Why Inquisita
Your AI is more capable than
the software you're buying.
We built AI software tailored to a specific workflow. Then we realized the software was the problem. Here's what we learned and why it matters for you.
Hitting the Limits of Custom Web Apps
Before the current version of Inquisita, we built a traditional SaaS product for legal discovery. It had a web interface, agentic workflows, and AI-drafted responses. User interviews told us it was simple and intuitive. We took that as validation.
But a simple interface for a narrow workflow led to its own problems.
Limited by design
Our system was built around a specific workflow, and every new use case required new engineering. We had designed for one type of legal proceeding in one jurisdiction, but requirements varied significantly across practice areas and localities. When customers asked "can your platform do this?" the honest answer was usually "not yet, give us a few weeks."
Meanwhile, the backlog kept growing. One customer asked if we could organize and extract data from 5,000 receipts. Even with modern development tools, reshaping our platform and rigid data model to handle that would have taken weeks.
The cost trap
To power our custom workflows, we needed high-intelligence models. We paid API rates directly to model providers and passed those costs on to customers.
At the same time, all of us on the team were heavy users of AI through our own subscriptions. We started noticing something: we were getting far more compute through a $200/month subscription than we could ever afford at API rates. Our customers were paying us a premium for the same intelligence they already had access to.
The more we thought about it, the more we realized this wasn't just our problem. No SaaS company can provide frontier AI to customers at a better rate than customers can get it themselves. The AI labs will always offer the best price on their own models. Any middleman either passes those API costs through, or quietly downgrades to cheaper models to protect margins.
Adoption friction
Our product was yet another app to learn, another integration to maintain, another training rollout. Our target users would have to learn our interface, then still use other platforms upstream and downstream. This is a familiar obstacle: does it integrate? Can I teach my team this, on top of the dozen other tools we already use? For a platform with limited functionality, the friction of adoption was hard to justify.
The Turning Point
These obstacles made it clear our software couldn't scale to meet customers' diverse needs. And our integrated AI workflows were quickly becoming antiquated.
One of our founders spent two weeks building a feature to generate Word documents in each customer's preferred format. Then Claude launched a docx editing skill that did the same thing. The pattern kept repeating. Custom agentic features, expensive to run and constrained by maintained prompts, were being outpaced by general-purpose AI that improved every month without our intervention.
The real tipping point was personal. During tax season, rather than buying dedicated tax software, one of our founders prepared and filed business taxes using just their AI assistant. A few hours of work, reviewed by a professional before filing. And we realized: every AI-savvy business is going to make the same calculation. Why buy narrow software when your agent can already do the job?
We stopped asking how to build better AI software. We started asking: how do we build infrastructure that makes AI agents far better at working with documents?
Four Primitives
We distilled every customer request down to what they actually needed from a document intelligence system. Whether it was legal discovery or sorting 5,000 receipts, the asks kept converging on the same four operations:
Query
Search thousands of documents in seconds by keyword, metadata, or meaning.
Organize
Group documents into collections that align with a task. Your agent builds the groupings; your team reviews them through the Inquisita interface and builds on them later.
Analyze
Get targeted answers across thousands of documents per minute. Every document is processed. If something fails, you know about it.
Share
Work persists. The next agent or person picks up where the last one left off.
These four operations, exposed as tools any AI agent can use, turned out to be flexible enough to handle use cases we never designed for.
Design Tenets
When designing our new system, we made a series of tenets that would ensure our product would work flawlessly for our customers.
Agents are first-class consumers
Every feature is evaluated for agent usability. If an agent can't discover, reason about, or act on it, the feature isn't finished. The tool surface stays small and discoverable.
Intelligence lives with the agent
Inquisita provides data and tools; your agent provides reasoning. Any agent, any provider. We keep expensive reasoning out of the platform and in the hands of your agent, where you already have access to the best models at the best price.
Bridge agent work and human understanding
The platform makes it easy for agents to show their work, not just tell. Whether that's linking to source documents or generating visualizations, the path from agent output to human verification should be short.
The data model is the foundation
Flexible enough for use cases we haven't anticipated, intuitive enough for agents to query without special guidance, clean enough to audit. We preserve original source data so every derived field traces back to what the source actually said.
Simplicity is the default
Every dependency, abstraction layer, and pre-processing step must earn its place. Simple and reliable beats clever and fragile.
+100% value over core agents
Every feature makes agents dramatically more effective than going it alone. Faster results, fewer tokens, better outcomes. If an agent could vibe-code a solution faster than using our tool, we haven't earned our place.
Work compounds
Everything an agent does in Inquisita leaves behind structured context the next agent can build on. Your 50th matter should be as navigable as your first. No agent starts from scratch.
Where We Are Now
Inquisita is now a broad platform capable of handling far more than legal documents. Its general-purpose design suits any document-intensive industry: regulatory, journalism, M&A, academia, insurance. We support 88 languages, so international work just works.
Because we built around agents rather than any specific AI provider, we ended up with something we didn't originally plan for: a true multi-agent platform. Your documents, collections, and analysis results live in Inquisita regardless of which agent you use. Switch providers, use two at once, let different team members use different agents. Your data stays put.
Our favorite part as founders: since we built on primitives and let the agent handle the reasoning, when a customer asks "can I do this with these documents?" the answer is usually yes. We've had firms get their whole team working in Inquisita in a day, because there's almost nothing to teach. Your team already knows how to talk to their agent.
