The Setup
PayPal put an AI SDR agent into production 14 weeks ago. Not a pilot. Full deployment across a 200-rep North America mid-market sales org. The target: 8,000 merchants monthly who stop processing payments mid-onboarding and fall through the cracks.
No sales team has the capacity to work 8,000 nurture cycles every month. So those leads sat untouched. Now an Agentforce agent runs a 10-nudge cadence on every single one. No fatigue, no cherry-picking high-commission accounts, no weekend gaps.
Meeting conversions are running 50% higher than human-only workflows.
Eitan Saban, Head of Sales for North America Mid Market at PayPal, shared the results at SaaStr AI 2026. The agent is not replacing reps. It is feeding them qualified meetings with full context already loaded. Reps start further down the funnel.
Why It Worked
PayPal feeds the agent conversation data from Gong, account data from Seismic, and structured CRM records. The agent learns why deals were lost and what drives lower conversion. It writes back to Salesforce with determinism, which matters in a regulated industry where "probably correct" is not good enough.
The data was not perfect before launch. Salesforce runs its own SDR agent at 70% deliverability, better than human reps, because the agent has omniscient understanding of curated data. But you do not need perfect data to start. Deploy the agent and it will show you what data structure you actually need. Action beats a six-month cleanup project.
The Headless Shift
Salesforce President Adam Alfano framed it clearly: the CRM is now tooling agents execute against, not a place humans log into. SaaStr's own CEO runs full campaigns through an AI agent with API access to Salesforce, no UI seat required. The agent pulls pipeline data, sponsors, historical revenue. He says "email this list" and it ships.
The model is commodity. The data model is the moat. If your CRM is still built for human data entry instead of agent inference, you are building the wrong stack.
What This Means
PayPal onboards over 100,000 merchants monthly. Trailing twelve-month revenue sits at $33.73 billion USD. This is not a scrappy startup testing agents on 50 leads. This is a Fortune 500 payments company running agents on thousands of leads in a regulated environment and posting conversion lifts in one quarter.
The playbook: identify leads humans will never touch, deploy agents on high-volume nurture, feed them conversation and account data, hand qualified meetings to reps. Rinse and repeat.
If you are waiting for perfect data or worried about replacing reps, you are solving the wrong problem. The win is not the agent closing deals. The win is the agent doing grunt work so reps show up to merchants as the best version of themselves.