The Context Engine

Kadhir Mani

(6 minutes)

<section data-section-id='8db60ac6-1585-4ca2-9ec0-8012de04c2c3'>Every company now has thousands of agents trying to get work done, and almost none of them know what the organization actually knows. The information exists. It's just scattered across Slack threads, PRs, call recordings, and the heads of a few tenured employees, unreachable by anything that isn't human. </section> <section data-section-id='20a95cd4-c4a5-432b-af2a-29fdb16bf9a2'><h2 id="the-problem">The Problem</h2>It's difficult to give agents the right bits of context they need to execute correctly, especially if you want autonomous systems that continually scale across your entire organization. After working with many companies, we've realized the core issue is that gathering the right pieces of information for agents to act on is surprisingly hard. This organizational context lives scattered across many tools (systems of record) and people (tribal knowledge). We have also realized that bringing all that organizational context together is rapidly becoming a big data problem, following the same 4Vs:<table><colgroup><col style="width: 92px;"><col style="width: 783.601563px;"></colgroup><tbody><tr><th style="width: 75px; background-color: rgb(242, 243, 245); border: 1px solid black; vertical-align: top; text-align: start;">V</th><th style="width: 75px; background-color: rgb(242, 243, 245); border: 1px solid black; vertical-align: top; text-align: start;">What it means for organizational context</th></tr><tr><td style="width: 75px; border: 1px solid black; vertical-align: top; text-align: start;">Volume</td><td style="width: 75px; border: 1px solid black; vertical-align: top; text-align: start;">The sheer amount of context generated daily (Slack messages, PRs, call recordings, docs, tickets) has outgrown any human's ability to track it.</td></tr><tr><td style="width: 75px; border: 1px solid black; vertical-align: top; text-align: start;">Velocity</td><td style="width: 75px; border: 1px solid black; vertical-align: top; text-align: start;">The speed at which decisions, code, and product reality change, especially now that AI agents produce output faster than people can absorb it.</td></tr><tr><td style="width: 75px; border: 1px solid black; vertical-align: top; text-align: start;">Variety</td><td style="width: 75px; border: 1px solid black; vertical-align: top; text-align: start;">Context lives in structurally different formats and systems (chat, video, code, documents, tribal memory) with no shared schema connecting them.</td></tr><tr style="height: 50.4492px;"><td style="width: 75px; border: 1px solid black; vertical-align: top; text-align: start;">Veracity</td><td style="width: 75px; border: 1px solid black; vertical-align: top; text-align: start;">Much of what's recorded goes stale or gets contradicted (superseded decisions, abandoned patterns, outdated docs), so the truth is uncertain even when the record exists.</td></tr></tbody></table> To tackle this problem, we built a context engine. </section> <section data-section-id='121e2d73-2b47-46cc-b598-ba3f3920cfe1'><h2 id="the-solution">The Solution</h2>A context engine is infrastructure. It's a system that continually pulls and reconciles information from every source that matters (systems, applications, people, and agents) into a single source of truth, specifically designed to address the 4Vs of organizational context. Two properties make it an engine:<ul><li value="1">Continuous reconciliation. It doesn't just store data once. It keeps pulling from source systems and resolving conflicts as they arrive, so the truth it holds stays current.</li><li value="2">Actionability. You can query it for answers and, just as importantly, act on it. That's what makes it the right foundation for running agents on top.</li></ul> </section> <section data-section-id='f796c28a-fa05-4c8f-951f-de8658b6f199'>And these five key attributes make it address the underlying problem:<ul><li value="1">Unified (Volume) <ul><li value="1">It needs to bring together the huge volume of information spread across many places into a single source of truth.</li></ul></li><li value="2">Living (Velocity) <ul><li value="1">It stays current automatically as source systems rapidly change, so the context inside reflects the world as it is right now.</li></ul></li><li value="3">Queryable (Variety)<ul><li value="1">It needs to transform the varied input schemas into a single format that we can easily run queries against.</li></ul></li><li value="4">Governed (Veracity)<ul><li value="1">Every asset carries version history, ownership, and access controls, enforced live at query time so we can verify truth over time.</li></ul></li><li value="5">Easy to contribute to (Veracity)<ul><li value="1">Context flows in as naturally as it flows out, so that humans and agents can propagate corrections back into the substrate as required.</li></ul></li></ul> </section> <section data-section-id='8494bbdf-82c0-4d02-8806-07e79b5794ad'><h2 id="why-now-the-multiplayer-first-ai-agent-forcing-function">Why Now? The Multiplayer-first, AI Agent Forcing Function</h2>AI agents are the most capable collaborators ever built. They're also profoundly amnesiac, every session starts from zero. It's especially difficult when you need everyone's agents to build on each other in a multiplayer first world. </section> <section data-section-id='8efacf06-ca5b-48ba-a5ad-416e9b8c6a97'>Getting agents to work together is hard, harder than getting a single agent to work well. The moment a second person or a second agent enters the picture, every private chat becomes a silo. One teammate solves a problem with an agent, and the next teammate starts from zero, because nothing about that session was ever shared, indexed, or made queryable by anyone else. That gap compounds with every handoff. A decision made in one session has to be re-explained, re-typed, or re-discovered in the next, and the organization pays that tax over and over. This is exactly why the substrate matters more than the agent. A context engine is what lets a decision made once become durable and accessible to every future session, human or agent, rather than being trapped in a transcript nobody else can see. None of this is easy to build. But the payoff, agents and people compounding on the same foundation instead of relearning it every session, is exactly why it is worth building. </section> <section data-section-id='de0faff5-bb6c-4274-83ce-a5a0a4f1ca0f'><h2 id="productnow-s-unique-approach">ProductNow's Unique Approach</h2>Everyone building in this space converges on the same idea: to pull organizational context into one place so agents can use it. Our differentiator is the form the context takes once it lands. We quickly realized, working with many companies, that trying to build a new abstraction, a new interface here, was going to be an uphill battle. The average person does not want markdown or some new format. They ideally want something they're already familiar with. And if you ask the average person today what context looks like, they'll say text. And we're all familiar with how to deal with text: in a document editor. So we started with that assumption. How do we build a context engine without changing the end interface? It needs to be just a document, viewed in a normal-looking web document editor. </section> <section data-section-id='ca32a217-1915-4b40-bab0-020d7577227b'>We then systematically solved the hard problems in our document editor that our context engine needed:<ul><li value="1">AI-native document format<ul><li value="1">We needed humans, agents, and our indexes to read the same underlying data without any translation layers for minimal latency and maximum correctness.</li></ul></li><li value="2">Agent-only editing<ul><li value="1">Agents need to reason about and edit our documents without any human intervention and done in such a way that scales as the models get better.</li></ul></li><li value="3">Human + agent multiplayer editing<ul><li value="1">We knew the majority of our documents would be written by agents, but they needed to play nice with each other and with any human authors concurrently editing; otherwise, the system would clobber edits and lose truth.</li></ul></li><li value="4">Semantic querying, filtering, and ranking<ul><li value="1">Agents and humans need to retrieve the most relevant, up-to-date information super efficiently.</li></ul></li><li value="5">Version control<ul><li value="1">Documents need a simple way to snapshot to cleanly capture version history so humans and agents could walk back to prior versions that have some semantic meaning as opposed to simple edit history.</li></ul></li><li value="6">Provenance tracking<ul><li value="1">Every piece of content has to be traceable to where it came from so we could maintain trust against external systems.</li></ul></li><li value="7">Permissions.<ul><li value="1">Authorization has to be enforced in our system since not all humans and agents may have a seat in the external system for us to copy permissions from, and we cannot leak information.</li></ul></li><li value="8">Rapidly scalable ETL<ul><li value="1">The number of tools we use is exploding, so we needed a system that would let us build integrations rapidly, including allowing our customers to build custom connectors as needed.</li></ul></li><li value="9">Agent-driven reconciliation<ul><li value="1">As we rapidly ingest new information, agents need to reconcile the new content with existing content to maintain a clean version history and keep context current.</li></ul></li></ul> </section> <section data-section-id='21e76ec2-cb28-4410-ada5-d4b0a99bb497'>We believe adoption compounds when the system is instantly legible to the average person. A context engine nobody can intuitively use or trust never reaches critical mass, no matter how sophisticated its retrieval is under the hood. We've solved many hard problems along the way to maintain that. Want to learn more? Shoot us a note at <a href="mailto:hello@productnow.ai" rel="noreferrer">hello@productnow.ai</a>. </section>

Published

September 7, 2026

Keep reading

AI-native data integration without an ontology

Every data integration effort historically required an ontology — slow to build, locking clients out of their own system, never truly self-service. AI changes that equation: continuous ingestion into living documents, agent-managed pruning, and a deep graph — integrations in 24 hours, not months.

Storing prototypes from external tools

An explanation of how we store external prototypes in the warehouse and what it took to get there.

CRDTs are not enough when your co-worker is an AI agent

CRDTs converge; they do not coordinate. That distinction becomes critical the moment an AI agent joins the editor. Why multiplayer collaboration needs a coordination layer above merge, and the section-level lease protocol we built to keep agents well-mannered.