Most conversations about AI in contracting focus on what the technology can do: draft a clause, summarize an agreement, flag a risk. Far fewer focus on what makes those outputs reliable enough to bet a business on. In this installment of The Intelligence Brief, our video series exploring the ideas shaping AI-native contracting, Icertis Senior VP of Product Management Sundar Balasubramanian unpacks the Vera Foundation — the underlying architecture that now powers every Icertis application, starting with the reimagined Vera Copilot.

Three years of lessons, distilled into an architecture

The Vera Foundation isn’t a single feature or a marketing label bolted onto existing tools. It represents three years of hands-on experience building generative AI applications, combined with the rapid pace of innovation in underlying large language models. That combination matters: capabilities that weren’t possible even six months ago are now foundational to how Vera understands and acts on contracts.

The five-point harness

What sets this apart from point solutions that only draft or only analyze is a deliberate design choice — a five-point harness built specifically for the demands of enterprise contracting.

• Intent understanding. Before Vera does anything, it works to understand what a user is actually trying to accomplish — whether they’re searching for information or trying to get something done. That distinction shapes everything that follows.
• A hybrid RAG pipeline for deterministic answers. One of the more counterintuitive insights from the video: more than 80% of customer queries are looking for the same category of answer — a defined set of agreements — and users expect the same question to produce the same answer every time. Generic generative AI, built for open-ended creativity, isn’t naturally built for that kind of consistency. Vera’s response is a hybrid retrieval-augmented generation (RAG) pipeline paired with low-latency retrieval and pre-computed answers for common and even complex queries. The result is faster responses, higher accuracy, and — notably — greater token efficiency, since the system isn’t reasoning from scratch every time.
• Low-latency retrieval. Paired with the RAG pipeline is a system built for speed — pre-computed answers for common and even complex queries, so responses come back fast without sacrificing accuracy. The result is greater token efficiency too, since the system isn't reasoning from scratch every time.
• Contracts as families, not documents. Vera doesn’t evaluate an agreement in isolation. It understands the full contract graph — from a master service agreement to its statements of work, amendments, and supporting documentation — as a single interconnected family. In many industries, this isn’t a nice-to-have; it’s make or break. A question about liability, pricing, or obligations often can’t be answered correctly without the full context of related documents.
• A customer-specific ontology. Beyond a general knowledge graph, Vera builds an ontology unique to each customer — informed by their system of record. That means Vera has context on a business’s specific rules, playbooks, templates, contract-type relationships, and even organizational structure. Practically, this means users don’t have to re-explain their business every time they ask a question. Vera already understands the enterprise context, which translates into faster, more relevant answers.

Together, these elements are designed to solve a problem that’s dogged generic AI in high-stakes business contexts: the gap between impressive demos and dependable, repeatable answers enterprises can actually act on.

One foundation, an expanding set of applications

The Vera Foundation isn’t limited to a single tool — it’s the engine meant to power Icertis’s entire application suite going forward. The first proof point is the newly relaunched Vera Copilot. But that’s only the beginning. Icertis is reimagining discovery, ingestion, insights, risk, and obligations experiences on top of this same foundation, with a full transformation of the product suite expected over the next 6 to 12 months.

Newer products are already being born directly from this architecture rather than retrofitted onto it. Vera Engage, for example, was built on the Vera Foundation from day one — meaning customers get the benefits of the five-point harness without waiting for a future upgrade.

Why this matters for legal and contracting teams

For legal, procurement, and commercial teams evaluating AI tools, the distinction between “AI applied to contracts” and “AI built for contracts” is becoming the deciding factor. The Vera Foundation is Icertis’s answer to that distinction: an architecture built specifically to handle the ambiguity, interconnectedness, and organizational nuance that make enterprise contracts uniquely difficult for generic AI to reason about.

Watch the video above to see the Vera Foundation explained in full, and stay tuned for the next installment of The Intelligence Brief, where we’ll continue exploring the ideas driving AI-native contract intelligence.