The News: Shield and OpenAI
Shield, an IT service management company, announced on November 13, 2026, that they are utilizing OpenAI technology to automate interactions with legacy software portals. The system, known as Forge, allows IT technicians to offload repetitive navigation tasks to AI agents. This shift aims to reduce the manual labor required to manage outdated enterprise systems, freeing human technicians to focus on complex troubleshooting that requires professional judgment.
This is a significant development because it addresses the last mile of enterprise automation. Most businesses have already digitized their core processes, but they are still bottlenecked by legacy portals that lack modern APIs. Shield is proving that AI agents can bridge this gap by interacting with legacy software as a human would, rather than trying to force the software to behave like a modern, API-first application.
This is not just about speed; it is about reclaiming the time that IT professionals spend on low-value, repetitive tasks. By deploying these Forge agents, Shield is effectively creating a new layer of abstraction over outdated systems. This allows the organization to keep its existing infrastructure while gaining the benefits of modern automation. It is a pragmatic approach that avoids the massive cost and risk of a full system replacement.
What is a Forge agent?
A Forge agent is an autonomous software component designed to navigate user interfaces within legacy enterprise portals that lack modern application programming interfaces. Unlike standard scripts that break when a button moves, these agents process visual and structural data to interact with the system as a human would. They execute clicks, input data, and navigate menus, effectively acting as an interface layer between modern LLM-based logic and rigid, outdated legacy software environments.
These agents operate by observing the screen, identifying the necessary fields, and executing the required actions. They are trained to understand the context of the workflow, not just the location of the buttons. This is a critical distinction. Traditional automation relies on hard-coded coordinates or specific HTML selectors, which are notoriously brittle. If a developer changes a label or moves a button, the script breaks.
A Forge agent, however, uses the underlying OpenAI models to recognize the intent of the interface. If the Save button moves from the bottom right to the top left, the agent understands that it is still the Save button and continues the process without interruption. This resilience is what makes agentic automation viable for real-world enterprise environments where software is rarely updated or maintained.
How does Shield automate legacy portals?
Shield automates legacy portal interactions by deploying Forge agents that interpret the user interface of older systems in real-time. Instead of relying on brittle API connections that often do not exist, these agents observe the screen, identify the necessary fields, and execute the required actions. This allows technicians to hand off the mechanical process of data entry or system navigation to the agent, ensuring that the work is completed without the technician manually clicking through every screen.
The process begins with the agent analyzing the current state of the portal. It identifies the relevant input fields based on the technician's request. For example, if a technician needs to update a user's permissions, the agent navigates the menu hierarchy, locates the correct user profile, inputs the new permission settings, and confirms the change.
Throughout this process, the agent maintains a log of its actions, providing a transparent audit trail. This is crucial for IT compliance. The agent does not simply do the work; it provides a verifiable record of exactly what was changed and when. This allows the human technician to review the agent's work at any time, ensuring that the automation is performing as expected.
Why is legacy software the next frontier for agents?
Legacy software is the next frontier for agentic automation because it is where most enterprise work gets stuck. Large organizations cannot easily replace core systems that have been running for decades, so they remain trapped in manual, high-friction workflows. By using AI agents to wrap these systems in a modern, intelligent interface, companies can finally automate the last mile of digital work that was previously considered impossible to touch without a full system overhaul.
The sheer volume of legacy software in the enterprise is staggering. From banking mainframes to healthcare records systems, these platforms are the backbone of modern business, yet they are notoriously difficult to integrate with modern tools. For years, the only solution was to build custom middleware or to accept the inefficiency as a cost of doing business.
Agentic automation changes this equation. By leveraging the visual and linguistic capabilities of modern LLMs, we can now teach agents how to use these systems. This unlocks a massive amount of trapped value. It allows companies to modernize their workflows without having to modernize their core infrastructure, which is a much faster and more cost-effective path to digital transformation.
Are agents replacing technicians or upgrading them?
Agents are upgrading technicians by removing the low-value manual labor that currently consumes the majority of their time. The goal of the Shield implementation is not to remove the human from the loop, but to elevate the human to the role of an oversight manager. When the agent handles the rote navigation and data entry, the technician is free to apply their expertise to the specific, high-stakes issues that actually require human intuition and complex problem-solving capabilities.
This is a fundamental shift in the nature of work. Instead of being operators who spend their day clicking buttons in a legacy portal, technicians become architects of the automation. They define the workflows, set the parameters, and handle the exceptions that the agent cannot resolve.
This shift empowers the technician to do more, not less. It allows them to manage a larger volume of issues, improve their response times, and provide a higher level of service to their clients. It is a transition from manual labor to knowledge work, and it is a positive development for anyone who wants to focus on higher-level problem solving.
What does this mean for a one-person agency?
For a one-person agency, the Shield and OpenAI development signals that you can now automate workflows that were previously locked inside proprietary, non-API-friendly software. You no longer need to convince clients to upgrade their entire tech stack just to achieve automation. You can build agentic layers that sit on top of their current systems, providing immediate value by eliminating manual bottlenecks and allowing you to scale your output without scaling your headcount.
This is a game-changer for the solopreneur. In the past, if a client had a legacy system, you were limited by the manual effort required to interact with it. Now, you can offer automation as a service for these legacy systems. You can build, deploy, and manage these agents for your clients, creating a high-margin revenue stream that is based on the value you provide, not the hours you bill.
You are no longer selling your time; you are selling the efficiency of your automated systems. This is the key to scaling a one-person agency. You can handle more clients, take on more complex projects, and deliver better results, all while keeping your overhead low.
Why should you build your own data pipelines?
Building your own data pipelines is the only way to ensure your agentic workflows remain stable and under your control. When you rely on third-party tools to handle your data integration, you are at the mercy of their uptime and their API limitations. By constructing your own pipelines, you create a proprietary asset that connects your inputs to your AI agents, giving you a competitive advantage that is difficult for other agencies to replicate or disrupt.
A data pipeline is simply the infrastructure that moves data from your source to your agent. If you use a pre-built tool, you are limited by what that tool can do. If you build your own, you can customize the flow to meet your specific needs. You can add validation steps, error handling, and logging that are tailored to your business.
This is not as hard as it sounds. With modern tools, you can build robust pipelines using low-code or even no-code platforms. The key is to own the logic, not just the connection. When you own the pipeline, you own the workflow, and that is where the real value lies.
How do you audit your own legacy workflows?
You should audit your legacy workflows by identifying every task that involves repetitive clicking in a browser or desktop application. Look for processes where you or your client spend more than thirty minutes a day navigating through screens to input data or retrieve information. These are the prime candidates for agentic automation. Map these workflows step-by-step, noting exactly where human judgment is required and where the process is purely mechanical and predictable.
Start by creating a spreadsheet of all your daily tasks. For each task, estimate the time spent, the frequency, and the level of complexity. Then, rank them by automation potential. The best candidates are high-frequency, low-complexity tasks that involve a lot of screen navigation.
Once you have identified these, document them in detail. Take screenshots, record your screen, and write down the exact sequence of steps. This documentation will be the foundation for your agentic workflow. It will help you see where the bottlenecks are and where an agent can actually help.
What is the trap of over-automating?
The trap of over-automating occurs when you try to force an AI agent to handle tasks that require nuance, empathy, or high-level strategic decision-making. If you automate the wrong parts of a workflow, you risk losing the human touch that clients pay you for. Always maintain a clear boundary between the tasks that are purely transactional and those that are relational. If an agent makes a mistake in a transactional task, it is a bug; if it makes a mistake in a relationship, it is a business failure.
The goal is to automate the process, not the relationship. Use agents to handle the data entry, the research, and the scheduling. But keep the communication, the strategy, and the decision-making in your own hands.
Your clients hire you for your expertise and your judgment, not for your ability to do data entry. If you automate away the parts of your work that provide the most value, you are not scaling; you are commoditizing yourself.
How do you start this week?
Start this week by selecting one repetitive task in your business that relies on a legacy portal or a slow, manual interface. Document the exact sequence of clicks and keystrokes required to complete that task. Once you have the sequence, research the available agentic frameworks that can interact with desktop interfaces. Build a simple prototype that handles just the first three steps of the process. Test it for reliability, refine the prompts, and then expand the scope.
This is a five-day plan. Day 1: Identify the task. Day 2: Document the workflow. Day 3: Research tools. Day 4: Build the prototype. Day 5: Test and refine. Do not try to automate the entire workflow at once.
Start small, prove the concept, and then build on your success. This is how you build a robust, scalable agentic system that actually delivers value. Focus on the low-hanging fruit first, and you will see the results in your efficiency and your bottom line.
Sources
OpenAI — Shield helps technicians hand off legacy portal work with OpenAI — https://openai.com/business/customer-stories/shield