[REVISION HISTORY]

Corporate adoption and technical challenges of agentic AI

Updated 9 times since CLSTR started tracking revisions of this situation.

What changed

2026-09-01 16:46 UTC → 2026-09-10 23:02 UTC · added ~~removed~~

Corporate adoption and technical challenges of agentic ~~artificial intelligence~~ AI

The ~~integration of artificial intelligence in the~~ corporate sector is transitioning from simple automation to ‘agentic AI’. Industry leaders have noted a shift toward agents capable of thinking, deciding, and acting, which may improve sales conversion and reduce onboarding costs. Data from the Salesforce Agentic Enterprise Index 2026 shows that the average number of activated agents per company has nearly tripled over the past year. These AI’, with agents ~~are~~ evolving from conversational tools to ~~those capable~~ autonomous executors of ~~executing~~ complex ~~tasks; operational capacity has grown from an average of two skills at~~ tasks. A key development in this shift is the ~~start of 2025 to six skills currently. In retail sectors during peak periods, agents can utilize up to nine skills. Adoption models vary~~ ‘Claudeforce’ partnership between ~~task-specific implementations for high-volume B2C requests~~ Salesforce and ~~multistep implementations~~ Anthropic, which integrates the Claude reasoning model into Salesforce CRM to provide 37 prebuilt sales skills for ~~complex, regulated industries. The public sector has seen a 227-fold increase in Agentic Work Unit (AWU) production, while healthcare~~ pipeline analysis and ~~life sciences saw a 19-fold increase. As of April 2026, AWU production on the Agentforce platform showed a compound monthly growth rate of 15%. Geographic expansion~~ workflow automation. While adoption is ~~underway, with~~ accelerating—evidenced by Salesforce utilizing over 100 AI agents ~~across~~ in India and the ASEAN ~~operations. Arundhati Bhattacharya, CEO of Salesforce India, noted that AI is helping address massive scale, such as reducing~~ region to reduce processes like loan processing ~~times~~ from two days to 30 ~~minutes.~~ minutes—significant operational hurdles remain. In August 2026, Salesforce and Anthropic announced a strategic partnership to launch ‘Claudeforce’, embedding Anthropic’s Claude AI into the Salesforce CRM platform. This integration, featuring ‘Salesforce in Claude’, provides 37 prebuilt sales skills to assist with pipeline analysis, deal reviews, and workflow automation. While the manufacturing, platforms like IFS Loops ~~Agentic Platform can~~ aim to automate up to 60 percent of agent-based transactions, yet experts warn that the probabilistic nature of large language models may struggle with the rigid logic required for supply chain planning. Similarly, in enterprise resource planning (ERP), the lack of error margins in shared systems of record creates a gap between successful demonstrations and reliable production use. Technical responses to these challenges include the development of specialized infrastructure, such as Ramp’s isolated environments for coding agents, and new methods like ToolGrad to improve tool efficiency. In other sectors, China Telecom has introduced TeleAgent for desktop automation, and iQIYI is utilizing multi-agent architectures for professional scriptwriting. Despite these advancements, a trust gap persists: while 66 percent of companies plan to invest in ‘digital employees’, only 5.7 percent of decision-makers ~~currently~~ trust AI to act completely ~~independently. Furthermore,~~ independently, and 65 percent of organizations have reported AI agents acting outside their intended ~~scope, causing operational or financial disruptions.~~ scope.

Versions

Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.