Technology firms are developing smaller artificial intelligence (AI) systems that can be hosted locally within business premises, targeting businesses seeking to adopt the technology without investing in expensive computing infrastructure or relying entirely on cloud services.
Smaller, locally hosted AI systems, often called small language models or SLMs, are lightweight AI models that run directly on personal devices such as a laptop, desktop, or phone—rather than on remote corporate cloud servers.
In Kenya, tech firm Dell and NTT Data, a data centre operator and cloud service provider, are now targeting small businesses with smaller servers as AI adoption picks up pace in the country.
Andrew Ngunjiri, general manager for technology solutions at NTT Data, says small businesses now don’t need large data centres to benefit from enterprise AI, even as firms race to localise AI servers amid strict data protection and governance laws.
“You don’t necessarily need to have a whole floor or a building such as this to get the benefits of sovereign or enterprise AI,” said Mr Ngunjiri.
“You can start small. We do now have offerings that support what we might be able to call ‘AI in a box’, where you’re able to run some of the smaller models to run your business applications.”
He said the solutions could be particularly relevant to businesses with compliance or regulatory requirements that favour keeping their data within their own infrastructure.
Small language models are more compact and efficient than their large model counterparts. As such, SLMs require less memory and computational power, making them ideal for resource-constrained environments such as edge devices and mobile apps.
The shift toward smaller systems comes as Kenya seeks to expand the use of AI across its economy. The National Artificial Intelligence Strategy 2025–2030 identifies digital infrastructure, data governance, research and innovation as key pillars for developing the country’s AI ecosystem.
For smaller businesses, locally hosted AI could offer greater control over sensitive information and allow selected applications to operate without sending every request to an external cloud platform. However, the benefits will depend on the cost of equipment, maintenance, electricity and technical support.
Mr Ngunjiri said companies were increasingly adopting a hybrid model, combining public cloud services with private infrastructure depending on the nature of their workloads, costs and regulatory obligations.
“The discussion is not an either-or; it is a journey,” he said, adding that businesses needed to weigh their investments against the value generated by AI.
Kenya’s data protection framework regulates the processing and transfer of personal data, but does not impose a blanket requirement for all business data to remain in the country.
The availability of smaller systems could lower the infrastructure barrier to AI adoption, although the cost and commercial availability of Entity Data’s proposed SME solutions have not been disclosed.
