Aditya Agrawal, Co-Founder at Creed.gettyFor the past several years, the AI industry has been focused on one question: How do we build smarter models?Larger context windows, better reasoning and higher benchmark scores have dominated the conversation. But after building a voice-first AI companion, I've become convinced that the next major competitive advantage won't come from intelligence alone. It will come from how people interact with AI.When we first introduced voice at Creed, we assumed users would simply ask the same questions they were typing, only faster.Instead, people thought out loud. They paused, changed topics mid-sentence and shared personal details they had never typed before. Conversations became less transactional and more reflective, fundamentally changing how I think about the future of AI.Voice isn't simply another interface. I believe it will become the default interface for AI.We Think In Conversation, Not PromptsMost software forces us to translate our thoughts into structured inputs. We type concise prompts, edit our wording and decide exactly what we want before pressing enter.That's not how people naturally communicate.When people speak out loud, they process ideas in real time, interrupt themselves, reconsider answers and connect unrelated thoughts—all during a single conversation. That richer interaction can give a system context that simply does not exist in text.For founders building AI products, this is an important design principle. Voice shouldn't be treated as a feature layered onto an existing chatbot. If your product depends on coaching, education, healthcare, recruiting or customer support, conversation may be the product, not just the interface.Memory Will Become More Valuable Than IntelligenceIn my experience, users care just as much about whether an AI remembers them as whether it answers perfectly.People don't want to repeat their goals every day. They expect the system to remember recurring struggles, communication preferences and previous conversations. The more context the AI retains, the more natural every future interaction becomes.As foundation models improve, intelligence is likely to become increasingly commoditized. Context won't.​For those building the next generation of AI models, this means thinking about memory as a product decision, not just an engineering feature. Instead of asking what your AI should remember today, ask what it should remember six months from now that makes today's conversation more valuable.Product Experience Beats Benchmark ScoresThe AI industry loves benchmarks. Latency, reasoning scores and evaluation metrics all matter, but building a consumer voice product taught me that users often judge AI very differently than engineers do.People rarely say, "This model scored higher on a benchmark." They say, "This felt easy."That's an important lesson beyond voice AI.Founders optimize what's measurable. Users optimize for how a product feels. As models converge, product experience—not benchmark performance—could become the real competitive advantage.Voice Enables Proactive AIToday's AI assistants are largely reactive. You ask a question, they answer it, and the interaction ends until the next prompt.Voice changes that equation because continuous conversations create continuous context. As users naturally share more information over time, AI systems become capable of recognizing patterns instead of simply responding to isolated requests.A language-learning assistant could notice declining consistency and encourage someone before they abandon their goal. A career coach could recognize recurring interview anxiety and proactively suggest preparation exercises. A health application could detect changing routines and recommend adjustments before small problems become larger ones.None of these experiences require artificial general intelligence. They require persistent memory combined with natural conversation.This represents a shift in product thinking. The goal isn't simply answering questions faster, but designing systems that become more useful because they understand users over time.Optimize For Outcomes, Not EngagementTraditional software rewards companies for maximizing session length, daily active users and time spent in the product, but I am believing more and more that those metrics are incomplete for AI.If someone spends an hour talking to an assistant, but never changes anything in their life, has the product actually succeeded?The most meaningful AI products should create value beyond the conversation itself.Did the user complete an important task?Did they build a better habit?Did they make a more informed decision?Did they accomplish something they otherwise wouldn't have?Those are harder metrics to measure, but they're ultimately the ones that matter. As we build and improve additional AI platforms, this will require resisting the temptation to optimize purely for engagement and building products that actually help users move forward.To ensure you're optimizing for outcomes, start by measuring whether AI helps users accomplish what they came to do, rather than how long they stay engaged. Depending on the product, that could mean task completion, time saved, resolution rates, repeat success or reduced human intervention. The key is to identify the real-world job the product is meant to accomplish and measure how reliably and efficiently the AI completes it. With AI, a shorter interaction can often signal a better product.​The Next Interface For ComputingEvery major era of computing has been defined by its primary interface. The personal computer was built around the keyboard and mouse. The smartphone was built around touch. I believe AI will increasingly be built around conversation—not because speaking is faster than typing, but because conversation allows AI to understand context, build long-term memory and develop relationships in ways previous interfaces never could.​​Simply building the smartest AI won't be enough. Rather, the most successful AI will be the one that understands people, remembers what matters and fits naturally into everyday life. ​​The voice revolution isn't about replacing the keyboard. It's about changing the relationship between humans and computers altogether.​Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?