The future of work is here, and it’s digital. With the rapid evolution of artificial intelligence, businesses are on the cusp of a revolutionary transformation—one driven by AI-powered “digital labor.” This new era introduces AI agents capable of performing tasks once limited to human employees, offering unprecedented scalability, efficiency, and reach. As Salesforce CEO Marc Benioff aptly puts it, “This is really the rise of digital labor,” and its success hinges on thoughtful implementation. OpenAI Chief Executive Officer Sam Altman calls agents “the next giant breakthrough.” 

What Is Digital Labor?

Digital labor refers to AI agents that act as virtual employees, handling complex tasks with precision and speed. Think of these agents as high-functioning interns—indefatigable, logical, and incredibly well-read on your customer data, knowledge bases and policies. But, like any intern, their performance depends on the quality of their training: the processes they follow, and the context they are given, and the automations they can use. And then how they are managed, measured, and coached.

Agents are not “magic”’. Many of the time-proven disciplines and approaches for managing a human workforce directly apply. The difference is that our digital workforce needs far more direct guidance because it takes instructions literally.  The looseness of words – ambiguity – leads to seemingly random results. 

AI-driven agents deliver four key benefits:

• Efficiency: Tasks are completed faster, often with minimal human intervention.
• Accuracy: Programmed to minimize errors, digital agents excel in repetitive, detail-oriented work.
• Cost Savings: By replacing manual tasks, companies can significantly reduce operational costs.
• Scalability: AI systems can handle increased volumes without the need for additional staff.

Despite these advantages, digital labor demands clear direction and robust process design to ensure its effectiveness

Agent or assistant?

These AI agents fall into two broad categories. There is naturally some blurring  but it is useful to see agents in these groupings:

1. Agent-Style AI: This is where the AI agent is likely to be replicating the work of a human with “agent” in their tile; customer agent, support agent, travel agent.  So they are ideal for structured, predictable tasks like scheduling appointments, processing returns, or generating invoices. These are complex multi-step tasks and have a clear outcome which is probably using an existing workflow to post data into a system. So they could be autonomous.
2. Assistant-Style AI: The AI Agent is replacing the tasks delivered by someone with “assistant” in their title; executive assistant, marketing assistant, revops assistant. They are geared toward subjective, collaborative tasks such as summarizing meetings, identifying customer churn risks, or drafting product deals. Here the output is more likely recommendations that is assessed by a human. The agent really is assisting the existing teams to be more effective and is less likely to be autonomous.

In the short term, the agent-style Agents are going to be the most prevalent.  We estimate that 90% of agents will be this type in the next 2-3 years.  In the US there are over 3 million call center agents. Some or all of their jobs are at risk, particularly the highly repetitive tasks. As we gain experience and confidence with AI agents, their scope will expand to handle more and more complex scenarios.

---

Lessons in Process Reengineering

Introducing agents into the operation it isn’t about adding technology to existing processes. Instead, it requires process reengineering. We need to revisit the agent processes and streamline them to exploit the agent’s data and knowledge assimilation capabilities. We need to design an end-to-end process geared around an exceptional user experience. None of this is new. Methodologies like LEAN and Kaizen have been around for decades. In a recent interview, Microsoft CEO, Satya Nadella said that he is “going back to school” on LEAN to understand how to apply it in an AI context. LINK

This is not surprising. We know that engaging business users to map out business processes in detail and then making them available to employees drives up adoption and consistency which leads to higher performing teams. The challenge was always the diagramming notations like BPMN and UML were so complex and took too long to produce, and only the creators understood them. Alternatively, flowcharts were too simplistic and were unable to deliver the detail that is required to direct teams in the context of the overall end-to-end process.

The UPN (Universal Process Notation) process mapping standard, that is now championed by Salesforce, bridged that gap. It has been around for over 25 years. Its hierarchical mapping approach combined with a simple set of shapes enabled diagrams to be readable by everyone, and the top-level goals were encompassed in the lowest-level diagrams. Plus the ability to drill down and add attachments such as training materials, metadata links and requirements to the activity boxes means that the detail is available but not overwhelming. UPN is a single view that the business, executives, regulators and IT can all agree on.

There are plenty of examples where UPN has delivered staggering results for organizations of every size and industry. Here are just a selection:

• Investment bank: “If we had had our processes up-to-date when we started the restructure, we could have executed it in half of the time that we did.”
• Defense contractor: “Inefficient textual procedures have been replaced with visual processes which played a significant part in winning more than $30 million of new business.”
• Food manufacturer: ”It is a way of creating a step change in our ways of working as we look for new and innovative ways of staying ahead“
• Construction: “It provided demonstrable corporate governance, improving business performance and pinpointing shortcomings costing more than $25m”
• Telecommunications “After the merger we realized that Company X achieved more after six months than Company Y (parent of Company X) had managed after 2 years.”
• Oil and gas: “A process model helps articulate and pinpoint shortcomings in ways that are simply not possible when relying on language alone“
• Food and drink manufacturer: ” End-to-end process thinking to break functional silos is absolutely critical for our company to deliver our business model“
  
• Retail bank: ”Our cost per transaction has reduced from €16 to €3, and we have exceeded our target for automatic processing“
• Retailer: “We’ve achieved really dramatic improvements in terms of processing time and reduction of outstanding debt.”
• Construction engineering: “We’ve identified process improvement savings of $198,000 – per day”
• Medical equipment manufacturer: “We achieved an 80% reduction in textual documents and better focus of activities with the added bonus of a 50% reduction in time spent training new recruits.”
• Defense contractor: “This has resulted in actual savings of $8 million and further projected savings of $6 million per annum.’
• Automotive manufacturer: “Speed of access to processes and local procedures was increased by 34 times”
• Retailer: “The primary success criteria is mass adoption. Put simply it’s about helping the right people do the right things at the right time.”

Process reengineering in the agentic world

We’ve already established that agents require teams to apply disciplined process reengineering to streamline the processes before they are agentified. 

We have been promoting the whole notion of design thinking and process thinking. When it comes to AI, having the mindset of process pattern thinking, I think it’s going to be really critical, especially as we are held accountable to having ROI at the end of the day with improving these technologies to improve processes.

Juan Perez, CIO, Salesforce

But rather than blindly applying the approaches from the last 20 years to AI, we’ve looked at the approach for building reliable agents. And then managing, coaching, and promoting them. We’ve looked at the UPN notations and it has been enhanced for the agentic world – UPNA – (Universal Process Notation for Agents). It has not lost any of the readability and simplicity that made it so popular.

We’ve also looked at how AI can help build the diagrams and then coach the creators to build better processes. This is now possible. Mapping processes in live workshops was not as time-consuming as people imagined. But now AI has completely changed the narrative.  Instead of starting with a blank canvas, AI can draw the diagram based on your Salesforce implementation, a sketch, an existing diagram or a text document like a SOP or SOW.

To optimize digital labor, organizations must:

1. Redesign for Simplicity: Streamline processes before introducing automation.
2. Focus on Exceptional User Experiences: Consider the needs of customers, partners, and employees alike.
3. Leverage AI for Innovation: Use AI to uncover insights, spot trends, and execute tasks previously deemed impossible because of the volumes of data or insights that were locked up in unstructured data.

The only place Agentify comes before Simplify is in the dictionary

Ian Gotts, Founder & CEO, Elements.cloud

The process diagram is the agent design canvas

Each agent’s capabilities are described in an Agent Interaction Map (AIM) which is a hierarchical map that drills down to a diagram that describes the agent Topic (in Agentforce terminology). This is the agent’s end to end process to deliver a task  e.g.  Answer An HR Policy Questions; or Book Vacation And Submit For Approval. 

When well designed, it will automatically generate the agent instructions, actions, and guardrails. It provides governance and versioning for the agent. It also generates the test utterances.

What is more fundamental is that you are providing the unambiguous direction to the agent. You are “programming” using natural language. If you want to change the logic, the validation, the rules, or the guardrails, you simply drag lines or change the text on the diagram and regenerate the agent. 

The diagrams also define how and when the agent passes back to a human and how it provides a summary of the prior conversation and hand-off information.

Finally, the diagrams are how you provide governance and traceability to the agent’s performance and actions.

Pause for a moment to think about the implications of this.

The process diagram was how regulated industries directed staff and drove process improvements. The process diagram is how you direct agents. Change a diagram and you have instantly redirected an infinite digital workforce.

Here is a solution guide for building UPNA diagrams

---

Managing Digital Labor: Building a High-Performing AI Team

Much like managing human employees, digital labor needs to be managed:

• Onboarding: Introduce AI agents to your organization’s policies, standards, tasks, knowledge, and resources. But also how agents interface with the humans.
• Monitoring and Coaching: Continuously checking on the performance so that you can fine-tune their processes, knowledge, or actions
• Promoting and sacking: Extending the agent’s capabilities is the equivalent of promotions. What happens if you need to withdraw an agent – i.e. sack it? How do you transition back to pre-agent operations?
• Governance: Version controls, access restrictions, and change logs are required to maintain oversight. These are all encompassed in the diagrams. This is not just for regulated industries.

The  Agent Interaction Map (AIM) is the framework that ensures consistency and alignment with organizational goals. And it provides the necessary governance and visibility to manage agents. It is the “operations manual” for your army of agents.

A Call to Action: Embrace the Future

The rise of digital labor marks a turning point in how businesses operate. But its success depends on thoughtful integration, rigorous process design, and a commitment to fostering adoption across all levels of the organization. As we navigate this transformation, the focus must remain on leveraging AI not just for efficiency, but for innovation and growth.

By adopting a process-first mindset and treating AI agents as integral members of the team, organizations can unlock the full potential of digital labor. 

You would not your human team to be effective without clear direction. Why would agents be any different?