Teams split on agent autonomy and readiness

Companies are splitting on how far to let artificial intelligence agents act on their own, even as adoption data shows most workplaces are still stuck in pilot mode. (mckinsey.com) An agentic system does more than answer a prompt: it plans steps, uses tools, updates records, and completes tasks across software without waiting for a human after every click. McKinsey reported on January 28, 2025 that nearly all companies are investing in artificial intelligence, but only 1 percent of leaders said their organizations were “mature” in deployment. (mckinsey.com) That gap is showing up inside human resources. HR Executive reported on January 2, 2026 that Phenom’s benchmark study of nearly 500 organizations found 83 percent in the lowest two maturity categories, less than 1 percent at high intelligence maturity, and only 5 percent at high automation. (hrexecutive.com) The same report showed heavy experimentation but uneven rollout by function and industry. Healthcare organizations reached 90 percent adoption of automated candidate campaigns, financial services hit 68 percent use of artificial intelligence for candidate matching, and 88 percent of retail organizations still lacked advanced automated screening. (hrexecutive.com) A second line of research points to the same bottleneck from a different angle. Harvard Business Review reported on February 17, 2026 that 88 percent of companies said they use artificial intelligence regularly, but adoption often stalls because employees fear loss of relevance, identity, and job security. (hbr.org) McKinsey framed the obstacle higher up the org chart. Its 2025 workplace survey said employees were more ready for artificial intelligence than leaders assumed, and called leadership the biggest barrier to scaling the technology into real workflows. (mckinsey.com) The debate gets sharper when teams move from copilots to agents that can take action. In customer support, Alhena wrote on April 13, 2026 that many ecommerce bots stall at about 50 percent automation because teams launch them and stop coaching them. (alhena.ai) Alhena said those systems need weekly review, scoring, tuning, and threshold changes after launch. The company pointed to Klarna’s assistant handling 2.3 million conversations in its first month, cutting resolution time from 11 minutes to 2, then losing quality within a year badly enough that Klarna rehired human agents. (alhena.ai) The split in practice is now clear: one camp wants agents to own full workflows, while another says most enterprises still do not have the trust, governance, or operating habits to hand over the keys. The data so far shows broad interest, low maturity, and a lot of work after deployment. (mckinsey.com; hbr.org; hrexecutive.com; alhena.ai)