Agentic AI – systems that can autonomously perform tasks with contextual awareness – is gaining traction across the agrifood value chain. According to Dave Farnham, VP of AI and engineering at ClimateAi, the momentum is being driven by a convergence of factors: more powerful algorithms, increased computing capacity, and the maturation of AI agent frameworks.
“Companies are looking to move from descriptive dashboards to systems that can synthesize data and suggest sets of concrete actions,” Farnham told AgTechNavigator. “The number of processes to track has increased and the climate is variable enough that old-school decision-making isn’t cutting it. Agrifood businesses are increasingly looking for tools that translate large amounts of data into suggested concrete actions.”
What makes AI ‘agentic’?
In agriculture, agentic systems are not about full autonomy, but intelligent automation and augmentation. Farnham defines agentic technology as “automated in an intelligent and flexible manner,” capable of handling routine but context-sensitive tasks.
At ClimateAi, agentic systems are used to:
- Automate rote tasks that consume time but add little strategic value.
- Fuse disparate data sources to eliminate spreadsheet chaos.
- Surface key insights through natural language interfaces, making data accessible without coding expertise.
Where the opportunity lies
Interest in agentic AI spans the entire agrifood value chain, but Farnham notes that readiness depends on data maturity. Larger, digitally advanced players are leading adoption, while others lag due to fragmented or poor-quality data.
There’s no clear upstream or downstream trend, he said. It’s really about how ready an organization is to contextualize AI with its own operational data. “Engagement seems to depend most on data readiness and organizational culture.”
Inside the architecture: Multi-agent systems at work
ClimateAi’s agentic systems are built on large language models (LLMs) that act as orchestrators rather than analysts. These agents don’t generate new calculations but instead trigger workflows across ClimateAi’s proprietary modelling infrastructure.
“We’ve developed a multi-agent system with router and worker agents,” Farnham explained. These decide which tools to use, pass the right arguments, and return results – often summarizing them in plain language.
Trust, transparency, and the human-in-the-loop
One of the biggest hurdles to adoption is trust. Farnham emphasized that ClimateAi’s systems are designed for constraint and transparency. Every recommendation is traceable to its data source and method.
“Humans remain firmly in control,” he said. “Agents support by synthesizing data, stress-testing assumptions, and surfacing overlooked risks. It’s a co-pilot model, not autopilot.”
Climate data as a cornerstone
Despite the AI advancements, climate and weather modelling remain foundational. What’s changed is how these insights are delivered.
“Agents help surface climate insights contextually and on demand,” Farnham said. “They translate forecasts into business actions – like adjusting planting schedules or sourcing decisions – with less friction.”
Agents can automatically pull together data from multiple systems, like weather data, delivery schedules and production forecasts, and give custom dashboards or scenario views tailored to specific questions such as: which of my deliveries is likely to be impacted by recent or upcoming dry conditions?
“This represents a shift from manual data wrangling to intelligent orchestration,” Farnham believes, and can save “huge amounts of time and enable faster, more consistent, and confident decision-making”.
Barriers to scale: data and culture
Two major barriers stand in the way of widespread adoption, Farnham believes. The first is data quality and accessibility. “AI agents can only perform well if they have proper context in the form of data,” he said. “Many organizations still lack reliable, structured, historical data that captures past performance and decision-making.”
Next is cultural resistance. Decision makers may fear being blamed if AI-informed choices go wrong. “Overcoming this requires transparency, education, and clear human-in-the-loop governance,” Farnham noted.
What’s next for agentic AI in ag?
Looking ahead, Farnham sees agentic AI becoming a standard layer in digital farming platforms, offering a conversational interface between people, models, and data.
“I expect this technology to stick around and steadily integrate into existing systems,” he said. “I am doubtful that it will take over entirely, but it will become a standard layer in digital farming and supply-chain platforms: a flexible, sometimes conversational interface between people, models, and data.
“Some keys that will determine how deeply this technology takes root in digital farming include: how well it can support the way people really make decisions, and whether strong data governance can be maintained across all actors as innovation accelerates.”