--- title: "If Your Company Started Using OpenClaw Today, Where Should It Begin?" author: "Unique Research" sourcePublication: "Unique Research Substack" originalPublishedAt: "2026-04-03T10:02:22+00:00" canonical: "https://ffcap.cn/en/research/src-20260403-01html" source: "https://uniqueresearch.substack.com/p/src-20260403-01html" language: "en" --- # If Your Company Started Using OpenClaw Today, Where Should It Begin? _Original · Unique Research · 2026-04-03 · Shanghai_ _Editor's note: This is a historical workshop recap, not a deployment guide or a claim that the proposed workflows were implemented. The opening survey concerns attendees in this room; its approximate 60–70% figure has no supplied sample size or survey method and is not an industry adoption estimate. The four groups' proposals are discussion outcomes, not measured efficiency, marketing, inventory-forecast or investment results. “Lobsters” are the source's nickname for OpenClaw-based digital employees. “Blue team” preserves the Chinese source label for a challenger in project review, not a cybersecurity assignment. The author retains the final decision with people. The source supplies no precise workshop date; April 3 is the article's publication date. Wotu is a provisional rendering of the Chinese organizer name. Twenty-one source images still require content review._ WORKSHOP A Group of Cross-Border Business Practitioners Talked Behind Closed Doors for More Than Two Hours and Found That AI's Real Challenge Is Not Knowing How to Use It, but Deciding Which Business Function to Change First Companies' Relationship with AI Moves from “Knowing It Matters” to “Getting Ready to Act” “ The real difficulty with AI is not whether you know how to use it, or whether a tool is popular, but which problem it should help you solve first. Over the past six months, many companies have been talking about AI. Some use it to write copy, others to generate images or code. Some have also started seriously “raising lobsters”—using OpenClaw as a digital employee that stays online over the long term. But inside a business, the question quickly changes. It is not “is AI impressive?” or “is OpenClaw popular?” It is something more practical: Which problem should it help me solve first? At a closed-door workshop during “Going Global · Coexistence: The Fourth Global Brand Expansion Summit,” hosted by Wotu and WotoHub, Unique Research spent an afternoon working through that question with companies in cross-border e-commerce, international brand expansion, supply chains and content-led growth. The most interesting thing about this workshop was not another explanation of OpenClaw. It pushed people who had not really started using it toward a more concrete question: If you were to start using AI today, where should your company make its first move? The Most Telling Scene: Many Knew of OpenClaw, but Few Had Really Used It The workshop began with a simple survey of the room. The result was straightforward: roughly 60–70% of attendees had not fully explored or deeply used OpenClaw. Only a very small number had actually installed it and were running more than three “little lobsters.” That says a great deal. It is not that companies do not care about AI. Quite the opposite: everyone knows it matters and has a vague sense that waiting much longer may mean falling behind. But most companies still understand it at the level of “the tools are good,” “the trend is big” and “I'd like to try it.” Once they get to actual business operations, many get stuck: Can it handle customer service? Can it run content production? Can it calculate inventory needs? Can it connect operations? Can the boss use it to support decisions? Mr. Wu's opening presentation of more than thirty minutes was therefore necessary groundwork. The point was not to explain OpenClaw exhaustively, but to give people who had not used it a shared understanding: what it is, why so many people are discussing it, and how it differs from tools such as ChatGPT, Manus and Claude Code. But that was only the opening. The real focus came afterward. A Workshop's Value Is Not Explaining a Tool, but Making the Question Specific In the second half of the closed-door session, we did not keep explaining concepts. Instead, we split participants into four groups and asked them to discuss problems actually occurring in their own companies. The rules were clear: No vague problems, no problems belonging to someone else's company, and no general statements such as “I want to learn AI.” Each group had to start from its own business and reduce its problem to one sentence. Ideally, it would take a form such as: “How should I…” or “How can we use AI to solve…” The atmosphere in the room changed completely. People stopped discussing how “AI can do many things” and began asking: What, exactly, is the most worthwhile thing for AI to take on first in my company? It sounds simple, but it is crucial. For many companies learning AI today, the biggest problem is not unfamiliarity with tools. It is not yet knowing how to translate vague anxiety into an actionable question. That was the translation this workshop performed. Four Groups Put Almost All the Typical Enterprise AI Challenges on the Table The discussion covered four directions: data, decision-making, marketing operations and efficiency. What looked like a grouping exercise actually laid out four of the most typical anxieties cross-border businesses face with AI today. The First Group Discussed Efficiency Its question was: How can AI connect workflows across the entire cross-border e-commerce value chain? That question resonated with many people. Many companies are not entirely without AI. They use a little everywhere: someone generates copy, someone makes images, someone has installed two lobsters, and one department has built a small tool. But the company as a whole is not truly connected. The result is that AI appears widely used while actual efficiency gains remain scattered. The group's eventual direction was practical: do not begin by trying to build a vast, all-encompassing system. Start with a clearly defined role—for example, a “product lobster,” an “operations lobster” or a “design lobster.” Get one link working before replicating it in other departments. That conclusion is not flashy, but the author considers it right. What most companies truly lack is not imagination, but an executable starting point. The Second Group Discussed Marketing and Content Its question was: How can agents and AIGC automatically generate high-quality content to address social-platform traffic acquisition and creative-asset production? Behind that question lies very real pressure for cross-border businesses: more accounts, more platforms and a continual need for updates. Videos, images, scripts and publishing schedules all have to keep up. Reality often looks different. Assets go wrong, a Logo gets distorted, objects clip through one another in video, visual quality varies, and tools are scattered across platforms. Teams look busy every day, yet have not built reliable content-production capacity. Some companies at the workshop said they managed multiple brands and social accounts and needed to produce at least dozens of videos each week. At that scale, simply “getting AI to help” is no longer enough. Production must become process-driven. The group therefore converged on this direction: do not just find a few more models. Build an agent workflow that is as complete as possible, connecting scripts, images, video and publishing into a more stable production chain. Put plainly, participants began to realize: the content problem is fundamentally no longer a creativity problem, but a capacity problem. The Third Group Discussed Data and Inventory Planning Its question was: How can everyday operational and sales data support more reasonable inventory forecasts? This was an especially good question because it was not remotely “flashy.” It did not resemble the AI scenarios people commonly imagine: not a cool demo or an automatically written article, but a direct encounter with one of the hardest parts of cross-border business—inventory, supply chains and forecast accuracy. Many SKU entries, many variables, unreliable historical formulas and inventory planning still heavily dependent on experience: these are familiar, longstanding problems for many companies. The group's conclusion was representative too: do not rush to have AI produce forecasts directly. First build the data foundation. Bring scattered data together and organize Feishu, multidimensional tables, sales data and operational data clearly. Then use OpenClaw to collect data across more dimensions, and hand it to models for assisted analysis. This was something the workshop repeatedly reinforced in the author's account: when AI cannot be put to work, it is often not because the model is weak, but because the company's own data is not ready. The Fourth Group Discussed Decision-Making Its question was: How can AI improve the precision of decisions about initiating new-product projects? This group discussed its question longest, which was entirely understandable. The earlier groups mainly addressed execution. This one reached what business leaders really care about: Is this project actually worth doing? Should we invest? What information should inform my judgment? Why did so many earlier decisions go wrong? They did not arrive at a simplistic answer such as “let AI make the boss's decisions.” Instead, a more mature approach emerged: first let AI fill information gaps, monitor projects, record processes and assist reviews. It could even act as a “blue team”—a challenger helping the team see the project-initiation process more clearly—rather than replacing the final decision-maker. That is important. It suggests more companies are moving from asking “can AI replace people?” toward “can AI first make our judgments better grounded?” The Greatest Value Was Not That Every Group Had an Answer, but That People Began Asking the Right Questions On the surface, this workshop might look like four groups discussing four specific problems. What mattered more was that many participating companies realized, for the first time: one of the scarcest capabilities in the AI era may not be operating tools, but asking the right questions. The shift is subtle, but real. Previously, learning AI in many companies would drift toward “watch how others do it,” “which software is popular?” or “is there a ready-made template?” That approach is not useless, but it easily remains at the level of watching from the sidelines. This workshop progressed differently. It pushed everyone to begin with their own role, category and company and identify the most concrete question. Once a question becomes specific, things begin to move. Someone also noted toward the end that AI is not just a tool, but a methodology. It pushes people to relearn how to ask questions, communicate, break down problems and view answers critically. That may not sound exciting, but it directly addresses many companies' pain points today. Most do not lack an awareness that “AI is powerful.” They have not yet translated that into “what should our company change first?” So What Did This Workshop Ultimately Leave Behind? Not a standard answer. Nor a promotional poster declaring “OpenClaw can do everything.” It was more a set of practical points of agreement. First, do not rush to talk about going all in. Find the most suitable starting scenario. High-frequency, repetitive processes with clear rules, which do not immediately touch the most critical risks, often make the best starting points. Second, data foundations matter much more than people imagine. Many companies want AI, but their underlying data has not been organized, classified or tagged. They then find that even a strong model cannot do much with it. Third, documenting SOP and skills is crucial. If a company's accumulated experience, standards, style and processes are not written down and turned into skills, AI struggles to enter actual operations. Fourth, do not glorify any one tool. OpenClaw was the workshop's central tool, yet participants ultimately became clearer about something else: not every task needs a lobster. The key is to choose the most useful combination of tools for the task and the company's situation. The workshop did not make AI more mysterious. It made it more concrete. Rather than leaving people thinking “this is far removed from me,” it pushed them to confront a direct question: If AI really has arrived, what will your company let it take on first? Once that question receives serious discussion, much has already changed. From that moment, the relationship between companies and AI is no longer merely “knowing it matters.” It is getting ready to act. --- Original publication: https://uniqueresearch.substack.com/p/src-20260403-01html On-site reading page: https://ffcap.cn/en/research/src-20260403-01html