Reflections on the TLTF Summit and a Year of Building AI Agents in Immigration Law

AI is reshaping legal tech. From funding growth to groundbreaking tools, learn how startups like CaseBlink are navigating this rapidly evolving landscape.


Reflections on Legal Tech Innovation from the TLTF Summit

TLTF

I found myself reflecting on the transformative discussions that took place at the TLTF Summit today. Legal tech funding is surging, opening doors for new ventures and acquisitions. In addition, starting a company has never been more accessible, with lowered barriers and the advent of foundational large language models, there is a clear surge in innovation. AI-enabled corporate clients and consumers are increasingly bypassing traditional legal services, addressing their legal needs with advanced AI tools. This shift is already transforming how legal services are accessed and delivered.

For tech startups and innovators, this transformation requires us to rethink legal solutions from the ground up. In an increasingly competitive market, defensibility and unique value propositions are essential. For legal tech products, a good user interface and experience aren’t just extras - they can become what sets successful tools apart. 

I see these discussions as immediate, actionable principles for our work as early stage founders. The summit also reaffirmed what we’ve believed since the very early days of our startup, that the legal industry is undergoing a major transformation, and those who fail to adapt risk being left behind.


          

How We're Navigating a Rapidly Changing Landscape

About a year ago, when CaseBlink was in its earliest stages, I had countless conversations with immigration attorneys about the future of the field with AI. The general sentiment was cautious optimism. While many acknowledged AI's potential to assist, there was a strong belief that it could never handle tasks like reasoning through case strategy or assembling a fully compliant immigration packet.

At that time, concerns about GPT hallucinations were widespread. These issues came from the limitations of large language models both in terms of context windows, and/or the absence of contextual grounding needed for reliable outputs. Skepticism was justified, but it’s now clear that the way we work is undergoing an irreversible transformation. Whether you’re an engineer, lawyer, artist, the very foundations of your profession are shifting beneath you.

AI has been evolving faster than I imagined, despite a career rooted in data science and machine learning. For an early-stage startup like ours, this evolution is both an opportunity and a challenge. Infrastructure, frameworks, and models that seemed cutting-edge just a year ago can quickly become liabilities. Building in this era is exciting but also uniquely demanding. The speed of innovation requires constant adaptability and a willingness to rethink everything - every single day. It’s stressful, but it’s also a privilege to be at the forefront of it.

Solving Tomorrow’s Problems, Today

Our team has spent countless hours refining models to ensure outputs are accurate, contextual, and free of hallucinations. Early on, this focus on quality set us apart. But we were also able to recognize that sticking to yesterday's methods in a fast-moving environment would only hold us back, and so we chose frameworks that help us adapt to the models and frameworks of tomorrow.

What does it mean to solve tomorrow’s problems? It means pursuing a vision that doesn’t always make sense in the short term. Already, we’re seeing AI replicate and even surpass certain aspects of human intelligence. But what happens two or three years down the road? Will models reason, iterate, and collaborate so effectively that they autonomously perform tasks traditionally requiring exceptional skill - at lightning speed? Has that happened already? These are the foundational questions that guide our work. 

Right now, users are growing comfortable with AI getting them 60-70% of the way, with human reviews to fill in the gaps. But what happens when AI reaches 95%? What if, in three years, it delivers results that are consistently better than human experts? The systems we’re building today meet present-day needs with human oversight, especially in high-stakes fields like legal. Yet our challenge as founders is clear: To deliver immediate value while preparing to adapt for a future where technology evolves faster than we had all expected.

Khalil Zlaoui
CEO, CaseBlink