Highlights
- The AI promises have been strong, but there is some concern as to whether or not the reality meets expectations.
- The cost of AI has been driving businesses to cut back on AI usage, also known as the ‘tokenpocalypse’.
- These patterns aren’t new – the dot-com bubble represented a promise that didn’t match the timeline, and the digital transformation age was slow until the catalyst of the pandemic.
- FOMO is driving many decisions, and costs are driving even more. Staying level-headed, being vigilant, and validating ROI are key to long-term success in the AI space.
I am convinced that the AI revolution is well and truly upon us, and this is not a technology that will fade away. Artificial intelligence, at least in the form that it takes today, is a beneficial tool and is proving its worth in many ways. Our recent Salesforce Architect Survey 2026 showed that Salesforce Architects are using it for many reasons: from handing off menial tasks to accelerated learning.
This wasn’t the story for all our survey respondents, and it’s also not the same story that we’re seeing everywhere else in the world, Salesforce or beyond. AI shows big promise, but the uptake is slow and confusing to many. This article will discuss these hurdles in detail. I’ll also give you some personal advice on what I would be doing next, from many different perspectives.
The Promise
Artificial intelligence will change EVERYTHING. Life will never be the same. Jobs will be automated, work will become optional, and the world will become a utopia. I’m exaggerating, but there were certainly promises that AI would change the world to a degree somewhat like this.
In the Salesforce world, we were told that our productivity would skyrocket and we’d see a significant reduction in human error. Marc Benioff started teasing his AI vision for the future of Salesforce all the way back in 2014.
Salesforce’s brand identity has shifted from the world’s #1 CRM to the world’s #1 agentic enterprise enablement tool. Salesforce has updated the naming of core products to include the Agentforce name, and even dubbed its developer marketplace ‘AgentExchange’. Salesforce could rebrand to Agentforce, and it would feel like a calculated decision at this point.
So… why does the actual, real-world situation seem so different from the promised dream? Why is it that not every Salesforce customer has a totally different platform from the one they had five years ago? Agentforce has been around for almost two years, with Salesforce first adopting generative AI in 2023 with Einstein GPT.
The most notable promise from Salesforce was when Einstein GPT and Copilot were pushed back in 2023 – this is when messaging really pivoted and became more optimistic. The promise, ultimately, was that Salesforce’s AI capabilities would move from just prediction (i.e. machine learning) to generation. The promise was that AI could construct personalized sales emails, draft knowledge articles for service agents, and summarize content. This would all be done on the Trust Layer, which kept your CRM data protected while your workforce was overhauled.
Then came the era of the Agentic Enterprise, starting in 2024 with the promise that AI could begin taking actions on your behalf, instead of just generating suggestions. This promise shifted AI from the position of an assistant to that of an autonomous agent that operated independently.
Salesforce painted a picture of a world where digital agents could resolve complex customer service issues, negotiate basic sales, or run marketing campaigns from start to finish based solely on high-level guidance from human orchestrators. The humans would only be required when an edge case arose or when it needed strategic guidance.
What a world this was going to become! Businesses would have the option to evolve, or try to compete with robots that performed tasks in a fraction of the time that meatbags could. The messaging was clear: adapt, or be left behind.
The Reality
Here we are in mid-2026, and reality isn’t quite what it has been pitched to be. We’re absolutely beginning to see the benefits of AI, but some of the promises of automating entire workforces aren’t quite what they were chalked up to be. Agentforce adoption isn’t quite where we expected it would be, with Marc Benioff even addressing this directly at Dreamforce ‘25.
It’s clear we’re well on our way to the future that has been predicted, but the acceleration is slower than anticipated. Tokens are the bane of any technology enthusiast’s existence, with AI usage coming at a rather high cost. This cost is further exacerbated by poor implementation, meaning an AI bill but no real output shown as a result.
Many businesses are either hesitant to invest much further in AI, or are being buried in token hell. As an AI vendor, you’re walking a tightrope between funding and the cost of hosting these super-powerful computing machines. There’s a constant need to expand the physical compute capacity. Customers pay for that in the form of usage-based tokens. Dubbed the ‘tokenpocalypse’, customers who have been searching for use cases to wedge AI in are now looking at cost-cutting.
The Previous Digital Revolution
We’ve seen this before. During the late 90s, the dream was sold: the internet would change the way that we do everything. If a company added “.com” to its name, its valuation went up whether or not it actually added value to the business. Once investors started realizing this, the bust hit hard. The real value came as the internet evolved, when businesses started adding value through the web. Decades after the initial promise, businesses that had not adopted an online presence were behind the 8-ball when compared to their online counterparts.
The fear of missing out (FOMO) drove a major boom, followed by a major bust, followed by a new equilibrium years later. The dot-com bubble burst not because the internet was a fad, but because people misunderstood the timeline. Over the next few decades, infrastructure caught up, the tangible value was proven, and reality caught up to the vision.
The same thing occurred years later when businesses were being convinced that they needed to become digital-first in terms of their backend systems and business enablement. Salesforce was a major player here, adopting the “No Software” slogan from the get-go and promising that cloud-based enterprise software was the future. They were right, of course, but began speaking this truth long before most customers were ready.
Once again, we saw a delay in businesses truly understanding the value of digital transformation. Then, the pandemic happened. The message became: “adapt, or be left behind”. Businesses learned the benefits of digital transformation very quickly, and new tooling was rapidly developed and enhanced to support it.
The Likely Trajectory for AI: History Repeats
My personal belief is that AI is an absolutely invaluable tool that will destroy businesses and governments that ignore it. However, I’m not convinced that the time is now. Businesses may be feeling the urgency, but chasing a technology without realizing true value is not a good strategy. Adding an AI agent to your CRM for the sake of it isn’t going to be a good move for your business. You need tangible value, and a true ROI on your AI investment.
What I predict is that we’ll see the bubble burst relatively soon. Vendors cannot continue to absorb billions of dollars of investor funds for compute costs, which means the cost of tokens will go up. Businesses that aren’t seeing real value will likely deprioritize AI efforts in the short term. You’ll see an uptick in the number of businesses employing human beings to perform service tasks once again. But this will not last forever.
AI costs will eventually come down. Just like the post-dot-com boom era, AI will lose popularity until hardware catches up. There is a severe backlash against AI in the consumer space at the moment as well, with the cost of everyday computers going up as a result of memory and storage shortages attributed to the AI/data centre boom.
One trend that I see a lot is boycotting AI. This is emotional, and the decision to do so is plain stupid in my opinion. If you want to have a leg up on your competition, it’s important to at least consider the options available, and become knowledgeable about how to use these tools as they evolve. A boycott isn’t going to hurt the AI industry – it’ll only hurt you.
Final Thoughts
AI is the latest in a long string of scientific and technological breakthroughs that have reshaped how we operate as a species. We’ve seen adoption patterns like this before; the dot-com bubble and the slow uptake of digital transformations give us an idea of how this current wave of new, innovative technology may play out in the long term.
What are your experiences with AI, both in and out of the Salesforce world? Is it living up to your expectations? Exceeding them? Or falling short? I’d love to hear your stories – leave them in the comments.
