SRI LANKA MUST BUILD DATA SOVEREIGNTY BEFORE AI BOOM, EXPERTS WARN AT AAT CONFERENCE 2026

Sri Lankan organisations must urgently develop their own data sovereignty frameworks and invest in artificial intelligence (AI) literacy across all levels of the workforce, industry leaders said, ahead of the enactment of the Personal Data Protection Act (PDPA) in January 2027.

The remarks were made during a panel discussion titled Innovation, AI and Data in Finance: Data to Decisions, held on the sidelines of the AAT Sri Lanka Conference 2026, themed Precision of Power, at Waters Edge, Battaramulla, on 8 and 9 September.

The panel, moderated by Dehan Vithana, Lead Data Evangelist of MAS Holdings, brought together Sampath Jayasundara (CEO/ Director hSenid Business Solutions/ Chairperson SLASSCOM), Anisha Dharmadasa (Managing Director, Nawaloka Holdings), Professor Nuwan Kodagoda (Deputy Vice Chancellor, Faculty of Computing, SLIIT Campus), and Sanali Kaushalya (Country Director, Sri Lanka, Women in Tech Global).

Opening his presentation on organisational AI strategy, Jayasundara told finance professionals: “AI will not replace people, but AI will replace people without AI.”

He noted that while 97% of finance departments had adopted AI in some capacity, only 42% had broadly or fully embedded it into core operations. Citing the World Economic Forum’s 2025 report, he said 90 million traditional jobs would be lost globally and 170 million AI-transformed roles created, with entry-level positions giving way to roles such as Finance Data Translators and Strategic Business Partners.

Jayasundara said finance teams currently spent 90% of their time producing numbers, and AI could reduce this to 35%, freeing capacity for strategic advisory work. He warned leaders to monitor token consumption costs, which in Sri Lanka could exceed a junior account executive’s salary. He stressed that auditability and human-in-the-loop validation must be established upfront, and urged institutions to build an internal “Finance Brain” using private LLM architectures.

Accountability at the core of responsible AI

During the panel discussion, Kaushalya said accountability remained central to the ethical deployment of AI in corporate decision-making, cautioning against organisations outsourcing sensitive judgement calls to generative AI tools without adequate oversight.

“If you put employee data into Excel or HR systems and ChatGPT decides an employee should not get a salary increment, are you comfortable with that decision? You cannot tell an employee ‘ChatGPT generated that you are not getting an increment this year’,” she said.

She added that developers building internal recruitment systems needed to weigh the ethics of feeding candidate data, including CVs and photographs, into external AI models. “We cannot unethically expose candidate data. Organisations should build their own internal tools and applications within ethical policies to protect data,” she said.

On barriers to AI adoption, Kaushalya said access to AI tools within organisations was frequently uneven. “In a finance department, not everyone has password access. Authorisation should be scaled by seniority,” she said, adding that finance and human resources teams were often under-resourced for AI training compared to software development units.

Healthcare sector cites gains, acknowledges PDPA learning curve

Dharmadasa said Nawaloka Holdings had been an early adopter of AI diagnostic technology in the region and had recorded consistent benefits from its use in patient care and administrative functions. “It has always been opportunities for us, not challenges,” he said.

On data protection, he said patient and financial records had been secured through vaults and automated cloud storage systems for over three decades, though the organisation was still familiarising itself with the requirements of the incoming PDPA. “We are still learning about it as it will be enacted soon,” he said.

Data sovereignty and rising AI costs flagged as key concerns

Jayasundara said mature organisations such as banks, hospitals, and insurers required a dedicated data sovereignty layer, noting that Sri Lanka’s domestic computing infrastructure remained insufficient to meet AI processing demands. “We must rely on reputed cloud providers he said, adding that the PDPA, due to take effect on 1 January 2027, would require organisations to formally assign responsibility for data protection and ethical AI use.

He cautioned that startups, while freer to experiment, needed to closely monitor operational costs linked to AI tool usage. Citing a tenfold increase in GitHub Copilot pricing in June 2026, he said token budgets allocated for a month had in some cases been depleted within a day or two. “Startups should experiment with open-source AI tools while monitoring token usage. Don’t stop innovating, but carefully manage token consumption when commercialising products,” he said.

Universities urged to move towards agile curricula

Professor Nuwan Kodagpda said AI literacy, alongside the ability to critically evaluate AI-generated output, had become an essential skill across all sectors. He warned that uploading sensitive company data to free, publicly available AI tools carried the risk of that data being absorbed into public model training. “The most critical skill today is maintaining human-in-the-loop oversight and developing the ability to critique AI outputs,” he said.

He added that conventional four-year university degree cycles were no longer suited to the pace of technological change. “Universities cannot wait four or five years to update curricula; we must adopt an agile curriculum updated annually,” he said.

Panel rejects fear-driven narrative on AI

Responding to an audience question on why global discourse on AI tended to foreground fear over opportunity, Jayasundara pointed to Sri Lanka’s workforce potential, saying that training a million AI engineers from among the three million young people expected to enter the labour market over the next decade could help address the country’s broader economic challenges.

Dharmadasa said Sri Lankans were well positioned to adapt. “The advantages of AI far outweigh the risks,” he said, adding that younger professionals increasingly preferred to use AI to solve complex problems rather than perform repetitive tasks.

Professor Kodagoda said AI had lowered the barrier to technical creation altogether, noting that prompt engineering now allowed individuals without formal programming backgrounds to build software and websites, a shift being reflected in curricula across disciplines including civil and electrical engineering.

Panel’s closing recommendations

Asked for one practical step finance professionals could take immediately, the panellists offered the following recommendations: Jayasundara called for finance departments to be reorganised around AI capabilities; Dharmadasa urged a focus on data insights with direct revenue impact; Professor Kodagoda recommended assigning finance teams the task of critiquing AI-generated reports to identify blind spots; and Kaushalya advised professionals to learn a new skill weekly, with an emphasis on entrepreneurship.

Jayasundara, in closing remarks, urged organisations to begin incremental adoption.

“Start somewhere. Take small steps, establish data sovereignty guardrails, and be the person who uses AI to replace old ways of working,” he said.