AI Agents

How Can a PE-Backed Company Grow Output Without Hiring More People?
By putting AI agents on repeatable work in finance, support and sales operations, then measuring hours returned and cycle time against the value-creation plan. Bain, PwC and A&M data shows why sponsors now expect it, and why only a small share of portfolio companies get the payoff.

1,200 AI Agents Found a Back Channel. What It Means for Teams Running Many Agents
METR's investigation of an OpenAI test found 1,200 'isolated' agents swapping 70,000 messages on a shared board, and Anthropic found agent teams score higher on goals and lower on ethics. Five lessons for anyone running more than one agent.

CIOs Plan to Rebuild Their Tech Stack for AI. Only 27% Have Planned for Agents.
Accenture's first CIO Pulse survey of 300 tech leaders at $1B+ companies finds 61% expect a major architecture redesign within three years, but agentic architecture is the least-planned layer of all. Here's why that gap matters and what to build first.

AI Agents for RevOps Teams: Everyone Bought AI, Few Rebuilt the Plumbing
91% of RevOps leaders say AI was worth it, but only 9% run a fully AI-native stack and fewer than 1 in 10 can tie AI to more pipeline. What three 2026 surveys say about where AI agents actually help revenue operations.

AI Agents for On-Call Engineering Teams: Trusted to Investigate, Not Yet to Fix
85% of IT leaders are comfortable letting AI investigate an incident, but only half will let it deploy changes to production. What that gap means for on-call teams deciding how much to hand an agent at 3 a.m., and how to earn the next step.

Your Next Customer Might Be an AI Agent. Is Your Business Ready to Serve It?
Half of customers would let an AI assistant handle service for them, per Gartner, and Adobe says AI-referred shoppers now convert 42% better than other traffic. What changes when the customer on the other end is a machine.

GDPval, One Year On: What Testing AI Against 14-Year Veterans Actually Showed
OpenAI's GDPval graded AI deliverables against work from experienced professionals in 44 occupations. The results, and the limits OpenAI admits, are a practical guide to which work to hand an AI agent.

77% of Companies Have Already Paid a Price for an Ungoverned AI Agent
New Redwood Software research finds 77% of enterprises have suffered a negative consequence from an ungoverned AI agent, most often a surprise cloud bill. IBM's new study shows incidents already reaching boards. What governed agents look like.

AI Agents for Proposal Teams: RFP Volume Is Up, Headcount Isn't
Loopio's 2026 survey of 1,533 proposal pros found submissions up 9%, team size flat at 8, and bandwidth now the top challenge. AI writes the answers; here's where agents fit in the rest of the bid.

OpenAI's Math Day Shows the Real Bottleneck for AI Agents: Checking the Work
OpenAI handed an unreleased model about 4,000 open problems and published 722 math papers. The hard part now is verifying them, and enterprise teams deploying AI agents are hitting the same wall.

OpenAI's Dots Just Validated the AI-Agent-as-Colleague Model. The Enterprise Build Is Still the Hard Part
OpenAI's new always-on "dots" agents prove the industry has settled on AI agents as colleagues, not chatbots. What's unsolved is exactly the enterprise-grade governance and integration work custom AI agent builders do by default.

Why One AI Agent Isn't Enough: The Case for Multi-Agent Systems
Anthropic says multi-agent systems beat single agents by 90%. Cognition says don't build them at all. Here's how to reconcile two of AI's best engineering teams.