Barron's just wrote that its Salesforce stock pick flopped because of AI. They're half-right, but they got the mechanism wrong.
Barron's picked Salesforce at $263 in late December 2025. The stock closed at $170.92 on June 10 - down 35% from their entry, 38% from its 52-week high of $276.80. The write-up blames "AI" as the generic reason the thesis broke. That's the kind of headline that sounds right until you open the hood and find the actual problem.
The problem isn't that AI is hard. The problem is that Salesforce is selling an AI product that fails at the customer level at a 77% rate, while Marc Benioff projects $63 billion in revenue by fiscal year 2030 on a $41.5 billion base that has grown below 10% for 18 months straight.
The math doesn't work. Not even close.
The $800 million illusion
Salesforce's Q4 FY26 earnings in February reported that Agentforce annual recurring revenue reached $800 million, up 169% year-over-year. Benioff called it proof that "agentic AI is a tailwind for our business." Sell-side analysts nodded along. The number grew fast enough to be headlined.
But $800 million in ARR on a $41.5 billion revenue base is 1.9% of the business. That's not a growth engine. It's a rounding error wearing a keynote costume.
Even if Agentforce continues its 169% compound growth rate - which is a bold assumption for any enterprise software product - it would take four years of that pace to push Agentforce ARR to roughly $3 billion. That still leaves the $38 billion-plus legacy CRM business doing all the heavy lifting toward the $63 billion target, while growing fast enough to offset the fact that the company has been stuck below 10% revenue growth since mid-2024.
The $63 billion-by-FY30 target requires a compound annual growth rate of roughly 11.3% for the next four years. On a $41.5 billion base, that means adding $5.4 billion in new revenue per year. No company of that size has sustained that trajectory while pivoting to a new product paradigm.
The 77% failure rate nobody headlines
Here's the engineering reality that Benioff's keynotes skip. Multiple independent analyses - including an October 2025 review by Oliv.ai that surveyed B2B Agentforce deployments - found that 77% of implementations fail. Not because the platform doesn't work in theory. Because the data feeding it is broken.
This is the kind of detail that matters more than any keynote slide. Agentforce requires clean, structured CRM data to function. Most Salesforce customers don't have clean CRM data. They've been running the platform for years with incomplete records, duplicate entries, and fragmented customer profiles. You can't build agentic AI on dirty data. It's not an AI problem - it's a data hygiene problem that Salesforce created itself by selling a platform whose customers were never forced to maintain it.
The community-level evidence tracks. Salesforce admins posting on Reddit report pricing models that change every time they look, support tickets that take 10 days to get a response, and client-facing agent workflows they no longer trust because the outputs are unreliable. One user summarized it bluntly: they stopped trusting Agentforce for client-facing work because the trustworthiness problem was unsolvable in production.
77% failure at the deployment level is not a product success story. It's an execution collapse.
Benioff claims 12,000 customers are using Agentforce. That number means nothing without context. How many of those 12,000 have working, production-grade deployments? How many are running pilot programs that stalled? The "customer" metric in enterprise software has always been elastic - a company with one sandbox trial counts the same as one with a five-agent production deployment.
The incumbent failure nobody credits
There's another dynamic Barron's missed entirely. Salesforce's stock decline isn't just about Agentforce under-delivering. It's about what happens when the company that defined enterprise software for two decades stops being the company that defines it.
Growth below 10% for a company of Salesforce's size and age signals something structural, not cyclical. The SaaS multiple compression that gripped the sector in 2025 - the "SaaSpocalypse" narrative - punished every software company. But Salesforce got hit harder because its AI pivot was the most marketing-heavy and the least engineering-proven on the list.
Compare that to companies that built AI products from a technical base rather than bolting agents onto a legacy CRM platform. The market doesn't reward the company that announces the most AI features. It rewards the company that ships something customers actually use at acceptable cost.
The cross-currents
The picture isn't one-directionally bearish, but the forces favor the downside:
- Agentforce ARR growth is real - $800M at 169% YoY is genuine momentum, but from a tiny base on a massive company. It can move the needle only if the overall business stops slowing down.
- The $63 billion target is aspirational math - 11.3% CAGR on a $41.5 billion base requires Agentforce to become a $4–5 billion business and the core CRM to grow at mid-single digits without further deceleration. Both conditions are unproven.
- Implementation failure rate is the hidden drag - if 77% of deployments fail, customer satisfaction drops, renewal risk rises, and the AI upsell story becomes a churn accelerant rather than a margin expander.
- Valuation compression reflects the truth - at ~20x earnings (the stock trades near $171 with a P/E around 20), Salesforce has already repriced from its peak of 30x+. The multiple still supports a company growing at 10%+, not one that needs to reaccelerate to 14%+ just to hit management's own targets.
Directionally, the execution risk outweighs the revenue momentum. Agentforce can grow fast without fixing the company's core growth problem.
What Barron's should have written
The Barron's article frames the loss as "AI disappointed." That's backwards. AI didn't disappoint - Benioff's keynote math disappointed. The stock fell because investors finally priced in the gap between the $63 billion projection and the sub-10% growth trajectory, between the 12,000-customer claim and the 77% failure rate, between the $800 million headline and the 1.9% of revenue reality.
Salesforce is still a $140 billion company generating real cash flow from a real enterprise installed base. But the thesis that Agentforce saves the growth curve - that's the claim that needed engineering verification. It hasn't passed.
At $171, the stock has already absorbed the disappointment. The question now is whether Benioff's $63 billion target is a roadmap or a press release. The evidence so far suggests the latter.