> **来源：[研报客](https://pc.yanbaoke.cn)** # 2026 AI Shortlist # 3 Trends Shaping the Foundation of AI-Native Productivity # The big picture We're at an inflection point with AI at work. The excitement is real, the investments are massive, and the promises are bold. Yet for most organizations, the results aren't matching the hype. Leaders keep hearing that AI will revolutionize productivity, but what they're seeing on the ground tells a different story: People feel busier, not more efficient. Workflows are more fragmented, not more streamlined. And the quality of work often reflects the quality of the data powering it. Here's the thing: # This isn't an AI problem. It's a context problem, an integration problem, and ultimately, a workflow problem. Most companies are treating AI like a shiny new app, something to bolt onto existing processes and hope for the best. They're running pilots, tracking adoption metrics, and wondering why a significant number of their AI initiatives are delivering little return. Meanwhile, employees are quietly using AI far more than leadership realizes, but in ways that create more work downstream rather than actually moving the needle. The organizations that are succeeding with AI aren't just using more of it. They're using it differently. They're embedding it into the natural flow of work, with the context it needs to be genuinely useful. They're rethinking decades-old workflows that were never designed for intelligent collaboration. And they're moving from a focus on productivity alone to thinking bigger picture: strategically leveraging AI to create better experiences all the way around. That means they are shifting from asking "How can AI help us work faster?" to "How can we design work differently if AI were part of the team from day one?" This report explores three critical trends that separate AI activity from AI impact. These aren't predictions about some distant future; they're shifts happening right now that will define which organizations thrive in an AI-native world and which ones drown in a sea of generic, low-quality content. The gap between the two is widening fast. # Shelly Kramer Founder and Principal Analyst at Kramer&Co # Meet the experts behind the insights These trends were shaped by conversations with the founders of companies at the forefront of AI innovation and an industry analyst who keeps a pulse on what's happening now and what's coming next. Together, they offer a view into how AI is evolving from concept to capability, and how organizations can prepare for what's ahead. # Shishir Mehrotra CEO at Superhuman Founder of Coda # Max Lytvyn Co-founder and Board Member at Grammarly # Rahul Vohra Founder and CEO at Superhuman Mail # Trend 1 # Context will fix the Al productivity paradox # Status check Leaders are looking to productivity as the most immediate return on their AI investments. While there are pockets of progress, the impact hasn't scaled. In some cases, the opposite is true. This is the Al productivity paradox: Leaders expect Al to accelerate performance, yet people often feel busier, workflows are more fragmented, and the quality of output declines. AI at work today is both overused and underused. Many people rely on AI for quick wins like summarizing, polishing, or generating slides from text. And those use cases generally work well. The problem is that they're not the most valuable applications of AI, and they often create more content for others to consume—content that someone else will later use AI to summarize to digest. Meanwhile, higher-value use cases often fail because AI lacks the necessary context to move work forward, or the prompting isn't skillful enough to deliver the right results. Over time, the need for elaborate prompting will diminish, but the importance of giving AI the right context will only grow. Consider a familiar pattern: Someone asks AI to "turn these bullets into a proposal," then sends that output to their team. The next person asks AI to "summarize the key points" so they can skim the information. Both steps might feel efficient, but they don't actually move the work forward. The AI-generated expansion adds volume without adding clarity. The summary compresses it back down, introducing even more signal loss along the way. The result is more content, more steps, and more work for everyone involved. These kinds of "quick wins" aren't always wins. When AI is used simply because it's easy, it often adds friction instead of value. This imbalance has left most workplaces stuck in the messy middle. AI is in the mix, but not yet driving impact. People are using AI to make work faster, not better. To move work forward, leaders need to rethink how their teams use AI, from a tool that produces more to a partner that understands more. # Forecast If people keep using AI only for quick wins, workplaces will face a new kind of productivity crisis. Words, documents, and messages will multiply across already overloaded channels, but the content inside them will become increasingly hollow. People will find the knowledge base article they were looking for, only to realize it's beautifully formatted fluff. The more AI fills our systems with low-quality content, the harder it becomes to find the information that actually matters. The future of productivity and the foundation of AI-native work depends on AI that truly understands the work it supports. Al must be built into the flow of work, with access to the organization's knowledge bases, data, documents, and project trackers so that it understands its goals, priorities, and audiences. When Al carries that context forward, it can move beyond one-off task support to offer real partnership, helping people analyze, strategize, think more deeply, and make creative, well-informed decisions. Right now, people must manually provide AI with context by defining tasks and goals, supplying background information, and providing the necessary nuance for accurate results. Doing this well requires translating organizational goals and context into clear direction for AI. It's a shift from prompt engineering to goal engineering, where people focus on intent, outcomes, and constraints to get higher-quality results. But in the near future, Al-native tools that are deeply connected across tools and workflows will start to relieve this burden by bringing context to the person rather than the other way around. These AI systems will already understand the organization, remember the project, know what information matters most, and proactively offer support instead of waiting to be asked. When that happens, AI stops crowding the workplace with more content and starts being used for higher-value work, such as deeper thinking, sharper communication, greater creativity, and better decision-making. The result isn't just faster output; it's smarter, more impactful work that moves the whole organization forward. # Action items To solve for the AI productivity paradox, organizations need to focus on context, not just capability. This means equipping people with the right training and tools to create higher-quality, more meaningful work. Find the gaps where AI is underused. Help people move beyond surface-level tasks and identify where AI could add strategic value by supporting innovation, critical thinking, and creative problem-solving. Train people to guide AI effectively. Help teams build goal-engineering skills. Train them on how to define problems clearly, provide relevant context, and articulate intent, desired outcomes, and constraints. AI's value depends on the quality of the information it's given and the clarity of the goals it's working toward. Invest in AI that works with your organizational knowledge and context. Choose systems designed to integrate with your company's data, tools, and workflows so outputs are factual, relevant, and aligned with real work. Generic AI doesn't just produce generic results; it often produces wrong ones. □ Refocus productivity goals. Measure success not by the volume of output but by accomplishing real business goals. True productivity means less noise, clearer thinking, and a job well done. When context and clarity meet smarter, native tools, AI stops contributing to the productivity paradox and starts solving it. # Trend 2 # AI impact will follow # AI integration # Status check While AI pilots are skyrocketing across organizations, the return on those investments remains underwhelming. According to MIT NANDA's State of AI in Business 2025 study, $95\%$ of organizations are getting zero return from their AI pilots. The $5\%$ that do succeed share one defining trait: They integrate AI deeply into high-value workflows. That's where most companies are falling short. It's not that people aren't using AI; if anything, they're using it far more than leaders realize. McKinsey's Superagency in the Workplace report found that employees use AI tools nearly three times more frequently than their managers think. The problem is where and how that AI use happens. Most employees rely on generic, consumer-grade chatbots that live outside their normal workflows. They have to switch tabs, copy and paste context, and manually "ask" for help. This break in workflow is the first productivity killer. The second is that these standalone tools have no understanding of what you're working on. They're disconnected from your systems, your data, and your goals. Without access to the right context—whether that's a customer note in Salesforce, a relevant Slack message, or a previous project document—AI can't deliver guidance that's tailored, accurate, or actionable. The result is a lot of AI activity, but very little impact. AI that drives real productivity looks different. It is embedded into the flow of work, where it can draw from real context and act in sync with the tools people already use. When AI works where people already work, it eliminates friction and compounds productivity instead of fragmenting it. We've seen this pattern before. Superhuman Mail was built on the recognition that email was the biggest productivity problem hiding in plain sight. We spend more time in email than in any other work app. Despite that investment, we often reply late or miss messages entirely, which slows down deals, deadlines, and decisions. The issue wasn't effort; it was friction. Email hadn't evolved in decades, and professionals were losing hours every day to context switching and overload. We solved that by rebuilding email to eliminate the time and focus tax, not by asking people to change how they worked. The same dynamic is playing out with AI today. AI isn't struggling to drive results because it lacks enthusiasm or potential; it's because it's sitting outside of how people actually work. # Forecast The companies that will see measurable impact from Al aren't the ones just mandating more Al use. They're the ones integrating it into the natural flow of work, meeting people where they already work instead of asking them to work around the Al. The biggest friction today is that employees have to remember to use AI, and know how to use it correctly. They have to choose which chatbot to go to, craft a good prompt, and then move the output back into their real workflow. This "AI detour" model limits impact. The first wave of value will come from AI-native productivity platforms that are ubiquitous, proactive, and connected. Real productivity will come from AI that works everywhere you do, anticipates needs without being asked, and understands context across your data and systems. Ubiquity is important because it eliminates the "AI detour" model by integrating AI into where employees already work. But ubiquity is just the beginning. The real transformation will come from proactive AI: agents that help without being asked. Instead of waiting for prompts, AI will know what you need and when you need it, drawing on signals from your tools, habits, and preferences. Imagine getting a Slack message from your manager asking you to schedule a QBR. Before you can even open your calendar, your Al surfaces available times, recent sales metrics from Salesforce, and last quarter's deck. Or imagine opening your inbox in the morning to find every email you've received already paired with a draft reply, some even sent automatically once you've built enough trust in the system. It's the same principle that made Grammarly so powerful: Instead of waiting for you to ask for help, Grammarly proactively improves what's in front of you. Extending that principle across every workflow through ubiquitous, proactive, integrated AI-native workflows will transform isolated AI activity into meaningful, measurable impact. # Action items Turning zero-return AI pilots into AI programs with measurable impact requires more than enthusiasm; it demands thoughtful AI integration. Leaders should focus on reducing friction, building trust, and integrating AI where work already happens. Build toward ubiquity. Don't give employees more tools to visit; instead, bring AI to where they already work. Identify how they use generic chatbots today and integrate that value into existing systems. Choose AI-native tools that reduce context switching and keep teams in flow. Prioritize proactivity. Al's value shouldn't depend on how well someone can write a prompt. Choose tools that guide employees, where prompts enhance the experience rather than unlock it, and surface help when it's needed. Connect the context. Al shouldn't create new silos but break down old ones. Link tools so Al can draw from shared systems like CRMs and project trackers and deliver guidance rooted in real organizational knowledge. Redefine AI readiness. Go beyond prompt training. Equip employees with AI-native tools embedded in their daily workflows and teach them how to collaborate with these systems naturally and impactfully. Integration is where AI's promise becomes performance. Once AI lives inside the flow of work, it lays the groundwork for rethinking the very workflows that power modern organizations. # Trend 3 # Legacy workflows must be rebuilt for Al-native work # Status check Most of today's workflows were built for a pre-AI world. Decades of digital transformation added layers of tools and automation, but the structure of work itself hasn't fundamentally changed. Ideas still move from spark to execution through long, linear paths. People exchange drafts, decks, and discussions until abstract ideas finally crystallize into something tangible. Take a typical launch of a product, website, or campaign. Planning starts with meetings that turn into decks and docs, but real feedback doesn't happen until people finally see the thing itself. That's when priorities shift and the real thinking begins. Al collapses that delay. By generating early prototypes or concepts in minutes, it helps teams get to tangible work faster, spark sharper feedback sooner, and move from idea to impact in a fraction of the time. Our current systems weren't built for intelligent collaboration between people and technology. Most companies are still treating AI as an assistant, not a collaborator. The result is incremental efficiency, not transformative change. This pattern mirrors the earliest days of digital transformation, when organizations digitized paperwork rather than rethinking processes for the digital world. We're now doing the same with AI, bolting it onto existing workflows instead of redesigning those workflows from the ground up. The gap between what AI can do and what our systems allow it to do keeps widening. To capture the true value of AI, we need to pair it with new ways of working. That means rethinking the very architecture of work: how ideas grow, how people and technology share responsibility, and how the tools we use can adapt in real time to both human judgment and machine intelligence. # Forecast To move beyond marginal productivity gains, organizations need to re-architect workflows around what AI and people each do best, both separately and together. This means moving from retrofitting to redesign. Leaders should not just ask how AI can fit into current processes. Instead, ask how we would build those processes differently if AI were a teammate from the start. Al-native workflows start from the assumption that generation, summarization, and analysis can happen instantly. Those AI-generated artifacts aren't finished products, but they're ready to transform a blue-sky brainstorm into a feedback session or proof-of-concept review within minutes. This bridges the gap between abstraction and execution, allowing people to focus on what machines can't: judgment, creativity, and lived experience. In these workflows, collaboration will take shape around concrete artifacts, not abstract conversations. Instead of endless meetings and email threads, AI will help teams to generate something tangible early that everyone can respond to and refine together. Work will go from $0\%$ to $80\%$ almost instantly, allowing teams to focus their time and energy on the final $20\%$ that makes it uniquely personable and differentiated. This shift will also demand new work surfaces: the documents, slides, and emails we rely on today were designed for a pre-AI era. In an AI-native world, these static formats give way to dynamic, connected environments where people and AI agents will co-create in real time. These workspaces will integrate thinking, doing, and communicating across tools and data, enabling AI to act with full context while keeping people in control. When we redesign workflows from the ground up, the collaboration between people and AI agents becomes not just faster but fundamentally better. Ideas move more fluidly from concept to creation. Teams spend less time translating information and more time applying insight. The organizations that embrace this change will gain not just speed but also adaptability, which will be the true competitive advantage of the AI-native era. # Action items To build truly AI-native workflows, leaders must do more than adopt technology with AI bolted on. They need to seek opportunities for AI to support how teams naturally work, enabling people, data, and AI agents to collaborate seamlessly across systems and surfaces. Audit for friction. First, map out your organization's highest-volume workflows, not necessarily the most strategic ones but the ones that consume the most time. Then ask where AI generation, analysis, or summarization could make steps faster or smarter. Audit for abstraction. Identify where teams spend more time talking about work than doing it. Strategy decks, status reports, and planning docs are prime areas to reimagine with AI-generated prototypes, summaries, or live simulations. Explore new collaboration surfaces. Evaluate the tools your teams use most. Which ones force rigid, manual workflows, and which allow AI to flow naturally between thinking, doing, and communicating? Start from zero. Forget how a process works today. Ask: If AI were a teammate, what would this workflow look like? What steps would disappear, and what new ones would emerge? Once teams learn to rethink the architecture of work itself, they can turn AI from a bolt-on addition into a true collaborator. That shift accelerates progress, amplifies insight, and transforms how ideas move through the organization. # How Superhuman can help As organizations put these trends into practice, Superhuman's Al-native productivity suite provides the foundation to make it happen. Superhuman brings AI everywhere you work, however you work. It gives you a team of agents and a set of apps that proactively offer support in your flow of work and are connected to all the data and context required to make you feel superhuman. # Go: Your AI collaborators that already know your context Go works everywhere you work, understanding what you're trying to accomplish and proactively offering help. It learns your preferences and understands your project context to anticipate what you need next, no more explaining background or copying between apps. Go orchestrates specialized AI agents that can act directly in your documents, draft in your voice, and handle the tasks that interrupt your flow. # Grammarly: The agent that integrates across every surface Grammarly brings intelligent communication into the flow of work. Embedded across more than 500,000 apps and websites, it helps professionals and teams write with precision, consistency, and confidence. Grammarly's proactive, context-aware AI enhances clarity, adapts to your brand voice, and ensures every message is concise, aligned, and impactful. # Coda: Dynamic surfaces for Al-native workflows Coda redefines the document as a living workspace where people and AI work together in real time. It connects data, content, and actions across tools, turning planning docs, status reports, and projects into dynamic, interactive surfaces. Coda helps teams move from conversation to creation instantly, transforming how ideas become impact. # Mail: The most productive Al-native email app ever made Mail reimagines the inbox for the AI era. Imagine waking up to an inbox where every email already has a draft reply. You would simply edit, then send. Eventually, you would not have to edit at all. With Mail, your team can get through email twice as fast, respond faster to what matters most, and save four hours or more every single week. # About Grammarly Grammarly is the trusted AI assistant for communication and productivity, helping over 40 million people, 50,000 organizations, and 3,000 educational institutions do their best work. Companies like Atlassian, Databricks, and Zoom rely on Grammarly to brainstorm, compose, and enhance communication that moves work forward. Grammarly works where you work, integrating seamlessly with over 1 million applications and websites. Grammarly is a part of the Superhuman suite of apps and agents that brings AI wherever people work. Learn more at superhuman.com.