“Every employee is a manager of AI agents”
As AI agents take on more workplace tasks, employees are increasingly becoming responsible for directing, evaluating and controlling digital workers.
The world of work is entering the next phase of the AI revolution. In recent years, we have used artificial intelligence as a tool to assist with coding, writing, searching, analyzing and streamlining tasks. Now, AI agents are moving beyond simply helping employees get the job done. They are taking on tasks and roles of their own, managing entire processes and becoming increasingly autonomous. This shift is reshaping the role of the human worker, transforming employees from people who use AI into people who manage it.
This transition is taking tangible form with the expansion of Agentforce, Salesforce’s AI agent platform. The company is introducing a new portfolio of “ready-to-work” AI agents designed to execute business tasks across sales, service, commerce, employee experience and back-office operations. These agents connect to Customer 360, operating on the basis of existing organizational data and business processes. They come equipped with role-specific capabilities, while new Agentforce technologies enable them to pursue long-term goals, acquire new skills, collaborate with other agents and continuously improve.
The shift promises to transform not only how we work, but also organizational structures, team sizes and the skills employers seek. Employees will need to define goals for agents, manage multiple agents simultaneously, evaluate their outputs and determine when they can operate autonomously and when human intervention is required.
1 View gallery
From left: Shimon Hason, Hadass Admon, Shachar Efal
(Studio Art, Jenny Feingold, Hagar Bader)
While some tasks and roles may diminish or evolve, entirely new roles will also emerge. The question facing organizations is no longer simply what AI can do instead of humans, but how to manage a workplace in which humans and agents collaborate.
Agentforce enables companies to deploy predesigned agents for roles where AI can handle a significant portion of the workload. These include Casey, a service agent that resolves customer issues; Paige, an IT and HR service agent; Carter, who assists customers during the purchasing process; and Hunter, a sales agent that manages the workflow from research to customer outreach while working alongside sales staff.
Hadass Admon, Director of Product Management at Salesforce Israel, says the challenge for many companies has been getting AI systems up and running in the first place.
“AI has already entered the realms of service and IT, but the ‘cold start’ problem always exists. In other words, you have an agent, but it doesn't yet know how to handle customers or execute actions. Consequently, large enterprises with the necessary budgets turn to third-party companies for assistance with customized implementation, and the results are amazing.”
Many customers struggle with that “blank slate” scenario. The next step in the evolution, therefore, is the introduction of ready-made, “out-of-the-box” agents.
After two years of experience, Admon says, Salesforce’s product teams now work closely with customers across the AI space, and the transformation has extended throughout the company, including its sales and service divisions.
Organizational data naturally varies from one company to another. The challenge is determining how to build agents and AI systems around that data. That is where the out-of-the-box approach is intended to solve the “cold start” problem. Customers begin with existing automations, operational instructions and a model already in place, significantly shortening the implementation process.
Still, Admon emphasizes that the technology is not something a company can simply switch on.
“Adjustments are required, but instead of starting from a blank page, the organization begins with a significant portion of the work already done.”
The change could be particularly significant for small and medium-sized businesses.
“For an SMB that wanted to deploy agents but couldn't due to a lack of manpower and capital, the ‘out-of-the-box’ concept solves a critical problem. We are seeing shorter implementation times and increased tool adoption. This is no longer limited to giant corporations; it affects every company in the economy. The changes will permeate deeper and become more dramatic. Companies that previously couldn't afford this kind of work now have an administrator who can make adjustments and get up and running. It is the scale that is changing dramatically, not the essence.”
Salesforce data indicates that the use of agents is expanding. According to the company’s “Agentic Enterprise” index for 2026, based on Agentforce platform usage data from February 2025 through April 2026, the average number of agents deployed per organization has nearly tripled over the past year. The average agent is now capable of performing six skills, compared with two in early 2025, while the frequency of employee interactions with agents has risen by 300% during the analysis period.
If more of the work shifts to agents, what is left for the employee?
“The nature of the employee's work changes based on the types of inquiries they receive. Clearly, in the service sector, there is no longer a need to perform repetitive tasks, such as updating an address. Employees also use AI to execute system actions, such as updates and automations. In service and sales, their skill sets are evolving. Everything is becoming much more intense. We have to be constantly interacting with our screens.”
Does this mean every employee will need to have a tech-oriented mindset from now on?
“AI is turning all of us into technologists. Back when I was a service representative, the cognitive load was immense. In a way, this technology makes things more accessible, effectively elevating us all. With the click of a button, I can now create things I never could have dreamed of producing before, and it’s about quality, not just speed.
“My role as a product manager has changed significantly, too. I have more time for direct customer engagement and genuine interaction because I spend less time writing specifications and sitting in meetings with engineers. Our development pace has shifted as well. Salesforce used to release three versions a year, but now we roll out a new version every two weeks. The pace of work has changed, meaning people who can't keep up simply aren't a fit. Consequently, the profile of the ideal employee is evolving, and I’m seeing that ‘10X’ impact everyone talks about playing out in real time.”
At Incredibuild, one can already see what work looks like when agents handle a significant portion of the workload. The company, which has operated for more than 20 years developing technologies to accelerate software development, has expanded into AI over the past year, integrating agents into development, support and sales functions.
“Until now, the conversation around AI centered on whether it could write code, a sentence or a document,” says Shimon Hason, CEO of Incredibuild. “We have moved to a stage where AI can actually perform work, and the question becomes how do we manage that? Management has traditionally focused on people. Now, job interviews will ask not only about a candidate’s approach to managing employees but also about managing agents. Candidates will be asked to describe a work process involving them.”
When working with people, managers set goals and objectives and define areas of responsibility. Now, each of us is becoming a manager of agents. That requires companies to consider how to establish control and decision-making processes for teams in which they do not want AI making decisions autonomously.
The idea that AI will simply do the work and render human workers unnecessary misses a broader point, Hason argues. There has always been more potential work than people could handle on their own.
One of the most fundamental shifts in the next phase of the working world is that managerial capability is no longer the exclusive domain of managers. When AI can perform tasks in seconds that previously took hours or days, it generates a vastly larger volume of potential work and, consequently, a greater need for management.
“I believe that every employee is, in a sense, a manager, and the very concept of management is evolving right before our eyes. We focus too much on the idea of ‘AI replacing someone’ and not enough on the fact that work itself needs to be conducted differently, and that if managed correctly, we can achieve far more.
“One of the questions posed to us by our hundreds of global clients, spanning banking, automotive, aviation, defense, healthcare and gaming, is whether we can help accelerate AI adoption safely. This acceleration process is no longer just about mathematics and speed. It is also about ensuring safety. We’ve told employees that they shouldn't merely use AI as an aid, but rather aim for a state where AI can operate almost autonomously. The metric we track isn't ‘show me how much you use AI,’ but rather ‘how much of your output was generated with the help of AI.’”
At Incredibuild’s new AI venture, Hason says, 90% of the code is generated by AI. But the shift is not limited to programmers.
One example is the “Deal Desk,” the operational and commercial function responsible for tasks such as working with sales teams and issuing price quotes.
“The Deal Desk team, which used to consist of five employees, is now entirely AI-driven. We transitioned the former manager to a different role, although oversight and security processes remain in place. The technical support team used to have 13 employees. Today, there are two. AI handles 80% to 90% of support requests. The support manager, using the time freed up by this shift, has begun working on new infrastructure and is building a new product that assists organizations.”
Niva Oz, a Customer Success Manager at Incredibuild, illustrates how the use of AI agents can enable employees to take on more advanced roles. She previously worked as a Deal Desk Manager, a position involving significant operational work. By leveraging AI agents, she was able to free up time, focus on more impactful tasks and move into customer success.
“I worked on a sales support team in a role that involved a great deal of manual work: switching between systems, verifying and cross-referencing data, tracking reports, opening tickets and sending follow-ups to customers. Tasks that used to take me hours are now performed automatically by the agents. They access the systems to gather and cross-reference information.
“Thanks to the time that opened up, I was able to focus on more complex tasks, learn new areas and advance to a Customer Success role where I work directly with clients. Even in this new role, AI continues to work alongside me. Integrating data with sales systems allows me to approach every client with a more accurate picture, provide faster and more targeted responses, identify business opportunities and dedicate more time to work where my experience and value truly matter.”
These examples also illustrate why the debate over job losses remains unsettled.
Admon says, “I don't think it will lead to layoffs. Jobs are simply changing.”
Hason describes a more complicated reality.
“There are people we let go because the tasks they performed shifted to AI. Some teams were downsized, too. We all need to understand where we need to change. There are people who don't want to change, don't realize change is necessary or lack curiosity. Yet today, it is more critical than ever to possess personal drive and a willingness to learn and break new ground. As a company, we are learning how to manage AI. It is a new discipline in the business world.”
Alongside role reductions in some companies, data from IDC and Salesforce from 2025 also points to the creation of new employment opportunities. According to the data, 23% of employers created new roles as a result of AI adoption, compared with 14% that cut positions.
The same report projects that by 2030, agentic AI systems will handle 22% of the average employee’s weekly workload, particularly in marketing, sales and IT. In sectors including financial services, retail, healthcare, manufacturing and logistics, the impact is projected to exceed 20% of total workload.
“It’s not that we have less need for people. We are still hungry for more talent, because managing AI is a challenging task requiring effort, judgment and the ability to maintain control. We need to ensure that people know how to manage this powerful tool and develop the necessary processes and management frameworks,” Hason said.
This shift is creating not only new ways to perform existing roles but also entirely new roles and areas of specialization.
Admon points to the “Forward Deployed Engineer,” or FDE, an engineer who works directly with customers to implement technology within their specific environments. Hason describes growing demand for such engineers as the company expands its AI-driven development and sales efforts.
“Today’s salesperson has access to far more relevant leads, complete with concrete information, enabling them to sell more efficiently. This creates a heavier client load, necessitating the recruitment of additional sales staff and process support personnel, specifically Field Deployment Engineers who assist clients in implementing the product.”
The demand for these skills is already transforming how companies train their workforces.
Ness, a company with 6,500 employees, launched “Ness Forward,” an enterprise AI-driven upskilling platform designed to prepare its workforce for FDE roles. The platform offers seven training tracks: development, data, infrastructure and cloud, project management, SAP, QA and a general track.
The curriculum combines practical, real-world tasks and is tailored to an organization’s specific needs, systems and business goals. The aim is to translate theoretical knowledge into hands-on execution capabilities. Training also extends to management and headquarters staff, with the goal of bridging the skills gap and enabling employees to remain relevant and advance within a rapidly changing job market.
Shachar Efal, CEO of Ness, says: “Artificial intelligence is evolving at an exponential rate, yet organizations are getting stuck due to a human capital skills gap. Today, corporate AI training is no longer merely a tool for employee enrichment. It is a business growth engine and a critical financial hedge. Our adaptive learning platform enables every organization to precisely tailor training processes to its needs, turn AI tools into direct cost savings and ensure that every employee, from technologists to finance professionals, delivers maximum value and productivity in this new era.”
According to Salesforce’s skills model, work in this new environment relies on a combination of three skill sets: human skills, skills for working with AI agents and business skills. Underlying all three is AI literacy, meaning the ability to understand the technology, use it effectively and recognize its limitations.
That raises another question: What happens to junior employees when many of the tasks that once served as entry points into a profession are precisely the tasks AI can now perform rapidly?
Hason does not believe the need for junior employees has disappeared.
“It remains important to understand software and mathematics, and now it is also crucial to understand how artificial intelligence works so it can be leveraged effectively. The nature of work is changing, and that is fine.”
The broader question, however, is not simply how many jobs AI will eliminate. Intellectual labor is a resource, while AI can operate around the clock at a low marginal cost. The more important question is how companies will utilize and manage that capacity.
Organizational structures may become flatter, and small teams could eventually generate output that once required dozens of people. But that alone will not be enough. Managers will need to know how to articulate goals, establish testing processes and design workflows around AI.
“I think the way my generation currently manages AI is still somewhat clunky, but new methods, approaches and tools for AI management will inevitably emerge.”
