Since the beginning of the year, I’ve been participating in discussions about the promise and limits of agentic AI, which is generally defined as a system that enables AI to make independent analyses and decisions without much human input. It has created a second wave of public interest in AI, following the launch of ChatGPT in late 2022, which introduced much of the world to GenAI.

Why all the attention? Agentic AI is a big leap forward in realizing our dreams of a world where AI can not only do things faster and better, but, with our guidance, reason independently on our behalf. If GenAI is about productivity, agentic AI is about agency, a power we typically attribute to humans.

But if AI has human-like agency, should we expect it to reason like humans?

Last week, I spoke with two founders of Abstract, an AI startup that provides “real-time, contextualized policy intelligence.” They’re looking to tackle a longstanding problem for businesses: making sense of the accelerating volume of policy changes resulting from the plethora legislation at the federal, state, and local levels.

What sets them apart is that the agents apply context for interpreting and predicting the impact of these changes, the way a human policy analyst might do, but at scale. The end user — a human being — needs context, so agents need to be capable of providing it.

“AI models are already good at reasoning,” said CEO Pat Utz. “What’s missing is fine-tuned context. We ingest and enrich data, then pass it to LLMs to generate responses aware of the user or organization.”

Over the past five decades, the task of monitoring and responding to policy changes has become nearly impossible. The volume of federal restrictions alone has grown from 400,000 restrictive words in the 1970s to more than one million today, according to the Office of the Federal Register. There are more than 145,000 federal, state, county, and city government entities that pass 3,000 to 4,500 final rules annually. Adding to the complexity is a wave of federal deregulation under the current Trump administration, which is shifting regulatory responsibilities to state and local governments. On top of that, legislative documents are hard to read. To the average human, they make little sense.

To keep up with the deluge, Abstract tracks all the aforementioned data to provide insights into risks and opportunities in context. By providing this level of context at scale, it has positioned itself for the “policy intelligence” market in several ways.

First, it expands the market beyond compliance, the primary focus of legacy Government, Risk, and Compliance (GRC). “Compliance is reactive. It kicks in once a regulation changes,” said Utz. “Abstract is focused on everything before that. We abstract the noise so we can identifyrisks and opportunities early, before compliance is even necessary. There is the proactive piece that provides an early warning system on how legal and regulatory changes may pose a risk to the organization.”

This proactive piece has drawn interest among businesses — some in the Fortune 500 — that don’t just want to keep pace with the change, but to stay ahead and capitalize on it.

“A large cybersecurity firm uses Abstract across 20 states to proactively revise contractor agreements, avoiding fines or taxes before compliance issues arise,” said Utz. “Law firms use Abstract to identify the best states or cities for their clients to open offices, based on regulations, taxes, etc.”

Second, context enables Abstract to verticalize for businesses that need to provide the high-level counsel they expect, including an analyst’s ability to see around the corners of a subject and make thoughtful recommendations. In addition to its work for large businesses, Abstract has made inroads with large national law firms in the Am Law 200.

Finally, Abstract is expanding its user base beyond in-house legal and regulatory departments to departments like HR, product, finance, knowledge management, innovation, and business development, which use Abstract to personalize outreach and insights for their clients.

“We want to structure and surface relevant information across all business units. International expansion is on the horizon, too,” said Utz.

Abstract isn’t alone in the U.S. market. Companies such as FiscalNote and Quorum also offer proactive policy tools, but according to Utz, they don’t deliver the context that sets Abstract apart.

Abstract’s sweeping POV on current and future users hearkens back to the founders’ original mandate: to democratize access to government data. Founders Utz and Mohammed Hayat — who conceived the company while undergrads at Loyola Marymount University in Los Angeles — along with their co-founder Matthew Chang, a UCLA alum — had something in common: they each came from immigrant families that were frustrated with the lack of transparency and accessibility of government records in their home countries.

“During our family dinner time, my parents would discuss heartbreaking stories of how families were devastated by government policies that led to hyperinflation and corruption across business and government,” Utz told Medium. “Those discussions stuck with me and empowered me to launch Abstract with my co-founders.”

It’s an egalitarian view of government access, more difficult today perhaps than ever before. But providing access and context at scale, Abstract hopes to make a difference.