Agentforce as a stress test for enterprise AI readiness
Salesforce’s Agentforce is an enterprise AI agent platform designed to automate CRM, sales, service, and marketing workflows, but its sluggish adoption exposes deep, unresolved problems in data quality, system integration, and operational readiness across large organizations.
Agentforce was launched with bold promises that autonomous agents would transform how companies run customer-facing operations. Salesforce’s CEO said the company was “all in on Agentforce” when it launched in 2024, yet only 34% of customers have adopted it so far. KeyBanc reports that only about 23,000 of 150,000 customers are using the platform, and their downgrade helped wipe out a large chunk of the company’s market value as investors questioned whether this could be the next growth engine. Bernstein called Agentforce “still in early stage of adoption” and not a short-term growth driver. In other words, this is not a story about lack of AI hype. It is a story about CRM AI implementation colliding with reality.
Messy data and half-ready products: the real enterprise AI adoption barriers
KeyBanc’s CIO survey is blunt: customers’ data is “not in order to do meaningful AI work,” and “Agentforce, as a product, just isn’t there”. These are the core enterprise AI adoption barriers hiding behind the glossy demos. AI agents depend on clean, structured, connected data to make reliable decisions, but many enterprises still operate with fragmented CRM records, disconnected systems, and inconsistent customer information.
Initial users report spending as much time preparing and organizing data as they do using the AI, which defeats the promise of autonomous agents. This is not a minor implementation detail; it is a structural problem. The excitement around agentic AI assumes a level of data quality enterprise AI rarely has today. When the underlying customer data is riddled with duplicates, missing fields, and stale entries, AI agents become expensive error amplifiers. That is why so many deployments remain small proof-of-concept projects instead of full enterprise rollouts.
Salesforce’s growth story vs. the adoption reality
Salesforce insists Agentforce is the fastest-growing product in its history and points to customers going live in weeks instead of months. Marc Benioff has dismissed bearish analyst reports as a “bad call,” citing internal metrics to argue that the opportunity is larger than skeptics admit. There is truth on both sides: a low base can still support fast percentage growth, even while absolute adoption remains modest.
But the external numbers tell a colder story. Only about 23,000 of 150,000 customers are on Agentforce, and more organizations in KeyBanc’s survey expect to reduce Salesforce spending than increase it over the next year. Partners are only now converting Agentforce proofs of concept into deals, suggesting the sales pipeline is still forming rather than exploding. Salesforce is also contending with customer resistance to paying extra for AI features on top of rising CRM prices. The result is a tension between a polished AI narrative and a hesitant, cautious customer base that is unconvinced the product – or their own operations – are ready.
Data governance first, AI agents second
Agentforce’s struggles show that enterprise AI agents require serious data governance before deployment. AI agents are only as capable as the CRM and marketing data they rely on; when that data is siloed and inconsistent, autonomy collapses into manual babysitting. Organizations hoping to automate campaign execution, lead qualification, customer service, and personalization are likely to see greater returns from improving data quality, integration, and governance than from buying more AI agents before their CRM data is ready.
Salesforce seems to recognize this, quietly shifting investment toward data readiness. The company has added technology that automatically pulls customer data from external sources and expanded its data-management capabilities through acquisitions, including Informatica, to improve data integration and governance before customers deploy AI agents. That is the right order of operations. If enterprises treat AI agents as a shortcut around messy data and uneven processes, they will keep paying for tools their staff do not trust. If they treat agentic AI as the reward for cleaning up their data foundation, they might finally get the productivity gains they have been promised.
What Agentforce signals for CRM and marketing automation
The debate over Agentforce is less about one vendor and more about the state of enterprise AI as a whole. Agentforce’s adoption rate is a measure of enterprise AI readiness: the companies moving fastest will not necessarily be those buying the newest agent platforms, but those that already built the data foundation those systems need to deliver meaningful results.
For CRM and marketing automation vendors, the message is sharp. Agentic AI is not an add-on feature; it is a bet that customers have solved data quality enterprise AI problems that most have barely started on. Competitors and hyperscalers are also shipping their own AI capabilities, and analysts already expect Agentforce to be used mostly in Salesforce’s core CRM market rather than becoming a universal platform. Vendors that continue to sell AI agents into unprepared organizations will keep seeing stalled projects, downgraded ratings, and skeptical CIOs. Vendors that help customers fix their data and operational readiness before promising autonomy will own the next wave of enterprise AI adoption.
