Jacob Effron and Benedict Evans compare AI with earlier shifts in PCs, the internet, mobile networks, semiconductors and electricity. Benedict Evans treats those histories as sources of competitive patterns rather than precise forecasts, emphasizing that AI differs because its physical capability limits remain unclear.
Jacob Effron and Benedict Evans discuss why impressive model capability has not yet produced universal daily use. Coding has clear product-market fit, while many consumer and knowledge-work tasks remain difficult to isolate, describe and integrate into existing workflows, so better models alone may not solve the adoption problem.
Jacob Effron and Benedict Evans challenge simple predictions about job exposure. Work changes around tools, industries are affected through indirect business-model shifts, and models remain jagged enough to fail on short tasks while succeeding on much longer ones, making percentages assigned to occupations misleading.
Jacob Effron and Benedict Evans also examine model-lab economics and product strategy. Foundation models have high capital and marginal costs but weak network effects, product advantages can turn over quickly, and labs must experiment while searching for durable differentiation across models, applications and distribution.
Jacob Effron and Benedict Evans conclude that enterprise adoption requires organizational redesign, integration and professional services rather than simply giving every employee a chatbot. New consumer use cases likewise have to be invented around real behavior, and many early experiments will fail before enduring products emerge.
Watch on YouTube