Salesforce signed a definitive agreement on September 29 to buy Listen Labs, whose AI agents design customer studies, recruit from a pool of 50M+ participants in 120+ languages, run voice, video and text interviews, and build “digital twins” of customers to test messaging before launch. Terms were not disclosed. Press reports put the price around $2B, about four times the ~$500M valuation of its early-2026 round. It will sit in Salesforce AI Labs next to Fin, which closed on September 10, with close expected in Q4 FY27. The strategic logic is clear: research, service and marketing data in one loop that Agentforce can act on. The open question is integration capacity, with two AI acquisitions landing in the same quarter.

CRM fell about 4% on Monday into the high $220s as Meta’s enterprise launch hit the group. It rose 3.4% on Wednesday as the Listen Labs deal and an Oppenheimer Outperform reiteration landed. Oppenheimer pointed to its customer survey showing steady Agentforce engagement and efficiency-driven IT budgets. The stock closed Friday at $234.69, against $234.30 a week earlier. It is still about 8% lower than a month ago and 13% below the $269.11 high. Consensus edged up to $281.08 across 55 analysts, so the post-Dreamforce gap between analyst models and the tape persists for a third week.

Meta’s new Enterprise Platform bundles the Muse agent, Meta Business Agent, the Muse API and Muse Code. The distribution base is already large: WhatsApp business agents handle sales and service conversations for more than 1M businesses each week. The gaps are just as large. There is no price list, no SLAs, no named enterprise customers, and Meta itself calls enterprise “a somewhat different muscle.” For Salesforce, the threat is not a CRM replacement but Meta owning the agent at the conversation edge (WhatsApp, Messenger) and treating the CRM as a back-end API. Last week’s Amazon-versus-Muse access fight showed how contested that edge already is.

Zenity Labs showed that a prompt-injection payload planted in a public Web-to-Lead form could hijack an Agentforce agent when it later processed the record. The agent would query account data, including deal sizes, and exfiltrate it over DNS, getting around Trusted URL redaction, or send phishing to contacts. Salesforce received the report on June 1 and confirmed full remediation by August 19, so there is no live exposure. The point for buyers is structural: every agent that reads untrusted inbound data is an attack surface. That makes governance and trust layers a product requirement, which is the Dreamforce pitch, and it now has to be demonstrated rather than asserted.

Three days after the Meta selloff, application software led the market. ServiceNow rose 4% to $139.73, Intuit 4% and Adobe 3%, and IGV gained 1% versus 0.5% for QQQ, a sign of rotation into the beaten-down group rather than a broad tech bid. ServiceNow gave the bounce something to stand on. Annual contract value from its AI products is now above $1B, and its renewal rate is 98%. Those are exactly the metrics the “AI eats SaaS” thesis says should be eroding. Salesforce has not yet disclosed a comparable gross renewal number, and analysts will notice.

The work-management cohort rallied together on October 1 with no company-specific catalyst. The same day, HubSpot began a new EMEA pricing model for new portals that consolidates 73 SKUs into 23, alongside a Fall Spotlight built around a shared “Growth Context” data layer for its agents. Simpler packaging around a data layer that agents can act on is the same move Salesforce made with Data 360 and Agentforce bundles. The mid-market fight is now over who prices the agent, not the seat.

The biggest single-stock casualty of Meta’s launch was not a CRM peer but MongoDB, which lost about 21% on news that Desai was leaving. He was a ServiceNow product chief before MongoDB, and Meta hired an enterprise-software operator rather than promoting from within. That says Meta intends to sell to CIOs, not just developers. MongoDB’s September 29 analyst meeting was expected to reaffirm guidance. The bigger issue for every software board is talent: AI platforms can now pay to recruit proven enterprise leaders.

Silver Lake’s talks to take Workday private, valued at about $51B, remain unsigned. In the meantime Workday has authorized a $4B buyback and set its financial analyst day for October 13. Watch that date. A firm deal would set a private-market price for a seat-based HCM franchise in the AI era, which is the cleanest comparable available for Salesforce’s own multiple. A walk-away would suggest even private equity cannot get the financing math to work at current software prices.

Employers added 29,000 jobs in September against an 84,000 consensus, and unemployment rose to 4.2% from 4.1%. Rate futures moved to about 79% odds of a hold at the October 28 FOMC, after September’s hike. Friday’s session saw the Nasdaq up 1.2% and the S&P 500 up 0.7%, but the week ended mixed: the Nasdaq was fractionally higher, the S&P fractionally lower and the Dow down more than 1%. For enterprise software the read is two-sided. A Fed pause supports long-duration multiples, but a cooling labor market eventually shows up in seat counts and in the hiring-linked modules (HCM, sales seats) most exposed to headcount.