RIP the wiki.
RIP the wiki, the novel (30 years ago) place where all the important information was recorded for employees and new hires to reference as needed. It marked a small step toward digitizing the company’s DNA, a very small one. It didn’t, however, have anything truly useful. Not how the company ended up in its current state, not the impactful decisions and the thinking behind them, not the wrong turns and what they taught anyone. None of that was in there. Zero context. But you could always find the holiday schedule.
So the employee struggled to understand the business context, left to piece it together over time if they were given access. Maybe more importantly, the business couldn’t truly benefit from the richness of the employee’s experience: their experiences, skills, connections, and knowledge were only applied going forward. They carried their personal context inside their head.
This is going to change.
We were building it the whole time.
We have been moving toward this for years, mostly unknowingly. Small steps leading to a final destination.
It started with comms. Email moved conversations out of the hallway and onto the record. Then Slack and its cousins took the rest of it, the quick questions, the side channels, the decisions that used to happen in passing, and turned them into text that persists. The talk that used to evaporate started leaving a trail. “Can you slack me that?”
At the same time, the work itself moved off of individual compute devices and into cloud-based collaborative systems. The documents, the projects, the tickets, the workflows all migrated out of individual files and inboxes and into tools where the state of things lived in one place instead of on someone’s desktop. The data followed. Large warehouses like Snowflake and platforms like Databricks pulled the numbers out of a hundred scattered systems and into a single store you could actually query.
More recently, even more human-to-human interactions came online. The meeting and the phone call, the richest and most stubbornly human context in any company, began to be recorded, transcribed, and searchable. The conversation, the one thing that always disappeared the moment it ended, started leaving a trace like everything else.
Each of these was sold as its own tool solving its own problem. None of them announced themselves as steps toward a bigger idea. Every one captured a slice of context that used to live only in the moment or in someone’s head, and stored it.
The problem was that we captured it faster than we could ever use it. We built lakes of meeting transcripts no one would ever watch, years of Slack history no one would ever read, warehouses full of tables no human could hold in their mind at once. The context was there, more of it every year, and it sat inert. We spent years scaling data capture but lacked a reciprocating way to act.
The other side arrives.
Then the other side arrived.
For all that time we scaled the capture of context. What we never had was anything that could act on it at the same scale. Storing was easy. Using meant a human reading it back, and no human could read back a company’s entire history at once.
Scaled intelligence is what changed that. For the first time there is something that can take in the whole record, hold it at once, and reason across it, at the same scale we spent years building toward. The capture had been scaling for a generation. Now the intelligence to act on it scaled to match.
That is the arrival. Not a better tool for a narrow task, but the first force capable of meeting all that accumulated context on its own terms.
The context layer.
When the two meet, you get the context layer.
Not a metaphor, and not a single new product you buy. The context layer is what forms when a company’s entire accumulated context becomes readable and usable by an intelligence that can hold all of it at once. The substance was already there. The intelligence is what turns it into something you can actually draw on.
And notice what happens to the shape of it. The context layer used to be dozens and dozens of discrete applications, each holding its own slice: the email system, the chat tool, the CRM, the warehouse, the doc store, the ticketing system, the meeting recorder, each a silo with its own login and its own fragment of the truth. And so many point solutions that the finance department had to audit their usage. Context was real, but it was scattered across a hundred places, and no one and nothing could see across all of them at once.
That is what is now converging. The slices are collapsing into a single readable whole. Not because the old tools disappear, but because the intelligence sitting on top of them can finally treat them as one thing: one body of context, queryable, coherent, entire. The fragmentation that defined enterprise software for a generation is being dissolved from above.
It is already starting. Starbucks recently announced it was dropping licensed Microsoft software in favor of building its own bespoke solutions. That is not a story about software procurement. It is a story about a company deciding it would rather own its own context than rent someone else’s container for it. Expect more of that, not less.
This is the beginning of the context layer. And it is why context suddenly matters far more than it did even a few years ago. For the first time, a company can see itself whole.
A memory that compounds.
The context layer is built around the interactions between employees and the models used to access and interpret the central store of data. As employees interact with the layer, every one of those interactions becomes context itself.
The insights of one are scaled to all. The answer that worked, the path someone took to solve a problem, all of it gets captured and kept, and over time that accumulation becomes a collaborative memory: a growing body of usable interactions contributed by everyone who uses the layer and available to everyone who uses it next.
This is the company’s proprietary contextual intelligence, and it compounds. When one person solves something, the next does not start over, they inherit the path, and the lone-wolf discovery that used to die on a single laptop becomes something the whole company draws on.
There is a catch, and it is the important one. Not every interaction deserves to be remembered. Most questions are noise, most exchanges are one-offs, and a layer that naively keeps everything anyone ever asked fills up with junk, contradictions, and wrong answers someone happened to accept. Perfect capture does not produce a memory. It produces a swamp.
So something has to decide what is usable. Something has to promote the interaction that mattered and discard the thousand that did not, to separate what the company actually learned from what it merely said. That word, usable, is carrying the whole design. And deciding what is usable is a judgment, which means a human, or a human-built rule, sits at the center of it. The context layer is not switched on. It is built, curated, and shaped, continuously, by people deciding what is worth the company remembering.
Start building.
So what do you actually do with this?
You start building your context layer. Not next year, not after the technology settles, now, because the pieces are already in your company waiting to be connected, and the companies that assemble them first will compound a lead the others cannot easily close.
A few places to start.
Take inventory of where your context already lives. The comms, the documents, the warehouse, the transcripts, the tickets. You have been accumulating the raw material since the day the company started, without calling it that. The first step is seeing it as one thing instead of a dozen disconnected systems.
Decide who owns it. The context layer is not an IT project. It is the company’s memory, and memory needs a keeper. Someone has to own what the layer remembers, what it discards, and who it answers. Name that person and give them real authority, or the layer will fill with noise and answer to no one.
Build for capture, not just access. Most companies will roll out a chat interface and lose every interaction the moment it happens. The value is not in letting people ask questions. It is in keeping the answers that worked so the next person inherits them. If you are not capturing the usable interactions, you have a search box, not a compounding memory.
Give your managers their new job. The manager whose value was holding context in their head and parceling it out is about to find that job done by the layer. The ones who thrive will stop being the keepers of knowledge and start designing how their teams draw on it. Tell them that now, before the layer tells them for you.
None of this requires waiting for a breakthrough. The intelligence is here and the context is already yours. What is missing is the decision to build the layer deliberately instead of letting it assemble itself, badly, while you are not looking.
The companies that build it well will know themselves in a way no company ever has. That is the advantage. It is available now, and it is yours to take or to cede.
And there is a second question waiting behind it. Once the context layer is built, the relationship between the company and the tools used to build it becomes more charged. The intelligence that makes your context usable is not yours. It is rented. That relationship is about to get complicated.