Sri Lanka has been pulling sapphires, rubies, and a rainbow of other gemstones out of its earth for over 2,500 years. The Ceylon Blue Sapphire alone carries a reputation that commands premium prices in Bangkok, New York, and Geneva. Yet the island's gem and jewellery trade, one of its most storied export sectors, has struggled to translate that heritage into the revenue it should. The government set a target of $1 billion in gem exports for 2025, but actual annual earnings have hovered closer to $300 million in recent years. Much of the gap comes down to something unglamorous: an industry still run largely on informal networks, manual grading, and trust built on personal reputation rather than data.

That's exactly the kind of gap artificial intelligence is built to close. Not by replacing the gemologists and miners who've carried this trade for generations, but by giving them tools that make the industry more transparent, more efficient, and more competitive on the global stage.

The Trust Problem AI Can Solve

Colored gemstones are notoriously harder to grade consistently than diamonds. Two experienced gemologists can look at the same sapphire and disagree on origin, treatment, or clarity grade. That inconsistency costs the industry money and credibility, especially as synthetic and treated stones flood the market and buyers grow warier.

This is precisely where AI has started to move from novelty to necessity elsewhere in the trade. Swiss International Gemlab launched an AI-assisted grading system in 2026 that cross-references spectral and analytical data against structured databases to flag anomalies and reduce inconsistency between reports. Diamond labs have been using machine learning models for years to analyze color, clarity, and cut with far more consistency than the human eye alone. There's no technical reason the same approach  spectroscopy plus machine learning trained on a large reference dataset  can't be applied to sapphires, rubies, and spinels coming out of Ratnapura and Elahera.

For Sri Lanka specifically, this matters because so much of the "Ceylon" brand rests on provenance. An AI model trained on a proprietary dataset of Sri Lankan stones' trace-element and spectral fingerprints could let local labs (working alongside the National Gem and Jewellery Authority, or NGJA) certify origin faster and more defensibly than manual comparison, and do it at a price small dealers can actually afford, not just the handful of stones that go to Geneva.

Provenance One Can Prove

Buyers increasingly want to know a stone's whole story: where it was mined, how it was cut, whether it passed through fair conditions. Gübelin and Everledger built the Provenance Proof Blockchain to do exactly this for colored gemstones, and it now tracks more than 500,000 stones through over 500 organizations worldwide, from artisanal miners to retailers.

Sri Lanka has an opportunity to build a national version of this idea, pairing blockchain record-keeping with AI-driven verification at each handoff point. An AI system could flag inconsistencies in the paper trail, cross-check images of rough stones against the cut versions later submitted for export, and help the NGJA move from a paperwork-heavy licensing process toward something closer to real-time supply chain visibility. That's not just good for buyers abroad; it directly protects small-scale miners from having their stones' origin obscured or undervalued further down the chain.

Fairer Prices for Pit Diggers

A huge share of Sri Lanka's gem mining still happens in small, family-run pits (locally, "gem pits" or illam mining) where miners often sell to middlemen without a clear sense of what a rough stone is worth. Price discovery is opaque, and that opacity tends to benefit whoever has more market information, rarely the miner.

Simple AI tools could change this asymmetry: a smartphone app that lets a miner photograph a rough stone and get an estimated value range based on a model trained on historical auction and export data, or a translation and negotiation assistant that helps small dealers deal directly with international buyers instead of through several layers of intermediaries. None of this requires exotic technology. It requires someone building the dataset and the interface, and making sure it's usable on a basic smartphone in Ratnapura, not just a lab in Colombo.

Smarter, less destructive exploration

Gem mining in Sri Lanka is mostly artisanal and small-scale, which keeps it labor-intensive but also means a lot of digging happens speculatively, with real environmental cost to land and waterways when it doesn't pay off. AI-assisted analysis of satellite imagery, geological survey data, and historical yield records, an approach already used in mineral exploration elsewhere  could help identify which gravel deposits are actually worth mining before the ground gets disturbed. That's a win for miners' odds and for the environment they depend on.

Lowering Marketing Barriers

On the demand side, AI is already reshaping how gemstones get sold globally: visual search tools that let a buyer photograph a stone and find similar ones, AI chatbots that can answer basic buyer questions in multiple languages at 3 a m Colombo time, and generative tools that help small exporters produce professional product photography and listings without hiring a full marketing team. For an industry made up mostly of small and medium businesses trying to reach buyers in the US, India, Hong Kong, and the Gulf, these tools lower the barrier to competing globally without needing a big marketing budget.

Role of NGJA

None of this happens automatically. Sri Lanka's gem sector is dominated by small, informal operators who don't have the capital or technical literacy to adopt AI tools on their own, and there isn't yet a large, well-labeled dataset of Sri Lankan gemstone spectral data to train models on in the first place. Building that dataset is arguably the single highest-leverage investment the industry could make, and it's a natural role for the NGJA, the Gem and Jewellery Research and Training Institute, and university partnerships to play together, rather than leaving it to individual labs or foreign firms.

There's also a reputational risk in getting this wrong. If AI grading tools are adopted unevenly or without transparency about their limitations, they could just create a new source of disputes rather than resolving old ones. The technology has to be introduced as a support for gemologists' judgment, not a black-box replacement for it which, notably, is exactly the framing the labs already using AI abroad have settled on.

Opportunity for a Coordinated Push

Sri Lanka doesn't need to out-invest Swiss labs or Silicon Valley to benefit from this shift. It needs a coordinated push, most plausibly led by the NGJA in partnership with local research institutions and industry associations, to build a shared gemstone data infrastructure, pilot AI-assisted grading and provenance tools at a handful of trading centres, and put simple valuation and market-access tools directly in the hands of small miners and dealers. Do that, and the country's oldest export industry has a real shot at finally closing the gap between its reputation and its revenue.

Sources: International Gem Society on AI in diamond grading, Rapaport on Swiss International Gemlab's AI-aided lab launch, Rosk Gem News Report on SIG's AI model, Xinhua on Sri Lanka's 2025 gem export target, Everledger and Gübelin's Provenance Proof Blockchain, National Jeweler on Provenance Proof Blockchain reaching 500K tracked stones, Newswire on Gem City Ratnapura 2026.

(The author is Lecturer, Research Assistant, York St John University (London Campus) PhD (Reading), MSc in Data science(University of South Wales).  The views expressed are personal. He can be contacted at sugeeswara@gmail.com )