South Asia is home to more than 45 percent of the world’s estimated 752 million poor livestock keepers, a large number of which live in Pakistan (Robinson et al., 2011). Livestock serves multiple functions in the mixed-farming systems prevalent in South Asia (Thornton et al., 2002) and contribute significantly towards the livelihoods of the rural poor (LID, 1999). For example livestock holdings increase farm productivity through draft power and manure, provide additional income through the sale of milk and beef, and serve as an important avenue for savings in the absence of access to formal banking sector. The aforementioned attributes make livestock an essential feature of rural agricultural households in South Asia. Pakistan is the 6th most populous country in the world and is categorized as a low income agrarian economy (World Bank 2012). Agriculture is the 2nd largest sector of the economy, accounting for over 21% of GDP and providing employment to 45% of the total labor force (Pakistan economic Survey 2010). Livestock is the largest component of the agriculture sector and contributed 11.6% to the national GDP during 2010-12. Even though Pakistan is the 4th largest milk producer in the world (FAOSTAT 2010), dairy farming in Pakistan is based on traditional, labor intensive, small scale farming methods. In 2010, more than 70% of Pakistani dairy farmers had an average holding of 2-3 dairy animals and less than 2 hectares of land (Burki, Khan& Bari 2012). Given that 47% of all rural households in Pakistan owned livestock and livestock contributed 11% towards their total income (FAO, 2009), interventions in the livestock sector hold the key to rural poverty alleviation. The milk yields of dairy breeds in the developed world are 5 to 6 times that of local Pakistani dairy breeds. The large differences in milk yields are largely attributable to the systematic breed improvement strategies employed in the developed world. However, the poor genetic potential of Pakistani dairy animals can be significantly upgraded through the artificial insemination (AI) technology. But despite rate of returns in access of 100%, adoption rates have remained very low and according to Pakistan Livestock Census only 11% of cattle were artificially inseminated in 2006. Despite the potential gains, the track record of sustained poverty reduction through livestock sector development interventions is weak (LID, 1999). Moreover, the existing development economics literature has focused primarily on adoption of different crop technologies involving the use of fertilizer, pesticide and high-yielding seed varieties whilst livestock technologies have been largely ignored. This paper aims to addresses these gaps in the literature by examining the factors driving low rates of AI technology adoption in Pakistan. To this end, we develop a parsimonious, stochastic, structural model that captures the essential features of the economic environment of rural agricultural households. The structural model is calibrated using survey data from Pakistan, solved numerically and simulated to compute the welfare effects of different policy interventions. Multiple roles played by livestock limits the efficacy of static models and reduced form estimation techniques. Structural models have gained popularity in the development economics literature over the past decade (Suri 2011, Kochar 1999; Munshi 2004; Demekw & Meshchi, 2003; Foster and Rosenzweig 1996a; Foster and Rosenzweig 1996b and Rosenzweig & Wolpin, 1993) as structural models allow economists to explicitly model implicit costs that are known to significantly reduce returns from technology adoption. For example costs associated with credit constraints, incomplete insurance markets, risk and learning are often ignored in reduced form regressions. Therefore, we make a concerted effort to capture the essential economic features of rural agricultural households e.g. credit constraints, herd dynamics, production risk and mixed farming systems in parsimonious manner in our paper. The literature on risk and credit constraints is the key to understanding the AI technology adoption problem. Even though AI technology promises significantly higher incomes in the future, credit constraints may prevent small subsistence farmers from borrowing against higher future incomes. Likewise, the absence of insurance markets and low levels of wealth magnify the disutility of risks associated with new agricultural technologies like AI, which entail upfront outlays and uncertainty regarding future returns. Consequently, dairy farmers can potentially smooth consumption through the channel of income smoothing and/or asset smoothing. However, given the multiple roles of livestock in rural agricultural households, different consumption smoothing channels have different impacts on the technology adoption decision. We aim to compute the relative bearing of income smoothing and asset smoothing on the technology adoption decision. This knowledge can assist policy makers design interventions that achieve greater impact by increasing the overall rates of technology adoption and thus, potentially transform the lives of poor dairy farmers in Pakistan. Our preliminary analysis suggests that the in presence of extreme poverty and in the absence of insurance markets low rates of AI technology adoption are a result of the consumption smoothing motive of poor livestock farmers, realized directly through the channel of asset smoothing, as opposed to income smoothing. Therefore, supply side interventions like subsidized rates for AI technology and access to credit are less likely to increase adoption rates compared to policy interventions that reduce the risk exposures of households. In particular, technologies that reduce the relatively higher mortality risk of cross-bred cattle compared to local cattle have a large positive impact on AI technology adoption rates.