AI agents as digital labor: Challenges and solutions
What is digital labor?
Digital labor refers to work performed by non-human agents — from rule-based bots to sophisticated AI systems — that automate or augment tasks traditionally done by humans. This can include anything from customer service and data entry to more advanced decision-making and communication tasks. As AI capabilities grow, digital labor is shifting from simple automation to intelligent, adaptive work.
AI agents enter the workforce
One of the most transformative evolutions in digital labor is the rise of AI agents. they are autonomous software programs that perform tasks with minimal human input. They are fast emerging as the new form of digital labor in today’s workplace. Unlike traditional bots or scripts, these agents leverage large language models (LLMs) and machine learning to make decisions, interact with users and systems, and carry out complex assignments independently. In an era where businesses seek greater efficiency and scalability, such AI “workers” are becoming invaluable teammates rather than just tools.
Salesforce CEO Marc Benioff predicts that today’s CEOs will be the last to run purely human workforces, as we enter an age of AI-powered “digital labor” co-working with humans.
This blog post explores how autonomous AI agents are transforming the workforce by handling tasks traditionally performed by humans. We’ll define their key characteristics, look at real applications across industries like customer service in real estate and healthcare, discuss perspectives from executives like Benioff and UiPath’s Daniel Dines, and consider the challenges and future outlook.
The goal is to give business and product leaders an approachable yet comprehensive understanding of AI agents as digital labor and explain how to begin harnessing their capabilities, including via AI agent platforms like Sendbird.
Automate customer service with AI agents
What are the characteristics of AI agents?
What makes AI agents suited to serve as digital labor? A few defining characteristics set LLM agents apart from basic automation or static chatbots. Let’s review them.
AI agent autonomy
AI agents can work independently without step-by-step human direction. They perceive inputs, make decisions, and execute tasks on their own. In other words, an AI agent is an intelligent software program that can self-determine what it will do. Unlike a scripted bot that follows a fixed flow, an agent dynamically decides to gather more information, consult a database, or retry a different strategy to achieve a goal—all on its own.
AI agent adaptability
AI agents are learning-oriented and adaptable. They can improve over time and adjust to new inputs or scenarios. For example, when traditional chatbots break while facing an unexpected input, an AI agent can further abstract the conversation to react more effectively. This makes AI agents suited for ambiguity and complex processes.
As UiPath’s CEO describes, robots (RPA scripts) are great at repetitive tasks, but agents excel at adapting to changes and handling dynamic processes. This adaptability lets agents tackle the “long tail” of functions that previously weren’t automatable because they required too much judgment or context. Over time, an agent can be updated or trained to expand its skills, becoming more capable as it gains experience—much like a human worker developing expertise.
👉 Read on about AI agents vs. chatbots
AI agent scalability
Cloud-based agents offer global, scalable AI workforce power—they can handle large volumes of work and scale up more easily than human labor. Once an AI agent is developed, it can be deployed to serve many users or execute many tasks in parallel. It doesn’t tire or need a paycheck and can operate 24/7. In practical terms, a single AI agent can instantly take on the work of an entire team of humans around the clock. This makes them extremely efficient for high-volume and always-on functions like customer service or monitoring.
In addition, as demand grows, spinning up more instances of an AI agent is much faster (and cheaper) than hiring and training new staff. The result is an elastic digital workforce that can scale with business needs.
Examples of digital labor
AI agents are already influencing various job functions across industries. Below are a few areas where autonomous agents are making an impact.
AI customer service
Customer service is one of the most impactful and widespread use cases for AI agents. These intelligent agents can converse with customers to answer questions, resolve issues, or guide them through processes, essentially acting as virtual assistants.
A compelling example is AskRedfin, an AI assistant designed to streamline the home search experience. It engages prospective buyers with qualifying questions, interprets their needs, and uses listing data to show the most relevant properties.
By handling the bulk of the research and recommendation process autonomously, the AI agent reduces browsing time while boosting engagement. A human sales agent still steps in eventually, but this happens after a lead is qualified by an AI agent, dramatically increasing the conversion for sales.
This hybrid approach of AI for scale and human for depth is quickly becoming the blueprint for modern customer service.
AI agent platforms like Sendbird make it easy to seamlessly deliver similar experiences in mobile applications and integrate with support software to transfer customers to real people at the opportune time.
👉 Check out Sendbird’s introduction of AI for customer service
Onboarding with AI agents
Onboarding at scale can overwhelm HR and operations teams, especially when it involves hundreds of monthly gig workers, contractors, or seasonal employees. It’s repetitive, time-sensitive, and error-prone.
A logistics company can turn to AI and deploy conversational AI agents to walk new drivers through documentation, training modules, FAQ responses, and compliance steps. And if a question falls outside of its scope, it escalates to a human within the same chat thread.
The result? Faster ramp-up times and a better first impression for every new hire due to effective support.
AI administrative automation
The potential for AI agents in healthcare is massive, particularly for administrative and support tasks that consume care teams’ time.
For instance, the healthcare AI company Innovaccer recently announced its “Agents of Care” suite, which includes specialized AI agents for many clerical tasks:
One agent automatically books, reschedules, and manages patient appointments.
Another agent automates patient intake by collecting pre-visit information and coordinating follow-ups.
Yet another manages referral workflows to connect patients with the right specialists.
There’s even an authorization agent to speed up insurance prior authorizations by gathering necessary data and submitting requests—a typically tedious process for healthcare staff.
On the patient support side, a patient access agent can serve as a 24/7 multilingual help desk, answering routine patient inquiries and helping them navigate services or get information anytime.
By offloading these “busy work” tasks to AI, healthcare organizations can relieve their overburdened staff and let human professionals focus on patient care. AI agents are designed to integrate with hospital systems (like electronic health records) to fetch relevant data securely and provide informed assistance while maintaining compliance.
Imagine never having to wait on hold to book a doctor’s appointment and getting paperwork processed in seconds. AI agents are laying the groundwork for efficiency in healthcare.
👉 Learn more about AI agents in healthcare with 12 examples plus their pros and cons
Delight customers with AI customer service
Industry perspectives on AI agents
The rise of AI agents as digital labor has not gone unnoticed by industry leaders. Executives at the forefront of technology and automation are actively discussing and shaping this trend. Here, we highlight insights from two prominent figures—Marc Benioff of Salesforce and Daniel Dines of UiPath—on how AI agents are transforming work and what it means for businesses.
Marc Benioff (Salesforce)
As head of a major enterprise software company, Marc Benioff has been vocal about the advent of AI-driven work. He describes the current wave of AI agents and automation as “the rise of digital labor," and believes it represents a huge economic shift, potentially a $3 to $12 trillion global transformation as AI becomes embedded in business processes.
Benioff has even restructured Salesforce’s strategy around this idea: the company rebranded its AI assistant platform to “Agentforce,” signaling a move from simple AI copilots to more autonomous agents within its software offerings. These agents are envisioned to proactively assist with tasks like building marketing campaigns or answering complex customer queries, effectively acting as extra team members.
Benioff expects CEOs and leaders to prepare to co-manage a workforce that includes both humans and AI entities. In a recent interview, he bluntly stated, “I’ll be the last CEO of Salesforce who only managed humans.” This perspective emphasizes that ignoring AI agents is not an option for companies—those who embrace this unlimited, always-on workforce stand to gain a competitive edge in productivity and innovation.
At the same time, Benioff and others acknowledge it raises important questions about leadership, culture, and how to responsibly integrate AI into organizations (more on that later in the AI agent considerations section).
Still, the overall tone from Salesforce’s CEO is optimistic: digital agents will augment human workers, handle the drudgery, and enable people to focus on higher-level, creative, or relationship-oriented work. In his view, the AI agent revolution is not some far-off future—it’s here now as “an incredible new opportunity for all of us.”
Daniel Dines (UiPath)
From the automation and RPA (robotic process automation) side of the industry, Daniel Dines offers a complementary perspective. As CEO of UiPath, Dines champions “agentic automation”—essentially the fusion of classic RPA bots with AI agents to automate entire end-to-end business processes.
He explains that traditional RPA robots excel at repetitive, rules-based tasks (for example, data entry and form processing). In contrast, AI agents excel at understanding context, adapting to changes, and making decisions in less structured scenarios.
By combining the two, companies can automate workflows that were previously too complex to hand over entirely to machines. “Agentic automation is the natural evolution of RPA,” Dines says, highlighting that this next generation of automation will use both robots and AI agents working together.
In practice, this might look like an AI agent orchestrating a process—say, handling an insurance claim—where it uses RPA bots to pull data from legacy systems, then uses its own AI reasoning to decide the next steps, and perhaps asks for human input only if a special case or critical decision arises.
Dines notes this approach allows entire business processes to be automated—instead of only isolated tasks—and leads to “more substantial business outcomes, greater productivity,” while freeing human employees to focus on higher-value activities. He also envisions AI agents becoming “first-class citizens” on the UiPath platform, meaning they will be a core component of enterprise automation toolkits, just like human users and bot users.
The underlying philosophy in Dines’ commentary is that AI agents should augment and uplift the human workforce by taking over the dull, menial, and exhausting parts of jobs to allow people to concentrate on strategic, creative, or interpersonal aspects. This aligns closely with Benioff’s augmentation theme.
Together, the perspectives of Benioff and Dines paint a picture of a near future where AI agents are embedded across business functions—from customer service desks to back-office operations—working alongside humans and traditional software robots. The consensus is that companies should start preparing for this by investing in an AI agent platform and reorganizing workflows to take advantage of smarter, more agile business processes.
👉 Read our blog posts:
Reimagine customer service with AI agents
Agentic AI considerations and challenges for the workforce
While the promise of agentic AI in the workforce is exciting, it raises important considerations and challenges. Business leaders must navigate these to successfully and responsibly implement autonomous AI agents for customer service or face consequences ranging from poor customer experiences to regulatory missteps.
Job displacement concerns around the AI workforce
A common fear is that deploying AI agents could eliminate human jobs. It’s true that if an agent can handle a task end-to-end, it may reduce the need for staff in specific routine roles.
Some note that AI agents could potentially replace entry-level human agents or brokers for basic tasks. These concerns can understandably cause anxiety among employees. However, many experts argue AI will disrupt the workforce but not outright replace humans; instead, it will augment human workers’ capabilities.
History has shown that technology tends to create new jobs even as it makes others obsolete. AI agents could spawn roles like AI supervisors, trainers, or analysts to manage and improve those agents. Nonetheless, managing the transition is crucial.
Companies should be proactive in reskilling and upskilling their workforce so that employees can move into higher-value positions alongside AI. Clear communication about how AI agents will be used (and how human roles will evolve) is key to maintaining trust and morale during this transformation.
AI oversight and ethical governance
Handing autonomy to AI agents doesn’t mean “set and forget.” On the contrary, oversight is critical. Organizations must ensure that AI agents are making appropriate decisions and adhering to company policies, ethical standards, and legal regulations. This requires transparency in how agents operate.
Experts highlight that a major challenge in expanding AI agent usage is understanding what exactly the system is doing and how it’s doing it, and being able to verify that its outputs are correct.
Black-box AI that can’t explain its actions is risky, especially in regulated industries. For example, if an AI agent declines a loan or makes a medical scheduling decision, you need to know why. There’s also the issue of bias and fairness: if an agent learns from historical data that contains biases, it might perpetuate or amplify those biases in decisions (e.g., hiring, lending, etc.).
Ethical governance frameworks should be in place to audit AI agent decisions and ensure they align with fairness and inclusion goals. Human oversight is often implemented via a “human-in-the-loop” approach for sensitive tasks, where a person reviews or approves the AI’s actions.
Additionally, companies should establish clear accountability: if an AI agent makes a mistake, who is responsible? Business leaders and policymakers are actively discussing the need for guidelines and regulations around this and calling for a robust legal framework.
In summary, deploying AI agents requires careful governance, which entails continuous monitoring, the ability to intervene or shut down agents if they go awry, and compliance with ethical and legal standards. It’s wise to start with AI agents in low-risk roles and gradually increase their responsibilities as confidence in their behavior and guardrails grows.
👉 Learn more about AI risk for business
Managing AI performance
Like any employee (digital or human), an AI agent’s performance must be measured and managed. Businesses will need to set KPIs for their AI agents, such as the resolution rate for a support agent, accuracy for a data-processing agent, or response time for a scheduling agent. In addition, they should regularly evaluate whether or not the AI is meeting expectations.
Unlike humans, AI agents might not complain if something is wrong, so it's up to the organization to catch issues. One challenge is that when an AI agent does make an error, it could do so at scale (affecting many transactions quickly), so robust testing and phased rollouts are important.
Sendbird knows that expanding the use of AI agents will demand tools to inspect and verify the agent’s reasoning—essentially, AI observability. Emerging technology for AI “explainability” will help show why an agent made a certain choice, which is useful for trust and refining the agent’s algorithms.
Another aspect of performance management is continuous training: if the business processes or environment change, the AI agent might need an update or retraining to maintain effectiveness. This means companies should invest in keeping their AI knowledge bases and models up to date, similar to how we invest in ongoing training for employees.
Lastly, feedback loops are important—just as a human worker gets performance reviews, an AI agent should get feedback. This can be through reinforcement learning signals (rewarding desired outcomes in the model training) or simply by human oversight, such as catching mistakes and correcting them so the AI can learn.
In sum, treating AI agents as part of the workforce means also managing them as such: tracking their work, ensuring quality, and continuously improving their capabilities with transparency and accountability in mind.
Leverage omnichannel AI for customer support
What’s ahead for AI and the workforce
Looking ahead, the role of AI agents in business and society is set to expand rapidly. We are likely only at the beginning of this transformation. Here are two key aspects of what the future may hold.
Advancements in agent capabilities
The intelligence and autonomy of AI agents are improving at a remarkable pace.
Advances in prompting techniques like chain of thought and model distillation allow agents to break down problems into steps, evaluate tradeoffs, and make decisions more accurately. Pair that with long-term memory and autonomous execution loops, and agents can now carry out multi-step workflows independently, improving over time.
We're also seeing the rise of multi-agent collaboration, where specialized agents, each with a role, interact with one another to complete complex goals. Each agent operates with autonomy but coordinates through shared context and logic. It's the beginning of AI that works together, not just for us.
New infrastructure is emerging to support these capabilities. Standards like the Model Context Protocol (MCP) make it easier for agents to interface with tools and systems in a reliable, scalable way. Open frameworks like LangGraph, CrewAI, and AutoGen give developers building blocks for memory, tool use, and decision-making logic, turning what once took months to build into something you can now prototype in a day.
This evolution sets the stage for agents to graduate from experimental pilots to mission-critical teammates. They’re starting to mirror how people work—with judgment, coordination, and the ability to grow in capability over time.
The rise of the AI agent-to-agent (A2A) economy
As AI agents become more connected and collaborative, we're seeing the early formation of a new economic layer: the AI agent-to-agent economy. In this emerging paradigm, agents don’t simply augment individual roles—they transact, collaborate, and orchestrate work across organizations and systems. It’s a shift from automation to autonomy at scale. These agents can negotiate, execute tasks with other agents, and interface with digital infrastructure in real time.
The implications are massive—as businesses adopt agent-based systems, new market structures, job categories, and business models will likely emerge.
👉 Explore the 3 stages of the agent-to-agent economy
Reskilling and workforce transformation
The rise of AI agents is poised to reshape the workforce, though not by replacing humans wholesale, but by changing the nature of work itself. As routine, repetitive tasks become increasingly automated, many job roles will evolve, and entirely new ones will surface. Navigating this transition will be one of the defining challenges—and opportunities—for organizations in the years ahead.
Jobs centered on transactional tasks (e.g., basic customer triage, record-keeping, etc.) will be among the first to change, as AI agents handle these functions with greater speed and scale. But rather than eliminating roles, this shift creates demand for new ones: AI trainers, process analysts, workflow designers, and ethics stewards—people responsible for shaping how AI agents behave, what data they learn from, and how their performance is measured.
Beyond new jobs, most existing roles will also transform. A marketing specialist might rely on an AI agent to pull real-time reports, freeing time for strategy and creativity. A support rep may shift from handling simple tickets to overseeing complex escalations, acting more as a problem solver than a first responder. This doesn’t merely require technical knowledge—it requires new ways of thinking about collaboration between humans and machines.
For business leaders, this presents a clear mandate: invest in reskilling and upskilling programs that empower employees to work effectively with AI. That means teaching people not only how to use agents but also how to interpret their outputs, correct their mistakes, and apply their strengths in tandem.
Some companies are already introducing training programs to help staff adopt AI into daily workflows; others are redesigning teams to treat agents as digital coworkers embedded in the org chart.
Ironically, AI agents may help accelerate this shift. Learning agents, personalized to employee roles, could offer on-demand training, simulate real-world tasks, or guide workers through new workflows at their own pace.
Ultimately, the future of work won’t be a binary choice between humans and machines. It will be a partnership where human insight, judgment, and empathy are amplified by AI’s efficiency, scale, and consistency. Organizations that recognize this—and prepare their people for it—will be best positioned to thrive in the age of digital labor, aka the AI workforce.
Empower your support agents with AI
The future of work is hybrid: Humans + AI
AI agents are no longer theoretical. They’re already transforming how businesses operate by executing tasks quickly, scaling effortlessly, and learning continuously. As digital labor, they’re helping companies unlock new productivity levels across customer service, operations, and more.
But the real value of AI agents won’t come from replacing humans—it will come from pairing their capabilities with ours. The most forward-thinking organizations integrate AI agents into their workflows with the intention of reskilling their teams, setting ethical guardrails, and building oversight to ensure AI acts in service of people, not instead of them.
So, where to begin?
Start small. Choose a repeatable, high-volume task—such as handling support FAQs, appointment scheduling, or internal queries—and pilot an AI agent to manage it.
You don’t have to build from scratch. AI agent platforms like Sendbird make it easy to deploy them across your app, website, SMS, WhatsApp, and email. Built for enterprise scale, Sendbird gives you the tools to build, test, and improve your agents—all while ensuring compliance, security, and reliability.
Companies that act now—and wisely—will gain a serious edge. The future of work is a hybrid one: humans and the AI workforce collaborating. And it starts with the first agent you put to work today.
