Salesforce ended Friday near $164 after a volatile week shaped by opposing analyst calls. The consensus target remains roughly $246 across 52 analysts, but the market is applying a steep discount to that target because Agentforce adoption evidence is still largely operational rather than financial. The setup is asymmetric: credible proof of incremental ACV could rerate the stock quickly, while another quarter of usage metrics without revenue conversion would validate the lower-growth multiple now embedded in the shares.

Starting this week, new agents must be created in Salesforce’s redesigned Agentforce Builder rather than the legacy experience. That makes the migration more than a cosmetic release: admins and implementation partners now have to standardize on the new orchestration model, test existing workflows and update enablement materials. Forced platform cutovers are useful adoption signals because they reveal whether customers are moving beyond pilots, but they also concentrate support risk in the exact quarter when investors want evidence that Agentforce can scale cleanly.

Salesforce is packaging nearly two years of service-agent learnings into a prebuilt Help Agent and Customer Service Portal, both generally available in July. The more consequential change is commercial: pay-per-resolution pricing ties vendor revenue to completed outcomes rather than seats or conversations. If customers accept the model, Salesforce gains a cleaner proof point for AI ROI and a path to monetize automation even when human-seat counts fall; if resolution quality is inconsistent, the same pricing structure exposes the product gap immediately.

A new industry analysis argues that agents are less likely to erase SaaS than to compress interfaces and shift value toward data, permissions and orchestration. That framing favors entrenched platforms only if they open their systems to external agents while preserving governance and transactional integrity. For Salesforce, ServiceNow and Workday, the durable asset is not the chat surface; it is the trusted workflow and data layer beneath it. The risk is that model vendors capture the user relationship and reduce application vendors to lower-margin infrastructure.

Salesforce’s Summer ’26 release introduces a single agent entry point backed by specialized subagents and expands Setup with Agentforce for administrators. The product direction is clear: fewer disconnected copilots, more centrally governed task routing. That can reduce adoption friction and strengthen platform lock-in, but it also raises the bar for observability—buyers will need to know which subagent acted, what data it used and how each action maps to cost before they allow multi-agent workflows into core operations.

Salesforce says the planned Fin acquisition will complement its native Help Agent with quicker deployment paths for smaller service organizations. Strategically, this acknowledges that time-to-value—not model access—is the binding constraint for many customers. The combined portfolio could widen Salesforce’s addressable market below large enterprises, but integration discipline matters: overlapping agent products and pricing models can create the same complexity that buyers are trying to remove.

ServiceNow reports second-quarter results on July 22, giving investors the next clean comparison for whether agentic features are expanding platform consumption or merely defending renewals. The company positions its AI platform as model- and cloud-neutral, which may resonate with customers wary of single-vendor stacks. Watch subscription growth, remaining performance obligations and any disclosure separating AI upsell from core workflow demand; those metrics will set the valuation bar for the broader enterprise-software group.

HubSpot says organic traffic across its customer base is down 27% year over year while AI referral traffic has roughly tripled, and it is responding with AEO tooling, smart deal progression and specialized agents. The signal extends beyond marketing software: AI interfaces are changing customer-acquisition economics before they fully change back-office workflows. Vendors that help customers measure visibility and conversion inside answer engines can create new budget pools, but they also inherit attribution problems that traditional web analytics never solved.

Gartner forecasts 47% growth in worldwide AI spending this year, with model spending more than doubling, yet says most organizations still prefer AI embedded in existing applications over wholesale process redesign. That is favorable for incumbent SaaS vendors in the near term because distribution and trusted data matter more than raw model novelty. The caveat is margin: if model and inference costs rise faster than customers’ willingness to pay, vendors can grow AI usage while diluting the economics of the base subscription.

New research on agentic software economics finds that lower development costs make custom tools more plausible, but integration, maintenance, security and organizational switching costs still protect established SaaS systems. The practical implication is segmentation: lightweight workflow apps face the greatest substitution risk, while systems of record with dense data models and compliance controls remain defensible. Capital should follow vendors that can expose their data and actions to agents without surrendering governance or forcing customers into expensive custom integration.