Palo Alto Networks: AI coding spend to top $13 billion

Palo Alto Networks is framing AI coding agents not as productivity add-ons but as a new class of actor inside the enterprise. The company points to a surge in AI coding adoption, with spend on these tools projected to top $13 billion this calendar year and compound at more than 60% annually, and says the benefits for engineering teams are clear: faster output, higher productivity and lower operational overhead.

The problem, in its telling, is that these agents can assume a user's identity, modify file systems, run terminal commands and connect into core business systems. That breaks three assumptions traditional security was built on — known software, human users and human-speed actions. The body lists four resulting risk areas: shadow AI spreading before security teams know it exists; agents taking risky actions at machine speed; blurred identity because machines outnumber human identities 109 to 14 and agents often use human credentials; and uncontrolled connections that raise both cost and IP-leak risk. The cited numbers reinforce the point: 68% of organizations reported lacking visibility into developer AI tools in 2026, 57% of employees used such tools without formal IT approval, one documented intrusion chain involved an agent executing 17,600 actions in 4.5 days, and 93% of enterprises exceeded AI budgets with 20-30% of AI spend unaccounted for.

Palo Alto Networks' answer is a platform-centric approach: discover shadow AI, assess supply chain risk, protect identity and prompts, and govern data, actions and cost. The stakes hinge on whether enterprises can put that visibility and scoped identity in place before agents perform harmful actions — a race that, on the body's own evidence, many organizations are currently losing.