Is Fashion Missing The Post-Interface Market Opportunity?

Despite AI fever still being in full swing, the technology sector seems to have settled into two predictable grooves for how to describe and deploy it. Startups and scale-ups built over the last three years or so get to use the “AI-native” label to describe their interface, their integrations, their data model, or all three. Pre-existing technology vendors bill AI as additive, giving users a new way to interact with data that was already there, or extending the capabilities of the tools that current customers were already using.

But this forked packaging-up of AI is, as we see it, also camouflaging a unique chance for much deeper introspection and an opportunity to scale the size of the addressable market for a broad cohort of fashion technology solutions, despite the evidence being visible in the headlines several times over the last couple of months.

As a case in point: this week the retail intelligence platform EDITED announced its official model context protocol (MCP) server, which provides an easy connector between AI chatbots like Claude (or custom whitelabel agents), and the platform’s library of market and competitive insights.

The launch is billed as giving existing platform users a way to “query retail data and get instant answers inside [their] AI tools”. Implicit in this posture is the realisation that people who would previously have spent some or all of their day with EDITED’s own user interface open are now spending more time inside Claude, ChatGPT, or other model harnesses, so the company clearly sees an opportunity to both retain those users and to enhance the value they obtain from the product by making it available wherever they go.

This is also the model that Stibo Systems (one of the most recognisable names in master data management) used when it unveiled its MCP server at the end of June, saying that the standardised connection was intended to meet “customers where they are building, without adding new governance risk or model dependency”. And while Stibo’s framing indexes more heavily on the use of MCP to bring its enterprise data layer to internal and external-facing systems and agents, the overall presentation is largely the same: connecting existing systems, data, and capabilities to AI agents is a way let existing users, either direct or API-level, do more with the software, services, and intelligence we provide.

And to underline the point even further: last week saw Adobe launch their eponymous plugin for ChatGPT, which gives non-registered users access to a very limited set of tools, but that opens up a wider suite of everyday capabilities to existing subscribers, allowing them to perform a reasonable brace of tasks in a natural language interface, and giving them the ability to access existing Creative Cloud projects and files.

The Adobe plugin announcement also contains more evidence that technology companies see MCP / plugin deployments as additive or as acquisition channels, following users to the surfaces they’re spending the most time in, or funnelling new users towards existing products and subscriptions.

“The Adobe plugin in ChatGPT is designed to meet you where you’re already working, making it easy to move from ideas to high-quality content inside ChatGPT. When your project calls for deeper creative exploration, advanced editing or pixel-level precision and control, Adobe’s apps are where you can push your ideas even further.”

Baked into that paragraph is the idea that text-first chat interfaces are a different class of products to desktop or web-native applications. But, so far at least, these different interfaces are pointed at the same objective: extending the footprint of the existing license-based software selling paradigm.

But what if the MCP user was just that: a new class of user who only ever interacts with a tightly-curated feature set or a selected library of data, and who rarely, if ever, converts to becoming a fully-fledged user with a traditional “seat”?

This is an uncomfortable question for technology providers to ask  – partly because it blurs a lot of the lines between backend and frontend development work, but primarily because it challenges one of the core assumptions of selling SaaS tools, which is the idea that, no matter how many different tiers of licenses you sell, all of them are sold to people who will, at some point or another, sit down and interact with the core product.

We’ve previously referred to this as the “headless software” question, named after the eCommerce, CRM, CMS, and other platforms that are content to provide backend databases, tools, and capabilities that companies then build their own interfaces for. But a more accurate and up-to-date name for it might be the “interface introspection,” and it’s a process that readers of this analysis can try for themselves right now: if you create software, mentally strip away the entire interface (not just parts of it) and describe what you sell. 

For some companies the answer will be almost nothing; for design tools, collaborative workspaces, and similar solutions, the interface is the product, and it’s effectively impossible to separate the authorship from the UI.

For other companies that trade in data, information, insights, or intelligence, the answer is likely to be that most of their offer is still intact, and that the interface is secondary.

Then, for the kinds of solutions that occupy the middle ground between visual tools and panes for the presentation and manipulation of alphanumeric data, the answer will probably be conditional. There will be “write” roles where data entry, creation, and authorship would be markedly worse or harder to perform in a chat interface, and then there will be “read” roles where the user mainly extracts information and seeks answers – both of which could be replicated in a textbox that returned text, figures, tables, and graphs the way the EDITED MCP does.

For interface-first companies, the market opportunity doesn’t, at least not today, extend much further as a result of MCP access. These kinds of companies may sell some additional seats, or token buckets, because generative tools become easier for other teams to access, but realistically The Interline expects that most creative work that’s done on a node-based canvas will continue to be performed there, by users paying very similar license costs.

For data or enterprise systems companies, the opportunity is potentially much more interesting, because there is, in essentially every organisation, a much larger market for users who only need slimmed-down “read” access to information or features. And if those users are sold a dedicated license tier at a cost below the current strata that solution providers currently sell, then these could become a lucrative new slice of the userbase.

Consider this: would you, as a technology vendor, rather sell a hundred seats at $40 per month that include MCP access, or would you rather sell a thousand seats at $10 per month that were MCP-only? For intelligence and functionality-first companies, the work involved in serving both classes of users would be comparable, but the value would be significantly higher.

And this is working off the assumption that this is an either / or situation. In reality there’s no reason to assume that AI-only licenses would cannibalise existing “pro” or “desktop” class seats. At worst, users who start with MCP access would eventually move on to more comprehensive licenses, and their lifetime value would increase. At best, every AI license sold remains a net-new sale that would never have been realised under the traditional licensing model, because the user in question simply didn’t require a full seat.

The market has also shown a remarkable willingness to entertain pure usage-based billing, so there’s every chance that companies that currently position their MCP servers as extra value for existing users could also sell them, by the hour or by the token, to new users who operate before and after current customers in the product lifecycle.

It’s rare that one of these weekly analyses is so focused on an opportunity for technology vendors, because, despite being a largely optimistic publication, The Interline is not in the business of telling software providers how to make money. But in this instance, both technology providers and brand / retail buyers seem to be converging on the same idea… but just missing the opportunity to take it further.