Why I reckon Knowledge Architecture is having its iPhone moment

Hi & welcome here. I'm Binoy, a Knowledge & Information Architect.

If you're someone who has sat in a meeting this month while several intelligent people disagreed about the meaning, ownership, or scope creep of a term, I'm writing for you.

I'm also writing for people working across Knowledge Management, Technical & Application Programming Interface (API) Documentation, Developer Education, Content Design, Information Architecture, Semantic & Ontology Engineering, & Data or Information Management & Governance. These disciplines are often viewed as silos, or housed in separate teams. Increasingly, I see them being asked to solve different parts of a similar problem.

I have worked across evolving generations of the broader technical writing industry since 2003, in the UK, Sweden, Belgium, & India servicing Global markets. What follows is a flattening & compression for brevity that I will also expand upon in future issues.

3 aspects moved, on 3 different axes. I unpack each below.

Which is why the current shift feels different. For the first time, humans are no longer the only first-class consumers of organisational knowledge.

For my Gen Zs reading this ;-) (and I like your sense of humour), the iPhone did not invent the mobile phone. However, it definitively changed who the interface was for, what people expected of it, & what could be built on top of it. So many industries reorganised around the consequences.

Something similar is happening with Knowledge Architecture & 3 data sets point in the same direction. 2 of the 3 come from vendors with a commercial interest in that conclusion, so I flag it where it matters.

GitBook published its own traffic data for one week in April 2026. 61.2 million page views. Strip the crawlers, count only reading done for a reason, & the split is:

The trajectory is the part I cannot give you. GitBook's earlier figures count agent traffic as a share of all traffic. This one counts it as a share of intentional reads, with crawlers stripped out. Line them up & you get a rising curve. Put them on the same basis & the most recent week is fractionally lower than December.

So I'm not going to show you a trend. One week, one platform, one measurement. On that week, machines did more of the intentional reading than people did.

OBS (Swedish shorthand I've never shaken, it means "note this"), 2 caveats:

• This is a platform-specific measurement from a vendor with an interest in agent-ready documentation ;-)).
• GitBook publishes no methodology. No bot-detection method, no agent-versus-crawler definition. Treat it as telemetry, not research.

The Postman State of the API (2025), n over 5,700 respondents drawn from Postman's own user base & also published without a methodology, finds that 89% of developers use generative AI daily, yet only 24% actively design APIs with AI agents in mind. Which means 60% still design primarily for humans only, & 16% have not considered AI agents at all.

The Stack Overflow Developer Survey (2025), n over 49,000 respondents overall, finds that more developers distrust AI accuracy, 46%, than trust it, 33%.

The authoring line above is the one I compressed hardest, & anyone who lived through it may notice as well.

Structured content & docs-as-code are not 2 points on one line.

They are competing answers to the same question, & the question, IMHO, is what counts as the unit of authoring?

• DITA answered topics, XML, & a CCMS (Component Content Management System).
• Docs-as-code answered Markdown, Git, & the Continuous Integration and Continuous Delivery (CI/CD) pipeline Developers/Engineers were familiar with.
• Many businesses went straight from unstructured docs & pages to docs-as-code without ever touching DITA. The fork is more interesting than the line.

Many regulated industries took neither branch, & that is where this also gets very interesting. For example:

• Aerospace & defence run S1000D, the international specification for technical publications.
• Instructions for use sit under IEC 82079-1, from the International Electrotechnical Commission.
• Automotive functional safety has ISO 26262, from the International Organization for Standardization.

They are safety & legal obligations with audit trails attached, which is precisely why they are the slowest estates to move & the most expensive places for a definition to be wrong. I unpack this in future issues.

There is also a generation sitting between all of this & where we are now, & almost nobody counts it. OpenAPI & Swagger. The moment documentation stopped being written & started being generated from a machine-readable description of the system.

A person can read an API reference and bring years of tacit understanding to it. An agent cannot safely rely on what the organisation has left implicit.

Consider a field called status.

The value may be technically correct, but an agent still needs to know:

• whose status it represents
• which process or market defines it
• when it became valid
• how it was calculated
• which exceptions apply
• which source has authority
• what action should follow

None of those questions is likely to be answered by the value itself.

They depend on context, relationships, definitions, provenance, business rules, and governance. They sit across documents, systems, & humans. In many organisations, no one owns the seams between them.

This is where the distinction between data interoperability and knowledge interoperability becomes useful, which I will go into in the next issue, together with expanding upon the evolving Knowledge Architecture Layer & my evolving understanding of the granularity of Knowledge Debt, a form of Tech Debt coming due.

This matters far beyond Knowledge Management & Documentation silos.

I work in Energy Flexibility because Climate change, Electrification, Decarbonisation, and the AI/Machine Learning (ML) Agentic Era have created a once-in-a-generation opportunity to rethink the electric grid. I'm interested in the AI/ML-first knowledge layer required to support that shift, while trying to let my human values bleed into the work and make a meaningful green-energy dent.

We are moving from centralised generation towards millions of distributed assets called Distributed Energy Resources (DER) in industry speak. Consumers are also becoming producers and participants. Electric vehicles, heat pumps, balcony solar, batteries, and renewable generation must interact with markets, suppliers, network operators, software platforms, & regulators depending upon the market type.

Every additional participant creates new relationships, rules, exceptions, & dependencies. The complexity is not only increasing numerically. It is increasing semantically & ontologically.

Each organisation has its own terminology, operating model, data structures, & version of reality. Then there is the hot mess of structured & unstructured data to contend with. Even when data can move between systems, its meaning often does not travel with it.

The Energy Transition, Decarbonisation, & Electrification are not a singular technological substitution. They are a reconstruction of physical assets, operational systems, institutions, markets, and expertise. Each part must exchange not only data, but meaning, authority, context, and learned judgment.

The Energy Transition is accumulating knowledge debt alongside its more visible technical and infrastructure debt. We can install more renewable generation, storage, & flexibility. But those assets cannot coordinate themselves.

Go To Market (GTM) Engineering did not appear because someone invented an entirely new kind of work, despite what Youtube cringelords may claim ;-)

The work already existed across Revenue Operations (RevOps), Sales Operations, Growth, Automation, Data analysis, and technically inclined sellers. Different people owned different parts of the commercial system, but few owned whether the whole system produced growth.

Around 2023 to 2025, those fragments began to acquire a shared commercial identity: GTM Engineering.

The change was not merely a new title. Work previously treated as operational maintenance was reframed as a growth lever. Go-to-market became something that could be architected, instrumented, tested, and continuously improved like a product.

Knowledge Architecture appears to be entering a similar category-formation phase.

The underlying work already exists across every discipline I listed at the top. I have also had to wear those different hats to arrive at a solution to an issue presenting as a Knowledge Architecture problem.

Each discipline owns part of the knowledge system. Few organisations have someone accountable for whether the entire layer works across Product, Engineering, Service, GTM, and AI. That was inconvenient when the main consumer was human, because a person could ask a colleague or infer the missing context.

Agents expose the cost of that fragmentation. They retrieve what exists, parse what we have made explicit, and scale the contradictions we previously paid people to reconcile.

This is the forcing event.

Categories form when fragmented work becomes important enough that somebody must own the outcome. GTM Engineering formed because fragmented commercial systems were slowing growth. Knowledge Architecture is forming because fragmented organisational knowledge is now slowing humans, software, and AI agents simultaneously.

Knowledge Architecture is not a completely new profession. Nor is it a replacement for Technical Writing, Knowledge Management, Semantic or Ontology engineering, or Information Architecture. It is a name for a larger responsibility that requires inter-disciplinary approaches.

It asks whether organisational knowledge can move from the person who knows, through an accountable system, to the human or machine that needs to decide or act. It treats knowledge as a product with users, interfaces, quality controls, forensics, observability, & measurable failure modes.

We already have names for the people building the electric grid, financing its assets, designing its markets, & engineering its software.

We do not yet have a widely accepted commercial identity for the people designing the knowledge layer that allows all those systems to understand one another.

That may be about to change.

The Energy Transition may be one of the industries that forces Knowledge Architecture into existence as a recognised category. AI, the Agentic Era, & the rising cost of interoperability are all simultaneously occurring events that make its absence impossible to ignore.

And the Knowledge Architecture Layer & Docs are likely where we see the shift first.

Fortnightly, Monday evenings, built in the open. Corrections get published dated, rather than quietly edited. Starting with one of mine: I have been calling the Energy Transition the largest knowledge management challenge in human history, I put it in a conference proposal this month, & I could not find a single source for it. Withdrawn.

Next: the Knowledge Layer in 4 parts, data interoperability against knowledge interoperability, & the 4 kinds of Knowledge Debt at 4 different interest rates.