In the space of about two weeks, Microsoft, SAP and Salesforce all announced AI agents that work the front of the sales funnel. They research prospects, qualify leads and, in some cases, write and send the outreach themselves.
For a sales team, that sounds like extra hands. But an agent does not bring its own judgment about your customers. It follows the process and the data you already have, so it is worth looking at both before you switch one on.
Three vendors, the same idea

Microsoft’s Sales Development Agent, covered in its October 23 partner announcements, researches prospects, personalizes outreach, follows up and qualifies leads. It has been in public preview since October 8. In Copilot Studio, Microsoft also added Agent Accounts, a way to run agents for recurring processes with permissions, triggers and a memory of how the process ran before. That part is still in private preview.
SAP’s customer experience release on October 21 added a Deal Qualification Assistant. One agent enriches a lead with company context, another scores it on fit and engagement signals, and a third turns it into a sales opportunity. SAP says it is available to select customers, with no price or general release date announced.
Salesforce introduced Hunter at Dreamforce, an outbound agent that finds accounts and contacts, drafts and sends outreach under rules the team sets, and keeps working on deals over weeks and months. It is in pilot, with general availability planned later this year.
An agent follows the playbook you actually have

Notice what these agents depend on. SAP’s agent judges fit. Salesforce says Hunter learns how your top sellers win. Microsoft’s agent qualifies leads. All of that assumes a business has a clear idea of what a good lead looks like, and data that reflects it.
In many small and medium-sized companies, that knowledge lives in a few people’s heads. The CRM has duplicate contacts, old deals nobody closed and fields that were filled in differently by every rep.
An agent working from that will not fix it. It will act on it, faster and at a larger volume. If your definition of a qualified lead is fuzzy, the agent’s version will be fuzzy too.
Decide what it may do without asking

An agent that drafts an email is very different from one that sends it. A wrong draft costs a minute of a rep’s time. A wrong message sent to a prospect costs trust, and you may not see it happen.
The vendors seem aware of this. Microsoft built permissions into Agent Accounts, and Salesforce describes Hunter as working under rules the sales team sets. Those rules still have to be written by you.
A sensible start is to let the agent research and draft, and keep a person approving anything that goes out to a customer. Loosen that only once you have seen enough of its work to trust it.
Before you turn on a sales agent
These agents may well save time, especially on research and follow-ups that reps tend to skip. But none of the three vendors has published results from real customers yet, and pricing is still unclear for some of them. A few steps help either way:
- Write down what a qualified lead means for your business, in plain words.
- Clean the CRM records the agent will learn from: duplicates, dead deals, inconsistent fields.
- Start with research and drafts, and keep a person approving outgoing messages.
- Measure meetings booked and deals won, not the number of emails sent.
- Ask the vendor how the agent is priced before usage grows.
The agent will do what your process tells it to. The useful work is making sure the process is worth repeating.
If an AI agent started qualifying your leads tomorrow, would it know what a good customer looks like for you?
Sources
- SAP News Center: Autonomous CX Arrives, What’s New in SAP Customer Experience Q3 2026 (Oct 21, 2026)
- ERP Leaders: SAP now lets AI agents sort sales leads and handle service cases (Oct 24, 2026)
- GEroundtable: Microsoft Unveils AI Sales And Service Agents For Copilot And Cowork (Oct 24, 2026)
- Salesforce: Meet Hunter, Your New Outbound Sales Agent
