Piper engages and qualifies inbound leads today. Hunter works a pipeline over weeks and months, in pilot until November. Neither is switched on by a rep.
For two years the honest answer to whether software could do the SDR job was that it could draft the email but could not judge whether a lead was real. Piper is Salesforce's claim that it now does both, on your website and in your inbox, today — and Hunter extends the claim to outbound work that runs for months rather than minutes. What that leaves you is narrower than your current week and worth more than most of it. The agent qualifies a lead against what the buyer typed into a form; you are the only one who finds out what they did not type. If your company switches Piper on, the meetings landing on your calendar will already have been screened against stated criteria, and the unstated ones become the whole of your edge. That shows up in the first ten minutes of the call or it does not show up at all.
"Inbound pipe gen" is Salesforce's phrase for the top of the funnel you did not go out and get — someone fills in a form, opens a chat, replies to a campaign — where somebody has to decide whether it is worth a meeting and then write back. An agent here is software that makes that decision and sends that reply, not software that suggests what you might say.
"Long-horizon" is the genuinely new part. Most AI tools answer and forget. Salesforce says Hunter runs on a new long-horizon runtime built to keep working one pipeline "over weeks and months," which means holding on to what it already sent and what came back. That is also why Hunter is the one still in pilot.
Two pieces of vocabulary decide what you can act on. "GA" means generally available: it can be bought and switched on. "Pilot" means Salesforce picks who gets it. And an Agentic Work Unit is Salesforce's own counter for agent activity, so seven billion of them describes how much the agents did, not how much of it was right.
The short version: Salesforce now sells a named agent that works your inbound leads, and has dated the one that works your outbound list to November.
Salesforce published "Salesforce Expands Agentforce With a New Portfolio of AI Agents Built for High-Value Work" on 11 September 2026, naming seven agents built for named jobs rather than for a platform. Two of them are sales roles. Salesforce describes Piper as "your inbound pipe gen agent," which "works across websites and inboxes to engage, qualify, and convert inbound leads into pipeline for B2B sales and marketing teams." Hunter is "your outbound sales agent" and "works a sales pipeline from research to outreach, collaborating with sellers over weeks and months."
Announced: 11 September 2026.
Availability: Piper, Casey, Paige, Carter, Marshall and Fin are "GA now." Hunter is "Pilot now; GA November '26" and is the first agent to run on Salesforce's new long-horizon runtime. AI Skills in Agentforce Coworker is "Pilot now; GA October '26," Agent Optimizer is "GA October '26," and Multi-Agent Orchestration is "GA now."
Price: Not stated.
Regions: Not stated.
The customer numbers are worth reading slowly. Salesforce says "60% of Perk's sales pipeline" is built by Hunter, that "4x the conversation volume" is now driven by Asana's website agent Piper, and that "Customers deploy Piper in 45 days on average." Across the portfolio it reports "7 billion" Agentic Work Units "across Agentforce and Slack, including 3.2 billion in Q2 alone."
What it doesn't do: The release states no price, no editions and no regions, so it does not tell you whether Piper is included in the Salesforce your company already pays for. It does not say what Piper does when a lead does not qualify, who it escalates to, or what a seller sees before a lead reaches them — the word "convert" stands in for all of that. Hunter's November date carries Salesforce's own disclaimer: "Customers should make their purchase decisions based on features that are currently available." And the audit promise is made about one agent only. Marshall is the one Salesforce says "provides an audit record of every action"; that sentence is not written about Piper or Hunter.
SDR — This is the agent aimed at your work, and it is not a pilot any more. Piper is GA, and Salesforce's own example is Asana, where it drives "4x the conversation volume" on the website. Find out this week whether your company has bought it. Then prepare the argument you will need, because it is a real one: an agent can answer an inbound form in seconds and still cannot tell that the person who filled it in is not the person with the problem. Get yourself onto the accounts where that distinction decides the deal.
AE — Your exposure is Hunter, not Piper, and you have until November. Salesforce says Hunter works a pipeline "from research to outreach, collaborating with sellers over weeks and months" — that is your account research and your sequencing, running next to you. Take one target account this month and write down every call you make on it that is not research: who you decide to walk away from, which executive you go around, when you stop. That list is the part of your job that survives the November date, and writing it now is cheaper than arguing it later.
Manager — "60% of Perk's sales pipeline" built by an agent is the number your leadership will read out to you. Ask what it counts: pipeline created, or pipeline closed. Salesforce says the first and does not say the second, and the gap between them is exactly where your team's work lives. Then settle one coverage question before the tooling settles it for you — if an agent books the meeting, who owns the no-show, and whose number does it come off.
RevOps — Piper being GA means it can be bought and deployed without you, and Salesforce puts the average deployment at 45 days, which is one quarter. Decide now which objects Piper may write to, which leads it must hand to a person, and what your audit trail actually shows when it qualifies someone out. Note what is missing: Salesforce makes the "audit record of every action" claim about Marshall, its supply chain agent, and makes no equivalent claim about Piper or Hunter. Closing that gap is your work, not the vendor's.
When an agent has already qualified the lead against what the buyer stated, the pain the form never captured is the whole of your remaining edge. Map Non-Obvious Buyer Pain Before Discovery builds ranked hypotheses across workflow, political, reporting and cost-of-inaction pain, with a question to surface each one.
Two minutes, once a week. What changed in AI, and what to run because of it.