Settling on OSINT

As the founder of an AI company, I literally make a living off of the idea that AI is the future, but not every industry is for an AI agent (at least not yet). You can force AI into just about anything if you try hard enough, but it’s rarely worth it unless it meets a few very specific criteria, some of which are actually pretty unexpected and even counterintuitive at first glance. First off, there’s the simple question of whether AI is necessary—if you’re dropping in AI just for bragging rights, don’t despair when nobody notices. The true benefit of automation is clear when there is a palpable need to save time by streamlining mundane tasks or to achieve a level of quality that surpasses what humans can achieve, often because of scale or consistency.

Without a good chunk of data, ideally mountains of data pilling up over many years, an AI system feels more like a novelty. Industries that generate and process tons of information practically beg for a tool that can sift, analyze, and interpret everything at scale. Without data, there is no benefit that ChatGPT, Claude, or any other generalized language model can provide because they can’t address any problem stemming from generic data. That is obvious, but this next point is much more counterintuitive: outdated systems that aren’t new and shiny provide the best canvas for AI agents. Trying to integrate with outdated or patchwork setups might sound like a nightmare, but it ultimately means that automation can make a meaningful difference. These messy integrations are often where humans struggle most and AI can fill in the gaps.

On top of that, industries with specialized knowledge are much better candidates for AI adoption. Some industries are so niche, so jam-packed with insider jargon and complex workflows, that ChatGPT or an OpenAI API wrapper created in a day simply won’t cut it. If day-to-day tasks are repetitive, you might as well let an AI agent handle them, especially if it takes some effort upfront to aggregate the knowledge to do so because this creates a moat.

There’s also something to be said for focusing on tasks that come up constantly. Of course, this is true for any software business, not just an AI firm, but an AI system uniquely thrives on repetition. Past results can serve as training data allowing AI to learn incredibly complex but routine steps incredibly quickly and accurately. That consistency translates into tangible results—less manual work, fewer errors, and a lot more time to solve the bigger problems.

As for the most counterintuitive point on this list and probably one of the more unique takes you have seen on AI, my final point is that while you would assume that heavy regulation should scare prospective AI companies away, regulation can actually be a strong reason to introduce it from the perspective of an AI firm. With well-defined rules, rigorous documentation, and a trail of processes that entail precision, an AI agent can jump in to keep things on track, log every step, and reduce bureaucratic headaches — bureaucratic headaches which often create the need for AI intervention in the first place. Humans are terrible at stepping through complex hoops but AI is actually pretty skilled at following hyper-structured formats and processes. Not to mention, once this regulation-following capability presents itself as a feature, it can serve as a wide moat against any potential competitors looking to encroach on your territory.

Running down this list—actual need, heaps of data, complex integrations, specialized knowledge, time savings, repetitive workflows, and regulatory structures—we found several sectors aligning with my thesis on AI agents: energy, defense, healthcare, archival management, property management, and law enforcement. Some were over-saturated with AI solutions, some we just weren’t interested in, but law enforcement seemed comparatively neglected by AI firms and, on top of that, we were actually interested in the problem set as a firm. It’s got the data volume, the tricky systems, the specialized know-how, and enough repetitive tasks to keep any AI agent busy for a lifetime.

We had settled on an industry, providing we could find a good problem to work on, but we didn’t know quite where to start. The eye-opener was talking to the folks who work in OSINT every single day. We were told about the obsolete tools, the multiple tabs that are constantly open, and the challenges of finding simple information. That’s when it clicked that we weren’t just picking a market; we’d stumbled onto an opportunity to make a difference not just in law enforcement but across a complete set of diagonal markets where investigations play a major role.

Ultimately leading us to Intrace — AI Agents for Open Source Intelligence. (Don’t worry, we know AI features aren’t an excuse for a poorly designed platform and we plan on building our agents on top of the best platform possible — no AI shoved down your throat).

If this is exactly the proverbial nightmare — ok, maybe that was a slight exaggeration, but come on, the status quo isn’t great — we would absolutely, positively love to get your insights. We’ll build Intrace, but someone’s got to design it and it’s not gonna be us! Join the waitlist at http://intrace.ai/waitlist for early access and know that we treat feedback incredibly seriously.