How AI and digitisation are tipping the balance in the race to save plants and fungi

TL;DR

• Royal Botanic Gardens, Kew and ~400 collaborators show that AI plus mass digitisation is unlocking historic specimens for conservation, genomics and discovery.
• Practical wins include faster species ID, quantified shifts in flowering time (~2.5 days per decade), and genomes recovered from specimens nearly 200 years old.
• Big risks: major data gaps (only ~145M digital records exist, under 16% of global holdings), model bias, equity and legal issues, and the carbon footprint of compute.
• For executives: fund equitable digitisation partnerships, demand low-carbon compute, and adopt benefit‑sharing and governance standards (Nagoya Protocol alignment).

Why AI for conservation matters now

Digitisation (high-resolution imaging plus metadata capture) and machine learning are turning dried sheets in herbaria and fungaria (preserved plant and fungal specimen collections) into live datasets. Royal Botanic Gardens, Kew—working with roughly 400 scientists across 40 countries—has digitised 7.4 million specimens. That contribution helps push the global online total to about 145 million digital specimens, but those records likely represent less than 16% of specimens held in collections worldwide. The takeaway is stark: powerful tools exist, but most of nature’s archives remain unmapped.

That matters for business and policy because these archives are a time machine. They let researchers track phenology (the timing of seasonal events such as flowering), recover historic genomes (sequencing DNA from old specimens), and detect distribution changes tied to climate, land use or disease. For companies working in agriculture, biotech, or biosecurity, those insights can translate into new crop traits, novel enzymes and earlier detection of emerging pathogens.

What AI unlocks — concrete results

Computer vision, natural language processing to extract label data, and genomic pipelines together produce practical outcomes at speed and scale:

• Phenology at scale: An AI-driven analysis of millions of specimens found an average shift in flowering time of about 2.5 days per decade over the past century — a clear signal of climate-driven change that would be infeasible to measure manually.
• Faster, more accurate ID: In some tricky plant groups, trained models match or outperform human specialists for routine identifications, speeding surveys and flagging potentially vulnerable species faster.
• Genomic recovery: Improved sequencing means high-quality genomes can now be recovered from specimens collected up to ~180 years ago, making historic collections a genomic goldmine for drug discovery and breeding.
• Fungi as an untapped frontier: There may be ~2 million fungal species; ~90% are unknown to science and under 1% of known fungi have extinction-risk assessments. AI helps prioritize which specimens to sequence or survey first.

“Digitisation and new technologies make me increasingly hopeful that documenting and protecting life on Earth is achievable.”
— Prof Alexandre Antonelli, Executive Director of Science, RBG Kew (paraphrase)

A time machine for discovery

Specimens collected by past naturalists — including those from Darwin-era collections — are not curiosities. They contain DNA, annotation, and ecological context. AI links images, labels and geographic data across decades, revealing trends, previously overlooked species and genetic variants that are directly useful for crop resilience and pharmaceutical scouting.

Case study: Madagascar — digitisation that unlocked local value

When Kew helped digitise Madagascar’s ~37,000 specimens, the results went beyond a dataset. Local botanists gained searchable access to centuries of collections, accelerating species assessments and informing conservation priorities on the island. Digitised records enabled more accurate range maps and helped prioritize areas for field surveys and protected-area expansion — small but concrete wins that show how investment in digitisation yields policy-relevant science.

“Digitising Madagascar’s 37,000 specimens has unlocked centuries of knowledge and crucial insights for current biodiversity.”
— Landy Rajaovelona, Senior Botanist, Kew Madagascar (paraphrase)

Risks and trade-offs to manage

AI for conservation creates value, but also amplifies existing problems unless programs are designed intentionally.

Data gaps and bias

Most digitised specimens come from well-resourced institutions. Models trained on those data tend to underperform on species and regions that are underrepresented. That can skew priorities toward already-studied ecosystems and leave biodiversity-rich, lower-resourced countries behind.

Energy and infrastructure

Large AI models and genomic pipelines require substantial compute. Datacentre energy use and water use are real operational costs and reputational considerations. Conservation projects must weigh the carbon footprint of compute against the ecological benefits delivered, and adopt efficiency measures and renewable procurement where possible. Industry leaders and cloud providers now offer options such as carbon-aware scheduling, renewable energy contracts and regional low-carbon compute choices — use them.

“Fungi thrive in heat and humidity; as warm seasons lengthen, opportunistic fungal pathogens may spread into new regions.”
— Dr Esther Gaya, Senior Research Leader, RBG Kew (paraphrase)

Equity, legal frameworks and benefit-sharing

Historical imbalances persist. Many biodiversity-rich countries lack the resources to digitise, curate and host their collections. That’s not just inefficient — it’s an ethical and legal issue. The Nagoya Protocol under the Convention on Biological Diversity sets standards for access and benefit-sharing of genetic resources. Companies must align with those frameworks and negotiate material transfer agreements and benefit-sharing clauses to avoid biopiracy and reputational damage.

What business leaders should do now

AI for conservation is both a responsibility and a strategic opportunity. For executives considering investments or partnerships, prioritize actions that create scientific value while managing risk.

Action checklist for executives

• Fund equitable digitisation pilots: Partner with national herbaria and regional botanic gardens to digitise collections in biodiversity-rich, underrepresented countries. Outcome: new data sources and goodwill.
• Procure low-carbon compute: Require cloud providers to supply renewable-backed compute, carbon-intensity metrics, and model-efficiency consultation for AI workloads. Outcome: lower emissions and better ESG performance.
• Adopt benefit-sharing agreements: Align contracts with the Nagoya Protocol, include capacity-building, and ensure local researchers have access and co-authorship rights. Outcome: legal compliance and equitable partnerships.
• Invest in model auditing: Fund bias audits, diversify training data, and support regional digitisation to reduce performance gaps. Outcome: fairer and more generalizable models.
• Start small, scale responsibly: Run a one-country pilot (digitise X specimens, train a local team, evaluate outcomes), document KPIs, then replicate. Outcome: de-risked scaling and measurable impact.

How AI agents and AI Automation fit

AI agents and LLMs (such as tools used for literature triage and stakeholder coordination) accelerate administrative and research tasks: matching specimens to descriptions, extracting collection metadata, and drafting regulatory filings. Use them to automate repetitive tasks, but keep humans in the loop for governance, benefit-sharing negotiation and final scientific judgement.

KPIs and procurement choices to track impact

• Percent of national specimens digitised in target countries (goal: move from single-digit to >50% over X years).
• Number of local technicians trained and retained per project.
• Reduction in model error for underrepresented species/regions (bias metrics).
• Carbon intensity (kgCO2e) per major model training or genomic run; targets for year-over-year reduction.
• Number of benefit-sharing agreements executed and tangible capacity-building outputs delivered.

Practical business cases

• Agritech: Use wild genetic diversity uncovered by genomic recovery to develop drought- or disease-tolerant crop lines faster and at lower breeding cost.
• Biotech and pharma: Mine fungal genomes for novel enzymes or metabolites; partner with local institutions to secure access and co-develop IP under fair terms.
• Cloud and infrastructure: Offer tailored low-carbon compute packages and research credits for conservation projects, creating a new vertical for cloud sales aligned with ESG goals.
• Insurance and risk: Use phenological and distribution shifts identified by AI to refine climate-risk models for agriculture, forestry and supply chains.

Key takeaways & quick answers

• How much has Kew digitised?
  Kew has digitised 7.4 million specimens, contributing to about 145 million global digital records — still likely under 16% of total specimens held worldwide.
• Can AI accelerate conservation science?
  Yes. Machine learning speeds species identification, quantifies climate-driven phenological shifts (~2.5 days per decade), and enables genomic recovery from specimens up to ~180 years old.
• What are the main risks?
  Major risks include biased training data, environmental footprint of compute, and inequitable benefit-sharing that could marginalize biodiversity-rich countries.
• What should business leaders do?
  Invest in equitable digitisation partnerships, procure low-carbon compute, adopt Nagoya-aligned benefit-sharing, and fund bias audits and local capacity building.

Digitised collections and AI form a pragmatic playbook: map the gaps, invest where they matter most, and structure partnerships so benefits flow back to source communities. The technology — computer vision for ecology, genomic recovery pipelines and AI Automation — is real. So are the trade-offs. Thoughtful procurement, governance and an iterative pilot-to-scale approach give companies a way to turn conservation action into measurable scientific, commercial and reputational value while helping prevent the loss of vital plant and fungal diversity.

Executives ready to act can start by commissioning a small pilot with a botanical institution, asking cloud partners for low-carbon bids, and committing to Nagoya-aligned benefit-sharing as a non-negotiable term. That’s how to turn hope into impact.