AI Tools in the Archive: Sector 01 by the Numbers

Sector 01 is large. It is not the whole archive.

AI software is easy to overcount in the mind. It is in product announcements, investor decks, release notes, browser tabs, and the name of every second tool that would previously have been called an assistant, automation, or editor. The impression is that the entire software map has become one bright, humming room.

ARCHIVE-9 has a less dramatic answer. On August 28, 2026, its deep archive held 8,497 software records. Of those, 733 are filed in Sector 01: Artificial Intelligence. That makes AI 8.6 percent of the collection.

This is a count of primary classifications, not a keyword search for AI, and that distinction is important. A recruitment platform may use models extensively but be filed under Personnel. An image editor may lead with generative features but still belong in Design and Media. A support product may advertise an agent while its actual job is handling customer conversations. The archive gives each record one address so that a directory remains navigable.

SECTOR 01 IS NOT SMALL. IT IS SIMPLY NOT ENTITLED TO EVERY ROOM.

The result is not a claim about the total AI market, revenue, quality, or adoption. It is a live record census. Still, it provides a useful correction to the loudest version of the story: AI is a major software sector, while also being a capability spreading through many others.

Where AI sits in the current census

Sector 01 is the fourth-largest active sector in this snapshot. It sits behind Productivity and Compliance, Dev Toolbench, and Commerce and Operations, but ahead of Design and Media, Communications and Community, Treasury and Ledgers, Security and Watchtower, and the rest of the archive's smaller rooms.

The comparison is clearer in counts:

• Productivity and Compliance: 2,131 records
• Dev Toolbench: 840 records
• Commerce and Operations: 795 records
• Artificial Intelligence: 733 records
• Design and Media: 539 records
• Communications and Community: 484 records

AI is therefore not a fringe category. It contains 106 more records than Design and Media, a familiar and durable software family. But it is also only about one-third the size of Productivity and Compliance. The archive keeps returning to the unglamorous work of approvals, routines, documents, reporting, and keeping teams coordinated. Software's oldest chores remain very well populated.

That does not make the AI category less consequential. It makes the category more specific. A directory label should describe the product's central job, not merely the technology mentioned most often on its homepage. If every model-assisted product were placed in Sector 01, the category would become a filing error disguised as a trend report.

For a closer look at the records assigned there, browse the Artificial Intelligence sector. The variation is the point. There are tools built around models, tools that package a narrow AI workflow, and products whose usefulness depends less on the model than on the surrounding interface, data, or operational process.

The pricing signal has a large blank space

The pricing labels in Sector 01 tell a second, less tidy story. Of the 733 AI records, 320 are marked free, 56 freemium, and 25 paid. That is 401 records with one of those explicit labels. The remaining 332 are undisclosed, almost 45.3 percent of the sector.

Among records with a disclosed label, free is the dominant model: 320 of 401, or 79.8 percent. Freemium accounts for 14.0 percent and paid products 6.2 percent. It would be tempting to turn that into a grand claim about the economics of AI software. The archive declines the temptation.

First, a listing's pricing label is an observation, not a complete billing-system audit. A product marked free may have paid usage limits, an enterprise plan, a trial, or a price page that changed after the record was collected. Second, undisclosed is not the same as free, paid, or absent. It is information the archive does not presently have.

That missing segment is useful data in its own right. Most AI-tool statistics pages focus on the split they can count. The ARCHIVE-9 view makes the unclassified portion visible: 332 records do not expose a usable free, freemium, or paid label in the archive. For a maker, that is a practical discovery issue. A visitor comparing tools cannot evaluate a price that is not clear enough to be filed.

THE PRICE MAY BE FLEXIBLE. THE ABSENCE OF A PRICE IS ALSO A SIGNAL.

If you maintain an AI product, this is a small but concrete maintenance test. Can a stranger understand the pricing path without opening a sales conversation or interrogating a chatbot? If not, the product may be easier to demonstrate than to compare.

Why the primary-category rule matters

The archive has no interest in pretending that AI is separate from the rest of software. That would now be visibly false. The important choice is more modest: whether a category tells a visitor what a product does.

Consider three hypothetical products. A tool that generates marketing copy might belong in Artificial Intelligence if the model interaction is the entire product. A CRM that adds a writing assistant probably belongs in Commerce and Operations. A code host with an AI review button is still a developer tool. Each uses related technology. Each solves a different primary problem for a different operator.

This classification rule is the part a general AI-tool directory count usually cannot show. A directory of AI-labelled products answers, "How many things call themselves AI?" ARCHIVE-9 asks a narrower question: "Which software records are principally AI products after the label is separated from the job?" The answer, currently 733, is deliberately conservative.

That conservatism makes comparison more useful. It avoids treating every new feature announcement as a new market category and preserves the route a visitor actually needs. Someone looking for product operations software should not need to search Sector 01 just because every serious operations product now contains some intelligence.

It also explains why a maintained record matters. A product can change category emphasis as it changes shape. Owners can bring a record online, correct its description, and make the practical details easier to inspect. An ARCHIVE-9 badge then gives the product's own site a route back to that maintained record.

A large room, with boundaries worth keeping

Sector 01 is substantial at 733 records, and its visible pricing labels lean heavily toward a free entry point. Both are useful observations. Neither supports the lazy conclusion that AI has swallowed software, or that every product with a free tier competes in the same market.

The more durable finding is the boundary. AI is large enough to stand on its own in the archive, but the archive is still mostly software built around other jobs. Productivity, development, commerce, design, communications, finance, security, training, and data work remain distinct destinations even when models now sit inside their machinery.

For people evaluating tools, that is good news. A useful directory does not make a trend disappear. It puts the trend in the right drawer.

Method note

All figures in this report are ARCHIVE-9 data, queried read-only on August 28, 2026 from the deep_memory archive. The total is 8,497 records. Sector counts use each record's primary cat value. Pricing counts use the archive's pricing field for Sector 01, with FREE, FREEMIUM, PAID, and UNDISCLOSED treated as distinct recorded values. These are archive observations, not estimates for all software companies or all AI products.

THE CATALOGUE REMAINS OPEN. THE LABELS REQUIRE MAINTENANCE.