“South Africa is not short on ambition when it comes to artificial intelligence. What we are short on is execution,” announces Joshua Harvey, head of Growth at Specno, South Africa’s leading digital innovation agency, the company responsible for launching South Africa’s very own Innovators Den.

While global benchmarks show South Africa is around 35–40% behind the United States in AI readiness, enterprise data reveals an even wider gap in execution, with AI implementation rates in South Africa sitting at roughly half the level of the US:

“This gap reflects not a lack of will, but differences in skills, data infrastructure, organisational alignment, and the integration of AI into core business strategy” adds Harvey.

He argues that South Africa’s AI challenge stems from a number of execution barriers, not a single missing piece. These barriers are well documented in recent national and academic analyses:

Skills Shortage and Workforce Readiness

Last year, SAP reported findings that South Africa faces a critical shortage of AI-related skills, which threatens to limit the country’s competitiveness and the ability of organisations to realise value from AI technologies. Without coordinated investment in training, certification, and workplace upskilling, this gap will widen rather than close.

Organisational and Data Readiness

Successful AI implementation is not simply about acquiring tools; it requires robust organisational infrastructure and data readiness.

Research on AI adoption frameworks shows that readiness factors, including data quality, executive leadership support, IT capacity, and available resources, are core determinants of whether AI initiatives succeed or fail.

The study also shows that where organisations lack integrated data systems or strong governance structures, AI pilots often stall and fail to scale.

Low Adoption Despite Acknowledged Value

Even where business leaders recognise the benefits of AI, adoption lags. Studies of South African organisations reveal that many executives understand AI value in theory but are constrained in practice by limited IT maturity, risk aversion, and organisational culture factors that inhibit more transformative adoption.

What the United States is doing Differently

“U.S. firms and institutions have aggressively pushed AI into operational workflows, talent development, and business strategy. Even amid challenges, including debates about deployment scale and workforce impact , American companies maintain strong investments in practical AI applications, cross-functional teams, and data architectures that support industrialisation” says Harvey.

In the U.S., a concerted focus on problem-first deployment, where AI is aligned to cost drivers and operational impediments, has influenced both innovation and productivity.

While adoption is still uneven across sectors, the integration of AI into core functions such as supply chains, customer service, and decision support is demonstrably more advanced than in South Africa.

What can South Africa do to catch up?

Harvey believes that we cannot work from an aspirational future state:

“We must benchmark honestly against where we are today and commit to concrete shifts in capability and practice.”

AI literacy at executive and board level must improve, but this must be matched with practical implementation skills. From data engineers and machine learning operators to product managers who can translate business needs into technology outcomes.

South Africa’s strongest opportunities lie in sectors such as financial services, healthcare, energy, and logistics;  areas where inefficiency is measurable and improvements deliver real value. AI must address tangible, locally relevant problems that matter to the economy and citizens.

Government, Industry, and Academia must Collaborate

“Filling the skills gap and strengthening data ecosystems requires coordinated action across all sectors. National initiatives that support training, certification, and research collaborations can accelerate readiness and ensure South Africa’s workforce is prepared for the demands of AI-driven economic participation. South Africa stands at a pivotal point. The next 12 months will determine whether we remain followers, or whether we narrow the execution gap and build AI systems that are robust, ethical, and commercially viable in our context” concludes Harvey.

Businesses looking to operationalise AI responsibly, pragmatically, and at scale are encouraged to engage with Specno, a local digital innovation agency focused on turning complex technology into viable, production-ready solutions that solve real-world problems.

But anyone who has worked in data engineering knows: building and maintaining these models is painfully slow.

In Nigeria, banks often struggle to merge fintech acquisitions because their data definitions don’t align. In the UK, the NHS faces constant battles over inconsistent fields like admission dates, leading to reporting delays and errors.

Traditional modelling is manual, knowledge-intensive, and brittle often taking weeks or months before systems can “talk” to each other.

Enter Generative AI

Instead of long workshops and endless iterations, AI can now generate draft data models in hours by analysing business glossaries, existing schemas, and even raw datasets. It can suggest alignments, flag duplicate definitions, and automatically produce documentation for compliance.

Imagine a Nigerian insurer unifying its legacy systems with new digital products far more quickly. Or a UK hospital network automatically generating a shared patient data model that works across multiple trusts. Suddenly, integration that once felt impossible becomes achievable.

Why Now?

This idea is fresh. Large language models like GPT-4, capable of this kind of semantic work, have only become available in the last couple of years.

Early adopters are discovering that AI-assisted modelling doesn’t just save time, it preserves institutional knowledge, reduces human error, and produces clearer documentation for auditors and regulators.

The Caveats

Of course, AI isn’t magic. It’s only as good as the data it learns from, and biased or incomplete sources will lead to flawed models. Explainability also matters: decision-makers need to understand why the AI generated a certain schema. That means human oversight is still essential. The role of the architect doesn’t vanish, it evolves. Instead of manual builders, they become reviewers, validators, and guardians of quality.

Why It Matters

If widely adopted, generative AI could allow enterprise data models to evolve as fast as the businesses they support. For Nigeria, that means leapfrogging bottlenecks that have held back digital transformation.

For the UK, it means taming the complexity of massive, fragmented systems in healthcare and finance.

It’s not flashy like self-driving cars or chatbots. But make no mistake: generative AI in data modelling may be one of the quiet revolutions shaping the future of enterprise data.

About Soji Olaleru

Soji Olaleru is a data engineering and enterprise architecture professional with a focus on how emerging practices can transform the way organisations manage information. With experience at the crossroads of innovation. My work often draws on lessons from both Africa and the UK, where the challenges are different but the need for reliable, transparent data is universal

The reality consistently falls short of the promise. These tools are unable to interpret the fragmented, chaotic data pipelines inherent in enterprise systems, leaving data teams still spending 87% of their time organizing data and enterprises spending an average of $4.6 million every year on manual data analysis — until now.

Connecty AI, emerging from stealth with $1.8 million in pre-seed funding, has developed a context engine that tackles the inherent complexity in enterprise data. 

The round was led by Market One Capital, with participation from Notion Capital and data industry experts including Marcin Zukowski, co-founder of Snowflake and Maciej Zawadzinski, founder of Piwik PRO.

Today, enterprise data teams navigate complexity across three critical dimensions: horizontal data pipelines (including multi-source ingestion, multi-cloud data warehousing, data lineage tools, and cataloguing systems), diverse consumption patterns (spanning CRM systems, BI dashboards, and machine learning applications), and distributed human knowledge across roles like data engineers, analysts, governance teams, and functional managers. 

While early AI solutions attempted to automate data workflows by interpreting complex schemas, these models fall short in enterprise environments. Even 90% accuracy isn’t enough when dealing with real-world data complexity. 

Large Language Models need more than static schema files; they require a continuously evolving, cohesive understanding across systems and teams.

“Our experience has shown us that effective data management is about more than just technology—it’s about connecting the dots between data sources, business objectives and the people who use them,” said Aish Agarwal, CEO of Connecty AI. 

“Any ad-hoc ‘guerrilla style experimentation’ with LLM data agents can lead to a pilot application but it’s a lot harder to build a production level application that is reliable.”

At its core, Connecty AI does two things: first, it extracts and connects three-dimensional context from diverse data sources and use-cases while integrating real-time human feedback, creating an enterprise-specific context graph.

Second, it leverages this context to automate data tasks across various roles, using a personalized dynamic semantic system. 

The engine operates continuously in the background, proactively generating recommendations within data pipelines, updating documentation, and uncovering hidden metrics aligned with business goals.

During prototype development, Connecty AI has partnered with enterprises ranging from $5 million to $2 billion ARR, validating its approach on real-world data rather than public datasets like Spider. 

The platform connects to data warehouses like Snowflake or BigQuery in less than five minutes with no-code deployment. Early results have been compelling: “Our data complexity is growing fast, and it takes longer to data prep and analyze metrics. We would wait 2-3 weeks on average to prepare data and extract actionable insights from our product usage data and merge them with transactional and marketing data. Now with Connecty AI, it’s a matter of minutes!” said Nicolas Heymann, CEO of Kittl.

“We were impressed with the accuracy of responses from day one. Additionally, Connecty AI generated excellent suggestions to improve the schema descriptions and enhance our semantic layer. It offers a unified flow from prep to querying, nothing like that we’ve seen anywhere else,” added Aditya Upadhyay, Director Analytics, Mindtickle.

Founded by Aish Agarwal and Peter Wisniewski, Connecty AI emerged from their complementary experiences in the data value chain. 

At FL Studio, Agarwal encountered the inefficiencies caused by fragmented data systems delaying business insights, while Wisniewski’s experience building data platforms for Point72 hedge fund and a major European e-commerce player highlighted similar challenges from a data engineering perspective.

The timing of the duo’s emergence from stealth aligns with growing market demand. Businesses are demanding more from AI. According to Future Markets Insights, the global AI Analytics market is projected to grow at a CAGR of 22.6%, reaching $223 billion by 2034. 

As data complexity grows, organizations face increasing costs, with data teams consuming 12.5% of IT budgets—an average of $5.4 million annually, with 87% dedicated solely to data and platform maintenance.

“We are thrilled to back Connecty AI as they redefine enterprise data management with their deep context learning,” said Jacek Łubiński, Partner at Market One Capital. 

“The platform’s ability to unify and contextualize data across fragmented systems presents a massive opportunity for businesses looking to use LLMs for data workflow automation. The vision Aish and Peter have resonates with us and we’re excited to support them on the journey.”

Looking ahead, Connecty AI will expand its context engine’s capabilities across additional data sources and offer it as a service via API. 

In a market flooded with AI tools that promise to replace human analysts but deliver unreliable results, Connecty AI is taking a fundamentally different approach – embracing the complexity of enterprise data environments and augmenting rather than replacing human expertise.