Across industries, organisations face a structural risk: decades of expertise sit with specialists approaching retirement. When they leave, much of that knowledge leaves with them. What remains is documentation. Quite often it is accurate, but bereft of judgement, context and nuance.
An unexpected illustration of this challenge comes from the field of horticulture.
In Season Three, Episode One of Futurise, Rob Price interviews Steve Bustin, Chair of Board of Trustees for the Hardy Plant Society.
Across the UK, specialist plant societies have quietly preserved gardening knowledge for decades. The Hardy Plant Society, the largest of these groups, has around 5,000 members, dozens of local and special interest groups, and archives spanning more than 50 years. As it approaches its 70th anniversary, it faces a question familiar to every enterprise: how do you preserve experience-based expertise in a digital-first, rapidly changing world?
The Problem of Context
Consider the term “hardy plant.” On paper, it means a plant that survives winter. In practice, hardiness is a relationship between plant and place. A plant that thrives in Brighton may fail in Yorkshire. One that tolerates frost in free-draining soil may rot in heavy clay. The Gulf Stream enables parts of Scotland to grow species that struggle on the south coast. Hardiness depends on temperature, moisture, drainage, and exposure. It also increasingly depends on climate volatility.
Dahlias once had to be lifted each autumn; now in many regions they survive outdoors, with rot rather than frost as the primary risk. Snowdrops, which prefer cool woodland conditions, are being lost in gardens where hotter summers desiccate bulbs underground.
This is knowledge shaped by experience, observation and local context. Generic online advice, or even AI summaries, can provide broad guidance, but often miss the nuance that determines success.
Replace “hardy plant” with “regulated workflow”, “complex product line” or “regional compliance process”, and the analogy becomes clear. Context matters. Institutional knowledge is rarely generic.
Changing How Knowledge Is Transferred
Historically, gardeners learned through neighbours, allotments, talks and specialist journals. Today, younger enthusiasts turn to Instagram, YouTube and AI tools. They learn quickly, but differently.
A similar shift is happening in the world of business. Learning increasingly occurs on demand, through searchable systems and short-form content rather than apprenticeship models or in-person briefings.
The risk is not a lack of talent. It is the loss of tacit expertise.
In horticulture, many specialists are less comfortable with digital platforms. Their insights often remain in live talks or informal exchanges. In enterprises, experienced professionals often carry the unwritten logic behind processes. They make judgement calls, handle edge cases, and apply situational adjustments. These details rarely make it into policy manuals.
When they retire, documentation remains. Judgement often does not.
Where AI Can Add Real Value
AI can either flatten nuance or preserve it. The difference lies in how it is implemented. Used superficially, AI tools simply aggregate public information. Used strategically, they can help organisations capture and structure their own expertise.
Transcription tools can record discussions that would otherwise be lost. Interviews with senior experts can be converted into structured knowledge assets. Historical documents, case studies and internal publications can be indexed into governed repositories. Over time, this creates the foundation for organisation-specific AI systems. Systems that reflect institutional context rather than generic internet data.
For enterprises, this is not about replacing human expertise. It is about extending it. However, technology alone is insufficient. Responsible deployment requires:
- Clear governance frameworks
- Explicit consent from contributors
- Robust IP protection
- Defined access controls
- Ongoing curation and oversight
Without these safeguards, AI risks amplifying noise. With them, it can function as institutional memory infrastructure.
Extending — Not Replacing — the Human Layer
Despite digital change, the physical “garden fence” still matters. Gardeners learn by observing neighbours, visiting open gardens and asking questions in real time.
The objective is not to eliminate that human exchange, but to extend it.
For plant societies, that may mean hybrid events, recorded talks and searchable archives alongside traditional meetings. For enterprises, it means combining mentorship and human judgement with structured knowledge capture and explainable AI systems that support decision-making. The challenge facing a 70-year-old horticultural society is the same one facing regulated industries, engineering firms and complex global organisations: how to prevent decades of expertise from disappearing within a single retirement cycle.
AI, deployed responsibly and governed carefully, offers a way to preserve depth rather than dilute it. The question is not whether AI can summarise information. It is whether organisations will act in time to capture what truly differentiates them: context-rich expertise built over years of experience.
Like the best perennials, institutional knowledge can return stronger each year — but only if it is intentionally cultivated before it fades.
