The enterprise software market that built the public-cloud era — Salesforce, ServiceNow, Workday, Adobe, Atlassian, and the broader category of subscription software companies that charge per-seat pricing for business applications — faces the most significant strategic threat in its history from the agentic AI thesis. The argument is straightforward: if AI agents can perform substantial portions of the work that human users currently perform within these applications, then the business value of seat licences declines proportionally with the work that agents take over. A company that previously needed 100 Salesforce licences for 100 sales operations staff may need fewer licences if AI agents perform a meaningful share of the data entry, lead qualification, and pipeline management work that the licences supported.
The financial implications for enterprise SaaS are existential in the most aggressive interpretation of this thesis. Per-seat pricing has been the primary revenue model for the category for over a decade, and the high-multiple valuations the enterprise software companies have achieved reflect the assumption that seat counts and per-seat prices would continue to grow with customer adoption and renewal cycles. An environment where seat counts plateau or decline because work has been agentified represents a fundamental change in the unit economics that justify those valuations.
The honest analytical question is not whether agentic AI represents a real threat to enterprise SaaS — it does — but how different enterprise software companies are positioned to respond to that threat and which defences are actually working. The answers vary substantially across the category in ways that the broad “agentic AI disrupts SaaS” narrative does not capture.
Salesforce’s Agentforce Response
Salesforce has been the most aggressive enterprise software company in pivoting its product narrative toward agentic AI. Agentforce — the company’s AI agent platform — has been positioned as the response to the agentic threat: rather than losing seats to AI agents, Salesforce intends to charge for the agents themselves through a consumption-based pricing model that captures the work the agents do.
The strategic logic is coherent. If AI agents are going to perform work that previously required human users, the customer relationship that delivers those agents to the enterprise can capture revenue from the agent usage itself rather than from the human seats the agents displace. Salesforce’s incumbent customer relationships — the deep integration of Salesforce into enterprise sales, service, and marketing operations — provide the distribution channel through which Agentforce can be deployed to existing customers without requiring them to integrate with a new vendor.
The execution challenge is whether Agentforce can deliver agent performance that customers will pay for at prices that replace the revenue from seat licences being displaced. The early Agentforce deployments have demonstrated proof of concept in specific use cases — customer service automation, sales lead qualification, account management workflows — but the conversion from seat-based revenue to agent-consumption revenue has been gradual rather than rapid. Salesforce’s reported revenue mix continues to be dominated by traditional seat licences, with Agentforce representing a smaller but growing component.
The Salesforce stock performance in 2025 and 2026 has reflected market scepticism about the pace at which Agentforce can replace the seat revenue that may erode. The shares have lagged the broader software sector even as Salesforce has reported solid headline growth, with valuation multiples compressed to levels significantly below the company’s historical norms.
ServiceNow’s Different Bet
ServiceNow has pursued a different strategy that emphasises the platform integration angle rather than the agent product angle. ServiceNow’s Now Platform serves as the workflow infrastructure for enterprise IT, employee service, customer service, and an expanding set of horizontal use cases. The company’s positioning is that AI agents need an enterprise workflow substrate to operate within — visibility into the data, integration with the systems, and orchestration of the actions that agents need to perform — and that ServiceNow provides that substrate.
The strategic argument is structural. AI agents that perform business work need to interact with enterprise systems (HR systems, IT service management, customer service tools, finance applications) that ServiceNow already integrates with. An agent that wants to take action on behalf of a user — file an HR request, resolve an IT ticket, update a customer record — operates through the workflow infrastructure that platforms like ServiceNow provide. The agent does not displace ServiceNow’s value proposition; it requires ServiceNow’s value proposition to function.
The investor reception of this thesis has been more favourable than Salesforce’s positioning. ServiceNow’s stock has performed substantially better than Salesforce over the past two years, with valuation multiples remaining elevated as growth has sustained at high rates. The bull case is that ServiceNow’s platform position is structurally protected from agentic disruption because agents need platforms; the bear case is that the same dynamic could be replicated by competing platforms or by AI infrastructure providers that build their own workflow capabilities.
Workday and the HR Application Layer
Workday has been the most cautious of the major enterprise software companies in its AI messaging, which reflects both the company’s culture and the specific dynamics of the HR application category. HR software faces a different agentic dynamic than CRM or ITSM: the work that human resources professionals perform involves significant judgment, compliance considerations, and human-centric tasks that are harder to fully automate. The agentic threat to Workday is real but is concentrated in specific subcategories (payroll processing, benefits administration, talent acquisition automation) rather than across the breadth of the HR function.
Workday’s response has been more measured: AI capabilities integrated into the existing application rather than a separate agent product, partnerships with AI infrastructure providers rather than a clear-positioned competing agent strategy, and emphasis on the data and workflow integration value of being the system of record for HR. The reception has been mixed — the stock has performed reasonably but has not benefited from any clear AI narrative that has supported other software companies.
The structural question for Workday is whether the HR application category continues to require dedicated software at the scale that Workday’s valuation implies, or whether agentic capabilities reduce the breadth of dedicated HR software in ways that compress the category’s growth trajectory. The probable answer is somewhere in between — HR will continue to require dedicated software but the per-employee revenue may be pressured by automation of the higher-volume, lower-judgment tasks.
The Adobe Case Study
Adobe represents an interesting case study because the company’s primary competitive threat is not the agentic seat-displacement dynamic but the disruption of its creative software franchise by AI-native alternatives. Figma’s emergence as a collaborative design tool already pressured Adobe’s design business; the generative AI capabilities that have emerged for image, video, and design creation have created additional competitive pressure from new entrants.
Adobe’s response has been a combination of Firefly (its generative AI platform integrated into Creative Cloud applications), continued investment in the broader Creative Cloud subscription, and the controversial Figma acquisition attempt that was abandoned after regulatory pressure. The execution has produced mixed results — Firefly is genuinely valuable for existing Adobe users but has not necessarily expanded the addressable market the way the original AI thesis implied, and the competitive pressure from AI-native creative tools has been more significant than Adobe’s initial defensive narrative suggested.
The strategic lesson from the Adobe case is that defensive AI integration into existing products is not necessarily sufficient when the competitive threat comes from new entrants whose products are AI-native rather than AI-augmented. The same dynamic applies in principle to other enterprise software categories: the AI defenders may be protected from agentic seat displacement but may face different threats from AI-native entrants whose products are not bound by the constraints of legacy software architectures.
The Hyperscaler Threat Vector
The threat that may matter most across enterprise SaaS over a 5–10 year horizon is the hyperscalers’ move into application-layer AI. AWS’s Bedrock and the broader hyperscaler AI infrastructure, Anthropic’s enterprise positioning, and the AI agent capabilities that Google and Microsoft are building directly into their productivity stacks (Copilot in Microsoft 365, Workspace AI in Google) represent a structural threat to dedicated enterprise SaaS that operates above the application layer.
The argument is that if Microsoft can build Copilot agents that perform sales operations work integrated with Outlook, Teams, and the broader Microsoft 365 stack, then a portion of the value that Salesforce currently delivers can be captured within the productivity software that enterprises already use. The same dynamic applies to Google Workspace for the segments of the workforce that operate primarily in Google’s productivity stack. The structural advantage of operating within the productivity layer where employees already work is significant for capturing agentic work that does not specifically require the dedicated enterprise SaaS application.
This threat has not yet materialised at scale — Microsoft Copilot and Google Workspace AI are still in early enterprise deployment, and the integration depth with dedicated enterprise applications has limited the direct displacement so far. But the trajectory is concerning for the enterprise SaaS category because the hyperscalers have distribution, integration capability, and AI infrastructure advantages that the dedicated enterprise software companies cannot easily match.
The Honest Investor Assessment
For investors evaluating enterprise SaaS exposure: the category is not uniformly threatened, and the variation within the category is significant enough that selective positioning produces meaningfully different outcomes than category-level allocation. Platform plays like ServiceNow that benefit from agent infrastructure demand may continue to perform well even if traditional seat-licence pricing is pressured elsewhere. Application-layer plays like Salesforce face more direct seat-displacement risk and require the agent revenue conversion to work at the pace that justifies current valuations. Vertical-specific plays like Workday face category-specific dynamics that depend on how automation affects the underlying workforce in their target markets.
The valuations across the category have compressed meaningfully over the past two years as the agentic AI thesis has been absorbed by the market. The current multiples imply lower growth expectations than the historical pattern suggested, which means that the equity outcomes depend significantly on whether the categories deliver against the lower expectations or whether they disappoint relative to even the reduced bar. The honest position is that selective enterprise SaaS exposure remains attractive but the simple buy-and-hold thesis that characterised the category during the cloud computing era is no longer sufficient — active assessment of each company’s agent strategy, its platform vulnerability, and its hyperscaler relationship is required to navigate the disruption that is genuinely under way.
