What Salesforce Learnt About AI in 2025 and How 2026 Will Be Different

Last year was the year of the agent. Artificial intelligence crept to the forefront of countless sectors with the promise that the traditional chatbot was old news. In fact, even basic ChatGPT was so 2024. Autonomous agents – which had the power to work on your behalf – were here, whether you liked it or not. 

Now that we have entered 2026, the playing field is renewed once again. This year, we focus on the next step: figuring out how AI fits not only into your business, but into your team. So, whether that looks like having an AI ‘coworker’ or just another installment in an existing tech stack is yet to be confirmed, but how much can it help versus hinder?

AI’s Evolution (2025-2026)

Before we look ahead to what 2026 holds for AI in the tech sector, we must look back to 2025 and what it taught both businesses and consumers.

For perhaps the first time since artificial intelligence hit the mainstream, 2025 was the year that businesses not only had to prove that their AI tools worked, but that they were useful to their customers. Having an AI tool that could increase productivity, pick up menial tasks, or speed up processes was only one part of the equation – proving to stakeholders that this would happen in their business was the other. 
According to research from the research and advisory firm Keenan Vision, businesses that were pushing their AI tools were beginning to hook onto buyers eager to join the AI race and make the most of this innovative technology. However, many businesses that bought these AI tools faced a “pilot purgatory” where AI projects were getting stuck between the C-Suite’s expectations and the ability for teams to actually deliver, and ROI that was becoming increasingly more difficult to prove.

Not only that, but a lack of vendor-led education was further contributing to the AI skills gap, with different teams feeling confused and disjointed, and leaders just being drawn in by hazy promises with no real way to execute any deliverables.

Last year also taught us that AI is not something that will be going away anytime soon, positioning 2026 as a platform for substantial evolution. 

According to a report by the Organisation for Economic Co-operation and Development (OECD), most countries are currently placing a bigger emphasis on developing AI professionals than expanding general AI literacy, when the focus needs to be a lot more balanced. Preparing the next generation of AI-savvy workers is a necessity at this point, but if sales teams, marketing teams, or leadership teams lack the appropriate AI literacy skills, cross-company cohesion will continue to be a challenge.

AI in Salesforce

Last September, Agentforce, Salesforce’s proprietary AI tool, celebrated its first birthday. A year on from inception, the tool had grown in leaps and bounds, finishing 2025 with $1.4B in annual recurring revenue (ARR) and 18,500 total deals since its launch, up 50% quarter-over-quarter.

Salesforce’s Q3 was dubbed by us as the “quarter of Agentforce”, especially as the tool and Salesforce as a company had to navigate numerous challenges throughout the year, including confusing pricing models, slow adoption, and a lack of value realization. 

By the time Dreamforce rolled around in October, Salesforce’s entire MO appeared to center around AI and Agentforce. Dreamforce marked an opportunity to really gauge how Salesforce’s ecosystem was responding to the company’s intensifying direction, and the event certainly did just that. 

With swathes of the community questioning whether Salesforce had lost its way amidst the shiny allure of Agentforce, Parker Harris, Salesforce’s Co-Founder, had to set the record straight. Go all in, or “go obsolete”; this is what Salesforce was doing, and this is what Salesforce was going to continue to do. The path forward was not entirely clear, but it was decided. 

The ‘AI Replacing Jobs’ Debate

While 2025 showed that AI could open a lot of doors, it also made something else apparent: AI could close and seal many doors shut, too. 

2025 was not only the year of the agent. It was also, unfortunately, the year of AI-driven layoffs, with tech businesses cutting jobs in the thousands to dedicate more money to AI efforts or reduce spending on employees where it could (theoretically) be spent on AI tools that could do the job for a fraction of the price.

Here at SF Ben, we debated extensively over whether AI was being used as a scapegoat in these layoffs decisions and whether masses of jobs were suddenly being replaced by AI, and a lot of these questions seemingly remain unanswered. AI has been at least a partial reason behind layoffs and job displacement within the sector, but how many jobs it has actually affected is undetermined.

Current estimates suggest that up to 60% of current jobs will require significant adaptation due to AI, and that by 2030, 30% of current US jobs could be automated. We may not yet fully grasp the magnitude of the shifts occurring, but the landscape is undeniably in transition.

At the ASEAN Summit last October, Canadian Prime Minister Mark Carney explained the current AI climate well: “The old system is gone. The new one has not yet formed.” 

AI sits at the forefront of perhaps its biggest transformative period to date, where it will either push full-steam-ahead towards becoming a permanent fixture in teams and tech stacks, or crash and burn on its way to its destination. 

For companies that have already begun implementing AI, the pressure to prove its worth is at an all-time high, no matter the cost, and now, the initial effects of this are being felt. If a company pivots entirely to AI, existing and potential employees will likely be negatively impacted in some way. In fact, it has already started.

Coworker or Co-Helper?

If I were writing this in 2025 rather than 2026, I would say that, unfortunately for us humans, the ‘AI in the workplace’ dilemma does not run out of steam there. If AI doesn’t replace you, augment your work, or demote you, you risk it becoming your coworker. 

‘The AI Coworker is Here’ is the title of an article released by Forbes last July. While many of us were spending our summers in the sun, AI had snuck into our teams, our offices, and our work group chats like a thief in the night. Or had it?

What Forbes covered was largely true for the time: AI had become a competitor, not just a tool, and it was forcing professionals – especially in the tech space – to reconsider what it meant to be a ‘good worker’. As author Nono Bokete put it: “Doing what you were hired to do is no longer enough.”
This notion was circulating in the Salesforce ecosystem, too. Theories around virtual employees and plans around the CRM’s infamous ‘digital labor’ movement were throwing the future of work into question. Was AI really more of a coworker than a co-helper?

The Salesforce Case Study 

However, fast forward to now, and sentiments are a little different. One of 2026’s first AI trends has come to light, and Salesforce is one of the companies feeling the shift.

Salesforce originally promoted Agentforce as an autonomous AI that could handle customer issues largely on its own, with humans stepping in when necessary. But in real enterprise use, the AI has behaved inconsistently, sometimes giving different answers to the same problem or query, which has created a lack of trust, increased business risks, and, in some cases, driven up costs.

To combat this, Salesforce has added a new rule-based scripting layer known as Agent Script, which forces companies to define step-by-step logic so that the AI behaves more predictably. It goes a long way toward solving these problems, but it also shifts responsibility back to CIOs, who now have to design, test, and maintain these scripts themselves. 

Like with any step in the AI processes, it has its fair share of pros and cons. Although it gives businesses a potential solution to hallucinations or wandering AI, it also has the potential to increase costs, slow down AI rollout times, and mean that teams suddenly require new skills on top of the AI ones that they already have to learn. 

Not only that, but the attitude around AI needs to change, too. AI agents can no longer be seen as a digital employee or virtual coworker – they need to be acknowledged as complex systems. 

Sanchit Vir Gogia, the CEO of Greyhound Research, told CIO that this shift in mindset is necessary to move forward into a new year and era of AI. 

“AI is not a black box you license,” she said. “It is a system you operate. If you don’t have the people to operate it, you don’t have a strategy, you have a pilot.”

Final Thoughts

As we get ready to see what 2026 has to offer for artificial intelligence, agents, Salesforce, and the tech sector as we know it, we must understand one thing: the progress path will likely not be linear, and success metrics are likely to change more often than we think.

So whether AI is your coworker, demoter, or handy assistant this year, take the time to consider where you fit into the equation. The time to evolve is now.