SAP Booked 40,000 Meetings and Cut Manual Outreach 80% — Outreach's Benchmark for What Agentic AI Actually Delivers

SAP booked more than 40,000 customer meetings and created 10,000 opportunities using Outreach, reaching $1.5 billion in total pipeline value while cutting manual outreach 80% in its North American digital hub. Those are the kind of named-customer numbers Outreach’s CEO uses, framed around a conversation with SAP’s Utkarsh Bahadur at the company’s Unleash 2026 conference, to draw a line between “copilot” AI that assists a rep and agentic AI that plans and executes on its own. Amplitude consolidated five point tools into one, saving $600K while increasing average deal size 50%. Siemens hit up to 98% forecast accuracy across 190 standardized countries.

The underlying argument: sellers spend nearly 70% of their time on non-selling admin work — updating CRM records, drafting follow-ups, flagging at-risk deals — which is exactly the work agentic AI is built to absorb autonomously rather than assist with. Bahadur’s framing is precise about what that buys: “doing more with less,” not solving a demand problem AI can’t actually touch.

The same split shows up outside sales, in Canada Goose’s Agentforce deployment for customer service: the agent resolved 89% of high-volume, low-complexity contact-center inquiries — order status, warranty questions — autonomously, cutting peak-season wait times 13.9% and lifting satisfaction 21.2%, while human “Style Experts” were freed to handle the product-advisory conversations that actually need judgment. Canada Goose validated the deployment on digital channels first, then extended to voice in December, on the logic that errors in chat are more recoverable than errors on a phone call.

For GTM leaders sizing an agentic AI investment, SAP’s $1.5B pipeline and Canada Goose’s 89% autonomous-resolution rate are the same data point from two different functions: agentic AI’s real ROI shows up in what it takes off a human’s plate, not in what it says to a prospect.