AI agents are coming for every day tasks: What Goldman Sachs sees for the next AI growth wave
Another crucial factor is the declining cost of AI inference. Goldman Sachs expects cheaper token processing to make increasingly complex AI applications economically viable.
- Sep 20, 2026,
- Updated Sep 20, 2026 1:00 AM IST
Artificial intelligence could be entering a new phase as consumer-facing AI agents move beyond answering questions to taking actions on users’ behalf, potentially reshaping shopping, search, travel, advertising and other everyday digital activities, according to Goldman Sachs Research.
The shift marks a move from AI experimentation towards wider commercial adoption, with companies increasingly integrating AI into both internal operations and consumer products.
From chatbots to AI agents
The next generation of consumer AI could be significantly more action-oriented. Instead of simply recommending a hotel or flight, an AI agent could potentially search, compare and book it. It could also manage calendars, appointments and purchases.
Goldman Sachs Research said this represents a shift from a “conversational relationship” with AI to an “action-oriented relationship”.
But greater autonomy comes with a major requirement: trust. To perform tasks independently, agents may need access to sensitive information such as calendars, passwords and payment cards. Security and consumer confidence could therefore determine how quickly agentic AI becomes mainstream.
If consumers accept this model, AI agents could effectively become a new utility layer between users and the internet.
Shopping and advertising face disruption
AI agents could fundamentally change online shopping. Instead of consumers visiting multiple websites, comparing products and completing transactions themselves, an agent could potentially manage much of that process.
This could create a new AI platform layer between underlying technology infrastructure and consumer applications.
Goldman Sachs expects some mass-market AI agents to follow existing internet business models, including free services supported by advertising, alongside paid subscriptions for more advanced or ad-free experiences.
Advertising itself could also become more automated. AI could improve ad creation, targeting, placement and performance measurement, potentially increasing efficiency and reducing costs.
However, companies are simultaneously spending heavily on AI infrastructure, meaning those efficiency gains could initially be absorbed by investments in computing capacity and data centres.
$1.4 trillion AI infrastructure push
The consumer-agent boom is being accompanied by a huge infrastructure build-out.
Goldman Sachs Research expects US hyperscalers to maintain elevated capital spending through 2027 and estimates they could deploy $1.4 trillion in capital in 2027.
The expansion could nevertheless face constraints involving memory chips, electricity, land and other parts of the supply chain.
The bank also expects semiconductor shortages to persist in the near term. Expanding chip manufacturing capacity takes time, with a new fabrication facility potentially requiring about three years to build.
Falling AI costs could drive usage
Another crucial factor is the declining cost of AI inference. Goldman Sachs expects cheaper token processing to make increasingly complex AI applications economically viable. Its earlier research estimated that token consumption could rise 24-fold between 2026 and 2030, reaching around 120 quadrillion tokens a month as consumer and enterprise adoption of agentic AI expands.
The combination of falling inference costs and sharply rising usage could accelerate AI adoption and improve the economics of running large-scale AI services.
Safety remains a major question
The expansion is taking place alongside growing debate over the pace and safety of frontier AI development.
Executives at Goldman Sachs’ technology conference discussed whether the development and deployment of increasingly capable AI systems should slow. Goldman Sachs Research said the eventual impact would depend heavily on how regulation is designed and enforced.
For now, the bank does not expect those debates to materially alter its outlook for the AI infrastructure investment cycle through 2027.
For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine
Artificial intelligence could be entering a new phase as consumer-facing AI agents move beyond answering questions to taking actions on users’ behalf, potentially reshaping shopping, search, travel, advertising and other everyday digital activities, according to Goldman Sachs Research.
The shift marks a move from AI experimentation towards wider commercial adoption, with companies increasingly integrating AI into both internal operations and consumer products.
From chatbots to AI agents
The next generation of consumer AI could be significantly more action-oriented. Instead of simply recommending a hotel or flight, an AI agent could potentially search, compare and book it. It could also manage calendars, appointments and purchases.
Goldman Sachs Research said this represents a shift from a “conversational relationship” with AI to an “action-oriented relationship”.
But greater autonomy comes with a major requirement: trust. To perform tasks independently, agents may need access to sensitive information such as calendars, passwords and payment cards. Security and consumer confidence could therefore determine how quickly agentic AI becomes mainstream.
If consumers accept this model, AI agents could effectively become a new utility layer between users and the internet.
Shopping and advertising face disruption
AI agents could fundamentally change online shopping. Instead of consumers visiting multiple websites, comparing products and completing transactions themselves, an agent could potentially manage much of that process.
This could create a new AI platform layer between underlying technology infrastructure and consumer applications.
Goldman Sachs expects some mass-market AI agents to follow existing internet business models, including free services supported by advertising, alongside paid subscriptions for more advanced or ad-free experiences.
Advertising itself could also become more automated. AI could improve ad creation, targeting, placement and performance measurement, potentially increasing efficiency and reducing costs.
However, companies are simultaneously spending heavily on AI infrastructure, meaning those efficiency gains could initially be absorbed by investments in computing capacity and data centres.
$1.4 trillion AI infrastructure push
The consumer-agent boom is being accompanied by a huge infrastructure build-out.
Goldman Sachs Research expects US hyperscalers to maintain elevated capital spending through 2027 and estimates they could deploy $1.4 trillion in capital in 2027.
The expansion could nevertheless face constraints involving memory chips, electricity, land and other parts of the supply chain.
The bank also expects semiconductor shortages to persist in the near term. Expanding chip manufacturing capacity takes time, with a new fabrication facility potentially requiring about three years to build.
Falling AI costs could drive usage
Another crucial factor is the declining cost of AI inference. Goldman Sachs expects cheaper token processing to make increasingly complex AI applications economically viable. Its earlier research estimated that token consumption could rise 24-fold between 2026 and 2030, reaching around 120 quadrillion tokens a month as consumer and enterprise adoption of agentic AI expands.
The combination of falling inference costs and sharply rising usage could accelerate AI adoption and improve the economics of running large-scale AI services.
Safety remains a major question
The expansion is taking place alongside growing debate over the pace and safety of frontier AI development.
Executives at Goldman Sachs’ technology conference discussed whether the development and deployment of increasingly capable AI systems should slow. Goldman Sachs Research said the eventual impact would depend heavily on how regulation is designed and enforced.
For now, the bank does not expect those debates to materially alter its outlook for the AI infrastructure investment cycle through 2027.
For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine