Key Takeaways
- Caitlyn was developed specifically for research-intensive sectors where accuracy and traceability are critical.
- Agricultural organisations hold extensive research assets that are often difficult to access or apply in practice.
- The platform operates within a customer’s own cloud environment to address data sovereignty and governance.
- Early agricultural deployments show increased engagement with research and faster access to verified answers.
- Research institutions are increasingly viewing AI as long-term infrastructure rather than experimentation.
The Gap Between Research Production & Practical Use
Agriculture is among the most research-intensive sectors globally, supported by decades of trials, reports, and technical documentation. However, much of this knowledge remains under-utilised, often stored in formats that are difficult to search, interpret, or apply in time-sensitive decision-making contexts.
Josh Smith, CRO of Caitlyn, described this challenge as the starting point for the company.
“Huge investment in research, but very little of it reaching the people who need it in a usable form,” Smith said.
He noted that in agriculture, incorrect or poorly contextualised information can have direct financial, environmental, or compliance consequences, making accuracy and confidence essential.
Why Caitlyn Was Designed Specifically For Agriculture
Smith explained that the platform was not adapted from a general-purpose AI model, but built to handle the specific characteristics of agricultural research. These include long technical PDFs, scanned documents, tables, images, videos, and region-specific terminology that can vary significantly by crop or production system.
“We didn’t adapt a generic AI tool — we built a system specifically for turning complex research into practical, trusted answers,” he said.
The platform processes content by preserving document structure, extracting metadata, and mapping relationships between concepts. This approach is intended to retain context rather than relying on keyword similarity alone.
Trust, Context, & Data Sovereignty
Data governance remains a major concern for agricultural research organisations, particularly those managing shared research, sensitive datasets, or Indigenous knowledge. Smith said the platform was designed to operate entirely within a customer’s own cloud environment, ensuring that data does not leave their control.
“These values are foundational, not add-ons,” Smith said, referring to trust, context, and data sovereignty.
Organisations control what content is included, who can access it, and how it is used. Role-based access controls and audit logs are used to support accountability and compliance.
Verified Answers & Citation-Based Responses
A core design principle is that both user inputs and AI outputs are treated as untrusted by default. According to Smith, every response must be validated against source material and supported by citations.
“If Caitlyn can’t answer confidently, it won’t guess,” he said.
This approach is intended to reduce the risk of hallucinated or unreliable answers, which remains a key concern for AI adoption in regulated and research-driven environments.
Deployment Outcomes in Agricultural Organisations
Smith pointed to agricultural industry bodies that manage large volumes of science-backed research across written reports, videos, and audio content. In these deployments, farmers and advisors are able to ask questions in plain language and receive answers linked directly to verified research sources.
Reported outcomes include increased engagement with research content, reduced time spent searching for information, and clearer insight into the types of questions being asked by end users. Smith noted that this feedback can inform future research priorities and extension strategies.
Caitlyn as Infrastructure For Agricultural Research Systems
Smith described a broader shift in how agricultural institutions view generative AI. While interest has grown, adoption has been cautious, driven by concerns around data quality, governance, and public trust.
“The question has shifted from ‘can we do this?’ to ‘how do we do this safely and responsibly?’” he said.
Looking ahead, Smith suggested that research platforms could shorten the gap between discovery and impact by making knowledge searchable, measurable, and directly connected to decision-making. Beyond knowledge retrieval, he noted growing interest in decision support, compliance workflows, and integration with modelling tools, while emphasising that these depend on a reliable and trusted knowledge layer.
