Firms warned against building AI systems around single models

Artificial Intelligence. Photo: Jobaaj Learnings

Businesses should avoid building artificial intelligence systems around a single model, as rapid technological advances could leave them with inflexible infrastructure, rising costs and difficulties adapting to new capabilities, a technology executive has warned.

Vice-President of Technical Product Sales at Core42, Will Badr, gave the warning during a panel discussion on enterprise AI adoption at the AI summit in Abu Dhabi, United Arab Emirates, where technology executives examined the challenges of moving AI projects from experimentation into business operations.

Badr argued that companies needed to develop technology architectures capable of accommodating different AI models as the technology evolved, rather than relying on specific models that could become outdated.

“The competitive advantage here is not having access to frontier models,” he explained. “It’s going to be about having the right architecture, the right infrastructure that is designed around all of these models and adapts to this continuous innovation.”

He noted that while AI models would continue to change, infrastructure would remain in place for longer, making flexibility an important consideration for organisations investing in the technology.

According to Barr, AI systems also differ from conventional software applications because they can handle variable demands, draw on multiple data sources and, in some cases, use autonomous agents to execute tasks.

He warned that deploying AI on enterprise architectures not designed for such workloads could create technical debt, complicate data management and expose sensitive information to security risks.

The discussion highlighted a broader challenge facing businesses seeking to scale AI beyond pilot projects, as they balance the technology’s potential benefits against implementation costs, data governance and operational requirements.

Vice-President of Solutions at French AI company Mistral AI, Corentin Petit, urged organisations to begin AI transformation by identifying business problems the technology could solve rather than adopting it simply because it was available.

Petit outlined a framework for evaluating AI projects based on four considerations: whether senior management supports the initiative, whether it addresses an urgent business problem, whether it can reach production, and whether it can be delivered within three to six months.

Related News

• ‘Professionals must retain independent thinking amid AI growth’
• Nigeria, UK deepen ties on AI, digital trade standards
• Evening recap: FG approves Ebola stipend, Tope Alabi threatens lawsuit, other top stories 

He cautioned that companies risk investing in projects that deliver little value when they prioritise technology over clearly defined business needs.

Petit also highlighted the need to customise AI models to meet specific business requirements, improve speed and control costs.

Senior Vice-President of Data Analytics and AI at Solutions+, Kamal Shah, stressed that reliable data, clear accountability and governance were essential to ensuring that AI projects delivered sustainable business value.

He called for AI to be embedded in organisations’ core operations, with designated business owners responsible for projects beyond the proof-of-concept stage.

“AI demos and flashy use cases definitely get attention, but governance earns trust,” Shah remarked. “And without trust, there is no AI strategy.”

Shah further emphasised that data governance should be established at the beginning of an AI project rather than introduced after implementation had started.

Senior Director of Technology Ecosystems at DMCC, Belal Jassoma, maintained that both established businesses and newer companies could become AI-native if they had access to appropriate technology, infrastructure, reliable data, capital and pilot programmes.

He highlighted the importance of the wider business ecosystem in helping organisations move AI initiatives from experimentation into production.

The panel’s discussion underscored the need for businesses to align AI investments with operational priorities while developing infrastructure capable of adapting to changes in models, costs and capabilities.