We are at the cusp of a tectonic shift in how technology is
built, used and experienced. As large language models (LLMs) become widely
available—often open-source, commoditized and deeply integrated—intelligence
itself is becoming a utility. It is no longer the algorithm or model that gives
you an edge. It’s the experience you build around it.
Intelligence as infrastructure
Over the past few years,
billions of dollars and immense compute power have gone into training LLMs
using vast swathes of public data. But the frontier of artificial intelligence (AI) is shifting. The next
breakthroughs driven by synthetic data and massive inference workloads will be
expensive, making foundational model training the domain of a few global
players.
Fewer teams
will focus on foundational model training; the majority will innovate through
post-training-building optimizations, workflows and intelligence layers on
top. In this new era, LLMs function more like a utility—fast, cheap and
universally available.
What makes this shift even more profound is that
interacting with this technology is remarkably simple. You donʼt need to code
or configure—you just speak, in your own language. And the model does the rest.
As execution becomes commoditized, the real differentiator is no longer how you
solve problems but which problems you choose to solve, when you choose to
solve them and whether you have the clarity and conviction to focus where it
truly matters.
When moats disappear
Customer experience has always been important. But in the past,
it shared the stage with other defensible moats: proprietary technology,
execution speed, distribution networks, economies of scale. Companies could
dominate through a combination of IP, funding, partnerships or operational
excellence, even if their user experience (UX) wasnʼt best-in-class.
But in the era of LLMs, that equation has fundamentally changed.
Technology is no longer a moat—foundation models are widely available and
commoditized, offering the same baseline to all. Execution, too, has been
flattened by AI, which now automates everything from code and content to design
and marketing, allowing even lean teams to move at pace. Distribution is also
losing its edge, as AI-native interfaces like chat and voice are inherently
viral and platform-agnostic.
This leaves one sustainable advantage: experience. And it’s not
just about usability, it’s about how deeply a product understands, adapts to
and acts on behalf of its users.
Why the interface wins
What’s particularly striking
about the success of platforms like ChatGPT, DeepSeek, Cursor, Lovable and
Manus is that their underlying AI capabilities weren’t radically different.
In
many cases, they were simply thin wrappers over the same foundational models.
And yet, they achieved massive adoption. The key driver? Interface lock-in.
These products built intuitive, user-centric interfaces on top of powerful but
broadly accessible technology—turning technical parity into market leadership.
Subscribe to our newsletter below
ChatGPT’s mobile app brought LLM
technology into the hands of hundreds of millions of users. Its advanced voice
mode took things a step further, removing the friction of typing and making the
experience feel natural, conversational and intuitive.
Cursor reimagined the coding experience by embedding AI directly
into the coding environment, turning the integrated development environment into a place where code can be
reasoned about, refactored and generated collaboratively. Lovable did the same
for product development by letting you create full-stack apps just by
describing them without using code.
These products didn’t rely on
splashy marketing campaigns. They let
the experience speak. They relied on interface lock-in: the idea that a product
becomes indispensable simply because it understands the user better than
anything else. That is the new competitive advantage.
From transactional to conversational
Most apps
today are still designed around transactions—search, compare, book. The user is
expected to know what they want, where to click and how to get there.
Interfaces revolve around static entry points: a homepage, a hamburger menu
and a series of options neatly laid out on a screen. The user is in the driver’s seat, navigating a
fixed structure to complete a task.
But as AI
capabilities evolve, this mental model will begin to break. Tomorrow’s apps
won’t just respond—they will converse. They won’t just present options—they’ll
understand context. Interfaces will shift from being static to adaptive, from
menu-driven to intent-driven. They’ll be voice-first, visually rich and deeply
integrated across surfaces.
Most
importantly, the UX will become increasingly invisible. Instead of overwhelming
users with all possible actions up front, the interface will metamorphose based
on need. Buttons, prompts and suggestions will surface only when relevant and
disappear when not. The app itself will become a living, breathing
interface-fluid, anticipatory and responsive to the moment.
In this
world, users wonʼt need to learn how to use the app. The app will learn how to
serve the user.
Hyper-personalization through memory
The second
pillar of next-gen customer experience is memory. AI tools are increasingly
being designed to keep the conversation going—not just to engage but to
collect signals. Every query, every preference, every action becomes a data
point in a persistent user profile.
Done
ethically and with user consent, this long-term memory creates a compounding
advantage. Apps that remember who you are, what you like and how you behave
will soon be able to anticipate your needs before you even articulate them.
This predictive layer will power the most powerful form of personalization
we’ve seen yet, not just personalized marketing but personalized experience.
This
unlocks true personalization, not just collaborative filtering like “users who
bought A also bought B” but “you often take beach vacations with minimal
travel, here’s a direct flight and a hotel with a kidsʼ club and a pool with
slides.” The more the system knows, the
more useful and sticky it becomes.
The more
you use the app, the more it gets to know you. The more it knows you, the more
compelling the value proposition and the harder it becomes to leave.
Agentic
experiences: Doing, not just suggesting
Finally,
the most powerful shift weʼre about to witness is from recommendation to
action. AI-driven apps wonʼt just suggest or advise; theyʼll do things for you.
This is the leap from assistance to agency.
Imagine a
travel assistant that doesnʼt just tell you when ticket prices are low but
books them for you with your preferred airline, uses your credit card points,
fills in your passport details, checks you in and sends you the boarding pass—all while youʼre on a call or at dinner. Or a health app that doesnʼt just show
you lab results but schedules a follow-up consultation, books the cab, sets a
reminder and emails your insurance provider.
These
agentic applications will handle your research, make decisions for you,
complete purchases, make phone calls, book appointments and more. Agentic AI
is the true holy grail. And whoever builds the most trusted, intuitive and
reliable agent for each domain, be it travel, health, finance, education, is
going to win.
The real moat is experience
In the age of LLMs, access to intelligence is no longer the
bottleneck, experience is. The winners won’t be those with the best model
weights but those who make those weights invisible through beautiful
interfaces, persistent memory, seamless autonomy and deep empathy for the user
journey.
When users
come back to your product, it’s not because of your model weights. It’s because
of how the product makes them feel understood, helped, empowered and
relieved. Because in a world where everyone has access to the same models, the
best interface and the most human, anticipatory and agentic experience wins.
The
playbook has changed. The new moat is not algorithms; it’s experience.
