MadCap and Spekit Put Enterprise Knowledge on the Clock

A company can possess the right answer and still fail to use it. The approved product language sits in a slide deck. The latest safety procedure waits in a learning portal. A technical fix lives in documentation that a support agent cannot find quickly enough. Then an employee asks an AI assistant, gets a polished reply, and discovers too late that the source was last quarter's truth.

This is the enterprise knowledge problem after the chatbot boom. Production became cheap. Plausibility became abundant. The scarce thing is a current answer with an owner, a permission model and a short route to the decision being made. Two private software companies in Denver, MadCap Software and Spekit, are building around that scarcity from notably different directions.

They are not one company, and no public partnership between them was found. MadCap comes from technical communication. Spekit comes from sales enablement and digital adoption. Yet place their current products side by side and a useful map appears. MadCap is concerned with the supply chain of governed knowledge: how content is authored, structured, approved, reused and delivered to many endpoints. Spekit is concerned with the last mile: how a salesperson or AI agent receives the approved fragment in the middle of actual work.

The source has to survive the journey

MadCap was founded in 2005 by veterans of eHelp, the company behind RoboHelp. Its durable idea was single-source publishing: write and manage content in a form that can be reused across web help, print, knowledge bases and other formats. That history matters now because generative AI has turned structured content from a documentation preference into an infrastructure question.

The company's portfolio now spans Flare for technical authoring, IXIA CCMS for large-scale DITA content, Create for learning materials and Syndicate for aggregation and delivery. The acquisitions tell the story. MadCap added IXIASOFT in 2023 and Xyleme in 2024, broadening its reach from product documentation into component content management and learning content. Flare Online followed in 2025 with cloud collaboration and AI assistance.

In April 2026, MadCap introduced a broader version of Syndicate with semantic search, analysis, configurable portals and features aimed at making enterprise content usable by AI. Its product materials describe content as small, independently queryable objects with embeddings, metadata, version controls and access rules. An outside AI application can authenticate and query the governed material through an API. If nothing relevant exists, the stated design goal is to report that absence instead of inventing a convenient answer.

The costliest AI answer is often the one that sounds right after its source stopped being true.YesPress analysis

This is not glamorous work. It is taxonomy, ownership, permissions and lifecycle management. It is also the part that determines whether a retrieval system has clean fuel. A model pointed at an uncurated folder does not repair the folder. It makes the folder conversational.

The last mile is measured in seconds

Spekit begins closer to the employee's cursor. Co-founders Melanie Fellay and Zari Zahra launched the company at Salesforce's Dreamforce in 2018, originally tackling outdated documentation, decentralized training and the awkwardness of teaching people a changing CRM. Spekit announced a $2.54 million seed round in 2019, a $12.2 million Series A in 2021 and a $45 million Series B in 2022.

The present product organizes go-to-market knowledge into modular units called Speks. Each can carry metadata, sources, governance and access controls. The Knowledge Engine syncs material from other systems and tracks freshness and ownership. Its AI Sidekick then surfaces guidance, drafts and recommended content in tools where a representative is already working, drawing context from CRM, conversation intelligence and communications systems.

The distinction is behavioral. A traditional repository assumes the user will stop, search, judge the result and return to the task. Spekit tries to collapse that excursion. During call preparation, follow-up or deal-room creation, the answer should appear near the action it informs. In December 2024, Spekit acquired Cquence, whose technology extracted deal insights from conversations, CRM records and other unstructured sources. The acquisition pushed Spekit further from a training overlay toward a context engine for revenue work.

Fellay has called this approach “just-in-time enablement.” The phrase is marketing, but the operating idea is sound: relevance decays with distance. A perfect battle card found after the customer call has the practical value of a missed train.

What operators can steal

The obvious mistake is to choose between a content system and an AI assistant as if the interface were the architecture. The stronger lesson from MadCap and Spekit is that companies need a chain of custody for knowledge. Someone must be accountable for the source. The system must preserve versions and permissions. Retrieval must respect context. The employee should be able to inspect where an answer came from. Finally, feedback from the work should help identify gaps upstream.

A practical audit can begin without buying either product. Pick one high-risk question that employees ask repeatedly: a pricing exception, a compliance procedure, a product limitation. Then trace its path.

1. Find the authoritative source and name its human owner.
2. Record the approval date, review cadence and groups allowed to see it.
3. Break the answer into the smallest reusable unit that still makes sense.
4. Test whether a worker can retrieve it from the tool where the decision occurs.
5. Require a visible source and a safe “no answer” behavior when evidence is missing.

This exercise often reveals that an AI project is really a content operations project. The model can summarize and compose, but it cannot nominate a policy owner or decide that two conflicting documents have different authority. Those are organizational choices wearing technical clothes.

Two businesses, one pressure

MadCap and Spekit also expose a convergence that should make software buyers more careful. Documentation platforms now talk about AI agents, semantic retrieval and knowledge infrastructure. Sales enablement platforms now talk about governance, source transparency and content health. Category lines blur when the shared problem is trust.

The overlap does not make the products interchangeable. A global manufacturer managing regulated manuals in several languages has different authoring and audit requirements from a revenue team trying to coach a representative through an objection. MadCap's center of gravity is the controlled content lifecycle across technical documentation and learning. Spekit's is go-to-market execution inside daily applications. Buyers should start with the failure they need to prevent, not the AI label on the demo.

The economics differ, too. Bad documentation can multiply support tickets, delay training and create regulatory exposure across a product's life. Bad sales guidance can lose a live deal or send an off-message promise to a buyer in minutes. One failure accumulates through a content estate; the other can happen inside a single conversation. This is why measuring only searches, clicks or generated words misses the point. Teams should watch how quickly approved changes propagate, how often users receive an answer with a valid source, how many questions return no evidence and whether the guidance changes a measurable outcome. A knowledge system earns its place when it reduces uncertainty without hiding where the answer came from.

They should also ask for evidence. Which repositories can be synchronized? How are conflicts resolved? What happens when a source expires? Do permissions survive retrieval? Can the answer cite the approved object? How quickly does an update reach every endpoint? These questions are less exciting than watching a chatbot draft an email. They are much closer to the reason an enterprise buys software.

The larger shift is from documents as destinations to knowledge as a service. A manual still matters. So does a training course or a sales playbook. But each is increasingly a presentation of smaller governed units that may also feed a portal, an assistant, a search result or another machine. The page is no longer the only final form.

That makes time a design constraint. Content has to be current when a system retrieves it and close enough to the work to influence behavior. MadCap is tightening the first clock. Spekit is tightening the second. The enterprise knowledge stack will be judged by whether both keep time.