Today’s conversation features Shishir Mehrotra, CEO of Superhuman—previously known as Grammarly, which remains its flagship product. Shishir also served as YouTube’s chief product officer and sits on Spotify’s board. This interview was arranged with an initial focus on AI’s impact on software, platforms, and creativity but soon centered on a controversial Grammarly feature involving AI-generated writing suggestions attributed to real experts without permission. Grammarly released in August a feature called Expert Review, which offered AI-generated writing advice inspired by experts including journalists and public figures without their consent, sparking backlash and a class-action lawsuit led by Julia Angwin. Superhuman responded by first instituting an opt-out email process and then discontinuing the feature altogether, with Shishir publicly apologizing. Despite the tension and disagreement over AI’s extractive nature, Shishir participated in this candid Decoder interview to address these issues. Superhuman now positions itself as an AI-native productivity suite integrating AI assistance into everyday work across multiple apps—Grammarly, Coda, Mail, and a new platform, Superhuman Go, which allows anyone to build AI agents that operate seamlessly alongside user workflows and familiar software. The company’s emphasis is on AI blending unobtrusively into work rather than requiring behavior changes, reaching 40 million daily active users with an integrated experience across diverse digital environments. When asked about the Expert Review feature controversy—using real experts’ names without consent—Shishir acknowledged it was a “bad feature, ” misaligned with their strategy and poorly received by both users and experts. It was developed by a small team trying to meet user desires for expert feedback and to offer experts ongoing connections with their audiences, but it failed to deliver value or respect attribution properly. Shishir emphasized that impersonation was never intended, that each generated suggestion clearly attributed the inspiration source with links, and that while the claims in the lawsuit are argued to be without merit, the company chose to discontinue the feature early on, before the lawsuit, because it conflicted with their goals. Regarding compensation and attribution, Shishir stressed that proper attribution is fundamental but distinct from impersonation, which he rejects. He envisions a platform where creators can control how their expertise is used, build AI “agents” embodying their style or guidance, and monetize their contributions through revenue-sharing models akin to YouTube’s 70/30 split.

However, this system was not in place when Expert Review launched, which partly explains why it relied on unauthorized use. The interview also touched on legal nuances around name and likeness rights, commercial use, and the evolving challenges AI poses to copyright law. Shishir compared the situation to YouTube’s history with copyright controversies like the Viacom lawsuit and Content ID system, stressing that legal frameworks often lag behind technological innovation. He argued the industry should aim to exceed minimal legal standards to fairly support creators, emphasizing connection and monetization over mere attribution. Addressing public distrust and fear of AI—citing polls showing AI’s poor perception compared to even unpopular institutions—Shishir attributed this mainly to fears of job displacement rather than creator-specific concerns. He believes AI will create more jobs than it destroys by augmenting human work, positioning Superhuman’s tools as empowering users to become “superhuman. ” He acknowledged creators face mounting challenges as AI changes content distribution and monetization but views new platform paradigms as opportunities rather than threats. On the broader software industry and “SaaSpocalypse” fears, Shishir argued that network effects, integrated ecosystems, and consistent user experiences prevent easy disruption—even though building software is easier and AI tokens cost money. He does not believe major AI providers will simply replicate and kill products like Grammarly, noting that sustained innovation and value creation are crucial. Concerning the future of monetization and creator economies, Shishir highlighted ongoing shifts from advertising-based models to subscriptions and direct product sales, with creators leveraging deep connections to generate sustainable income. He envisions AI agents enabling creators to scale personalized assistance and engagement to audiences willing to pay for that value, although not all creators will adopt this path. The quality of these AI-driven experiences depends heavily on the creators’ effort in defining and training their agents. When discussing whether AI can replicate the nuanced creativity and editing style of a human editor like Nilay Patel, Shishir admitted challenges remain but expressed optimism based on Grammarly’s success and early agent deployments, such as sales or teaching assistance. He sees specific expert niches thriving initially where guidance follows clearer rules or can be effectively modeled. Finally, Shishir previewed upcoming expansions of Superhuman Go and reiterated commitment to building a platform where users and creators collaborate with AI, maintaining ethical standards and creating new economic models that empower creators rather than exploit them. Overall, this in-depth interview candidly explores the challenges and opportunities at the intersection of AI, content creation, attribution, and business models, from the vantage point of a CEO navigating recent controversies while shaping the future of AI-powered productivity tools. —Decoder with Nilay Patel