AI Diffusion Among Punjab Farmers in Pakistan: Associations with Perceived Farm Productivity, Profitability

The researchers collected questionnaire responses in person and through local agricultural networks, including farm owners, managers and supervisors.

For a farmer deciding when to irrigate or how to protect a crop from pests, the appeal of artificial intelligence comes down to practical questions: can it reduce waste, help avoid a poor harvest and leave more money after the bills are paid? In Pakistan's Punjab province, where farms produce much of the country's wheat, rice, cotton and sugarcane, those questions carry weight as growers face rising costs, uncertain weather and pressure on water supplies.

A study titled "Improving farm productivity and profitability through AI diffusion: evidence from Punjab, Pakistan," published in Frontiers in Artificial Intelligence, explores how farmers view these technologies and how those views relate to farm performance.

Aeman Khalid and colleagues surveyed 101 respondents and found that greater reported AI diffusion went together with stronger perceptions of productivity and profitability. Enthusiasm exceeded reported use, revealing a gap between seeing value in AI and putting it to work.

What 101 Farmers Revealed About AI on Their Farms

The researchers collected questionnaire responses in person and through local agricultural networks, including farm owners, managers and supervisors. Medium-sized farms of 12–25 acres made up 55.4% of the sample, with smaller and larger operations also represented. Wheat was the most common crop, followed by rice, cotton and sugarcane.

Participants rated statements on a five-point scale, covering their awareness and use of AI, its integration into farming, and their views on productivity and financial benefits. The study's measure of "AI diffusion" combined several dimensions, including awareness, reported use and future intentions, so it should not be read as a simple count of farms using advanced equipment.

The technologies discussed include tools for monitoring crops, predicting yields or pest problems, and managing irrigation. Sensors, drones and smart irrigation appeared in the questionnaire as examples of tools farmers might use.

About 53.5% agreed that their farms currently used such AI tools, compared with 71.3% who agreed that AI improves crop-management decisions. Around 63.3% planned to expand AI use, suggesting an appetite for further adoption among respondents.

The strongest enthusiasm concerned specific farming tasks: 75.3% believed AI could improve monitoring of farm conditions, and 81.2% believed it could help identify profitable crop-management strategies.

Better Farming and Better Earnings Were Closely Linked

Respondents with higher AI diffusion scores tended to give more positive assessments of productivity and profitability. The correlations were 0.617 for productivity and 0.607 for profitability, indicating reasonably strong positive relationships within this sample.

The closest relationship was between perceived productivity and perceived profitability, with a correlation of 0.749. People who rated the production benefits more highly also tended to rate the financial benefits more highly.

Separate regression models linked AI diffusion scores with 38.1% of the variation in perceived productivity and 36.8% of the variation in perceived profitability. The productivity model accounted for 56.1% of the variation in profitability perceptions. These percentages describe differences in survey scores; they do not mean that AI raised yields by 38.1% or profits by 36.8%.

Examining earnings alongside production gives the research practical relevance. A farm can grow more without keeping more money if equipment, maintenance and other expenses absorb the extra revenue. Better timing of irrigation, more careful fertilizer use and earlier detection of crop problems could improve the financial picture by reducing waste, although this survey did not measure those savings.

The Distance Between Interest and Everyday Use

The authors interpret the gap between positive views and reported use through three research frameworks: Diffusion of Innovations, the Resource-Based View and the Technology–Organization–Environment framework.

In everyday terms, these perspectives ask whether farmers find a technology useful and manageable, whether they have the money and skills to use it, and whether their surroundings provide the necessary infrastructure and support. A promising tool needs to fit the work of the farm and the resources available to the people running it.

High purchase costs, unreliable electricity, weak internet connections, limited technical knowledge and gaps in agricultural advisory services feature prominently in the paper's discussion. These explanations come from earlier research; the survey did not directly test which barriers affected its respondents.

That distinction matters because the findings cannot establish why any particular farmer has held back. The wider literature raises a concern that farms with more money and expertise may be better placed to benefit, leaving smaller operators with fewer opportunities to participate.

What It Would Take to Make AI Useful and Accessible

The paper recommends investment in rural broadband and reliable electricity, financial support such as subsidies or affordable credit, and practical training through agricultural advisory services. Collaboration between government, researchers and agricultural technology providers could help produce affordable tools suited to local crops, water pressures and farm sizes.

Shared equipment, cooperative advisory services and AI available as a service could spread costs across several users. The authors suggest gradual adoption of tools with visible benefits in irrigation, pest management or input use, giving farmers time to learn and adjust their routines. They acknowledge that performance can temporarily dip during that learning period.

This was a one-time survey of a convenience sample, rather than a randomly selected group representing Punjab's farming population. District and village details were not systematically recorded, restricting geographical comparisons.

Productivity and profitability were assessed through perceptions, including statements about what AI could achieve, rather than verified harvest records or financial accounts. Optimism may influence those answers, and the analysis cannot establish whether AI caused better performance or whether better-resourced farms were more likely to adopt it. The simple models also leave other possible influences unresolved.