it to a well-known car manufacturer. ‘Tesla came up with AI technology for cars ten years ago. Back then, we didn’t trust it. It’s no different now: AI has to be, so to speak, at least ten times better at a task before people trust it’.
‘It’s important that we learn to understand that AI is not 100 percent accurate. But neither is a human. That’s why we need to clearly delineate where AI can add value,’ he notes.
Ganseman looks from the perspective of the public sector. ‘Adoption is more difficult in government. It can work at the level of personal productivity. AI is so embedded in the software stack of vendors that you have no choice but to embrace it. But it becomes completely different when we’re going to have a counter clerk use AI. There, we still have many legal uncertainties’.
Need for Guidance
The AI Act should provide clarity here, but in Belgium, there’s still work to be done, Ganseman notes. ‘The AI Act will come into force step by step. There should be a Belgian AI Office, but I haven’t heard anything about that yet. As far as I know, it’s not in the new federal coalition agreement. Moreover, the regulators use a completely different definition of AI in the legal text. This way, you can’t find any guidance’.
‘We work on a different scale, but it would also help us if there was a clear regulatory framework,’ Callaert chimes in, who focuses on the banking and insurance sector with Paperbox. ‘Now we refer to GDPR and other European directives to show clients that our product complies with regulations, but Belgian rules would make it easier’.
Callaert shows understanding for the regulators. “To speak a bit in their favor: I understand that they are trying to find a description that is generally accepted. But it’s moving so fast that a general term is difficult to find. Even terms that have been common for years, such as self-learning systems, are already outdated.”
Companies should take matters into their own hands and not wait for regulators, Cooreman and T’Kindt believe. Cooreman: “It’s no secret that in Belgium we are not very innovative but rather risk-averse. Companies want to get a feel for a technology in small groups before rolling it out”. “That’s why it’s important that somewhere a human remains in the loop to build that trust,” T’Kindt adds.
Marketing Machine
The tech companies should also look at themselves. The launch of ChatGPT at the end of 2022 kicked off a marketing machine that seems to have no end. Every announcement must contain the acronym AI somewhere, causing companies to further hollow out the definition of the term, if it even exists. We ask the panel what they think about this.
“For us, it’s actually the opposite. I catch myself saying too little that we are an AI company. We could actually jump much more on the commercial story. It’s interesting for companies,” Callaert notes. T’Kindt supports his neighbor from Ghent. “AI certainly helps to open the door. But just saying you do something with AI is no longer enough. Companies expect you to be able to demonstrate return on investment.”
“It again depends on your definition of what AI is. For some companies, the smallest bit of automation already equals AI,” Robyns notes. “They then ask us what AI can do for them. I always answer that it depends on what their problem is. At Agilytic, we prefer to say that we also do AI. Trying to put a definition on it doesn’t help. The problems we want to solve should be central, whether with or without AI.”
“In any case, it no longer impresses me. If I read something about AI and it’s not clear to me what and how, I immediately tune out. We need to be able to see through the hype,” Ganseman adds.
AI Gurus
“All vendors talk about AI today,” Cooreman chimes in, not denying that his employer actively contributes to the buzz in the market. “Many companies have some kind of ‘AI guru’, but they can barely set up a proper workflow. It’s becoming harder for companies to separate the wheat from the chaff. Showing a demo is no longer enough; companies want to see what AI can do in practice with their data.”
“I think it’s becoming important to train salespeople well. There’s a trend in the market that tries to push you towards the full option. Companies want to start with step three, while they haven’t taken step one yet. We try to be careful with that,” says Callaert.
“It’s an interconnected phenomenon. It’s not easy to estimate where the technology will be in six months. I remain amazed by the evolution of voice technology: even support for the West Flemish dialect is already very good today,” T’Kindt says with a wink. “But it’s impossible to know now how much the technology will improve further, because this depends on so many factors that you don’t have control over.”
“Product roadmaps are becoming fluid. The technology is evolving so fast that it’s also becoming more difficult for vendors to accurately estimate how quickly software evolves from beta to general availability. The consumption models of AI are creating a new way of selling. The shift is also happening within our company: CRM salespeople need to be able to talk about data and AI. Salespeople need at least a bit of technical expertise,” Cooreman concludes.
This is the first editorial article in a series of three on the theme of AI in practice. Click on our theme page to see all the articles from the round table, the video, and our partners.