Salesforce failed to replace thousands of workers with AI. What can your business learn?
- Salesforce laid off thousands of workers, intending to replace them with AI, only to reverse course and try to hire many of those workers back.
- Organizations exploring deeper AI integration into their own processes should learn from this example by taking a more thoughtful approach that focuses on solving specific problems and respects the limitations of AI tech and the value of human workers.
- CEOs overseeing an AI integration should set clear, realistic goals, continually reevaluate strategy and prioritize clear communication with their teams.
Salesforce grabbed headlines in 2025 when it announced layoffs for 4,000 customer support workers it planned to replace with AI. But in a major reversal, the company now wants to hire many of those workers back, as executives now believe that they overestimated AI’s problem-solving capabilities while underestimating the value of their own workforce.
While this may seem like the story of a single tech giant confronting the ashes of its own ambition, leaders in other sectors ranging from construction to healthcare should pay close attention. Keep reading to learn more about where Salesforce went wrong and how you can avoid its mistakes as you integrate AI into your firm.
Where did Salesforce’s AI strategy go wrong?
Last year, leaders at Salesforce believed their Agentforce AI technology was powerful enough to replace almost 50% of their 9,000-person customer support team — and also began pitching Agentforce as a tool that could allow their customers to do the same. This proved to be not just a mistake, but a costly one.
- As a result of the layoffs, Salesforce lost an enormous amount of institutional knowledge, experience and crucially, the trust of both laid-off team members and their colleagues who remained (studies show that layoffs significantly affect a company’s remaining workforce).
- But the trust problem wasn’t just internal. Because Salesforce sold customers on AI as a major workforce replacement tool, its overconfidence has affected customer trust as well.
Behind this overconfidence was the mistaken belief that AI could flat-out take over large numbers of human roles. Perhaps one day, it will be able to, but as of now, the tech simply isn’t there yet.
So what’s a better way to leverage AI today?
How should your business pursue an AI integration strategy that avoids the Salesforce trap?
You can learn from Salesforce’s mistakes as you develop a firmwide AI strategy for your own business. The biggest lessons are to see AI as a tool to support your team rather than outright replace roles, and to approach AI integration as a thoughtful, deliberate process.
Consider these five pillars as you develop your AI strategy:
1. Understand operational debt
You can't rebuild your entire organization around AI overnight. Integrating AI takes preparation, including putting all of your data into a centralized, organized form, before you can even begin to think about automating processes or finding efficiencies.
This preparation is about paying off what’s called operational debt — essentially, cleaning up enough of the mess in how you do things now that you’re ready to operate differently. Salesforce didn’t do that, leaping into huge organizational changes before considering whether they were remotely prepared to make them.
2. Respect institutional memory
Your team possesses deep institutional knowledge of not just the work you do, but also how things operate within your organization. Both are valuable, and the latter is especially difficult to replace because it’s knowledge that is unique to your organization.
Salesforce created a major headache for itself by firing a big chunk of its institutional knowledge. That gap couldn’t be filled by AI agents and placed additional strain on the remaining team.
3. Don’t assume that AI equals efficiency
To executives, a major appeal of AI is the promise of working more efficiently. But that’s not always the end result.
A growing body of research shows that AI can sometimes produce more work, not less, forcing team members to spend additional time completing tasks they could finish more quickly on their own. Rather than assuming AI equals efficiency, explore whether an AI tool actually delivers before setting expectations or making decisions.
4. Remember the value of trust
Salesforce tried cutting its customer support team, only to find out that customers looking for support often prefer to deal with a human being. Whether it’s a potential new customer trying to decide if a product or service fits their needs or existing customers looking for additional help, people want to feel like they can trust a company they’re doing business with.
AI can provide information, but it can’t build trust. So how is your organization going to do that?
5. Embrace change management
As you do integrate AI into your systems and processes, how can you shift your team to doing work that only a person can do? These are typically higher-value tasks (this is how you build trust), but don’t assume your team will know how to pivot on their own.
You have to take the time to rethink roles, have conversations with team members, including executives, about how they spend their time and what could be done differently. Communication is huge here: Your team needs to know what you’re doing and why you’re doing it, and you also need to create room for your team members to offer their own perspectives on this change.
You should also provide training or upskilling opportunities to help your workforce adapt as needed.
6. Pursue AI integration as a process focused on problems and solutions
Finally, AI integration isn’t a one-and-done event. This is likely something your organization will be navigating for years to come, so it’s important to approach it as an ongoing, evolving process.
Take AI in steps. Start by focusing on specific problems you can experiment with AI to solve, prioritizing by what’s easy to implement or offers high potential ROI.
This problems-and-solutions focus will help keep your AI efforts more disciplined and more likely to succeed.
What should CEOs do to lead a successful AI integration?
CEOs overseeing an organization that’s exploring AI integration are typically not involved in day-to-day implementation. But they can help set a baseline for success by:
- Setting clear goals: Organizationally, you have to crawl-walk-run. Set clear, achievable AI goals, and update them based on your progress or challenges.
- Evaluate continuously: AI tech is changing so fast. Your firmwide AI strategic roadmap from six months ago may not be relevant today, so you have to continuously evaluate your strategy (including from the perspective of acknowledging a Salesforce-like overreach).
- Communicate, communicate, communicate: The success or failure of any organizational change often comes down to communication. Recognize that change is scary, create open lines of communication and help your team understand not just what you’re doing, by why it matters.
How Wipfli can help
We help organizations use AI to deliver results. Let’s talk about your goals and how we can help you achieve them. Start a conversation.
