AI’s Consultancy Era Is Just Around The Corner, And Platforms Are Making Big Strides With First Party Data

Is fashion about to have a groundhog day for service days?

The so-called “SaaS-pocalypse” has, in very short order, been through a few different sets of clothes. 

First, it referred to a sudden sell-off of enterprise technology stocks that was catalysed by the idea that, rather than being so heavily reliant on paying monthly (or annually-discounted) seat rates for cloud software, companies of all shapes and sizes were going to vibe-code their way out of license agreements.

In practice, that idea was premature, but not necessarily wrong. It would be naïve to say that AI has not changed the “build or buy” equation, but, so far at least, the actual successes have largely come from companies reducing their dependencies on bolt-ons, modules, and plugins, rather than successfully ripping and replacing bigger tranches of their enterprise tech estates.

That tearing-down could end up coming over time, though.

Then, the same word came to mean something related, but with a much narrower scope: the promise that it would be individual users, rather than corporates, that would be building their own software, driven by their own idiosyncratic use cases. 

The ongoing joke amongst both knowledge workers, and in the cohort of that particular kind of “prosumer” that pursues productivity and automation in their personal life (there are several of these here at The Interline), is that everyone wants to build their own dashboard to consolidate the output of multiple other systems, and that the future of software is bespoke only insofar as it represents a new aggregation layer on top of systems that people are already paying for.

There is an argument to be made that if you stack enough of these individual users together, what you end up with is a department of people that all wish their tools talked to one another better, and there is certainly the potential that these departments will Claude Code or Codex their way to brute force integrations (or at least aggregations) that the open market didn’t see the justification to build.

More than likely, though, these users’ needs will be served by technology companies offering new classes of “headless” licenses that allow people to query data from disconnect sources through connectors or MCP servers that all roll up into the oft-vaunted “super app” or “universal chatbox”.

Now, it seems like AI is coming for software as a service, not by disrupting its business model, but by borrowing the part of it that elicits the most sighs from customers, and potentially poaching the people whose time accounts for the biggest line item in implementation and ongoing maintenance: consultants, advisors, and “forward deployed” professionals of all shapes and sizes.

Earlier this week, the two frontier AI labs that are the most active in courting enterprise customers (Google, bizarrely, seems content for the time being to just shove Gemini into Workspace and let diffusion happen through sheer scale) were revealed, separately, to be working on joint ventures that would focus on building and / or acquiring service businesses that would, per Reuters, employ “hundreds of engineers and consultants to help companies put their AI models ​to work”.

To anyone with enough time in enterprise software selection, implementation, and support under their belt, this will be starting to sound very familiar: both on-premise perpetual and SaaS projects have a tendency to become progressively lopsided, where budgets are concerned, with the actual software component becoming a footnote to extensive service agreements that, despite promises of self-onboarding, seem to always become necessary.

It is not an exaggeration to say that one of the biggest beneficiaries of the SaaS boom has been management consultancy practices and digital transformation partners. A Salesforce report from 2019, which forecasted out to 2024, predicted that the ecosystem that would grow up around a single SaaS vendor would, over time, grow to dwarf the value created by the vendor itself.

This proved to be an underestimate, and a joint IDC / Salesforce report from just a couple of years later, in 2021 (so pre-AI) revised an already optimistic projection upwards to the point where “[the Salesforce] partner ecosystem […] will make $6.19 for every $1 Salesforce makes by 2026”.

When AI was added to the mix, in 2023, the methodology predicted that the Salesforce ecosystem would create “$2 trillion in business revenues and 11.6 million jobs” by 2028.

To underline the important part: the majority of that value, by a factor of 1 to 6, was to be created by services that existed because of the underlying software, but was not directly tied into software licensing.

This, in a nutshell, is the prime reason that a lot of fashion companies found the SaaS era in general to be something of a poisoned chalice. It’s unambiguously true that the software market for fashion has led to the explosion of businesses built on selling software that solves real strategic and operational challenges, everywhere from design to point of sale. But it’s also true that those businesses have supported an explosion several orders bigger of consultancy, advisory, and implementation service practices.

Over time, with “service days” overtaking software licenses in many fashion technology projects, the world tried to rebrand the consultant into a “forward-deployed software engineer” (a term pioneered by Palantir in 2020), but these were in-house, solution-focused consultants by another name – and they were distinct from independent consultants primarily because their playbook only needed to contain best practices for one specific solution.

Now, in AI, we’re hearing about the “forward deployed creative”. 

The Interline is not against this new, narrower branding, per se, because our own first-hand experience, combined with the conversations we’ve had with brands off-the-record, tells us that getting real value out of generative workspaces requires effort. But the truth remains that AI is proving to be less distinct from other enterprise software than we might have thought, because there is still a real need for people to step in and lead other people to value that clearly isn’t easy to find.

But as well as those new artistic roles, this week’s news suggests that fashion’s AI ambitions are also going to involve a lot of plain-old consultants. 

It’s also critical to point out that, as The Interline wrote just last week, the cost of using AI models – whether they consume and generate text, image, video, or other alphanumeric tokens – is also on the rise.  

So the hope, amongst the brands and retailers implementing new tools, is now that the value of AI won’t be eclipsed by the labour cost required to find it, or the token budget required to pay for it. And also that the creative empowerment potential of the technology won’t be counterbalanced by the creation of an even larger and more persistent management class that grows to eclipse the thing it was intended to enable and then get out of the way of.

Pinterest showcases an unexpected success story from AI implementation that builds on first-party data and decades’ worth of context.

Finally, this week, a Pinterest earnings call (not usually the readiest source of fashion technology analysis) revealed something of a lighthouse for both home-grown AI and adoption that appears to come with relatively little friction.

As picked-over by PYMNTS and CX Dive, the company’s Q1 earnings revealed ten consecutive quarters of double-digit user growth, which its leaders attribute to their pursuit of personalisation through AI.

That may not sound especially relevant to fashion, but on closer inspection it demonstrates both the importance of first-party data and context when it comes to delivering on personalisation, and the incredible investment of time and money required to obtain it. And Pinterest has also emerged as an unlikely example of just how far in-house AI development can actually go.

Talking about the pay-off Pinterest has seen from AI, CEO Bill Ready revealed that it relies on two context layers that are exclusive to the platform: more than ten years’ worth of visual search data (“Every search, click and save gives our AI more signal about who a user is and what they care about, which allows us to deliver more relevant and personalized experiences across the platform.”) and a proprietary taste graph that then correlates those signals.

And the company has, it now emerges, created its own image generation and editing model that has been trained, according to Ready, on Pinterest data alone. We’ve written previously, here at The Interline, about the future of programmatic visual content creation and optimisation, but in essentially every other instance, those automated workflows have relied on companies making use of closed-source, high-cost models like GPT Image 2 and Nano Banana. In Pinterest’s case, “build vs. buy” has been taken to its extreme, and the company has eliminated its exposure to the coming “token squeeze”.

As well as being instructive when it comes to how fashion should perhaps be thinking about its own consumer data stores and its own exposure to price-hiking from AI suppliers, it’s worth reminding ourselves of how Pinterest itself compares to the promise of agentic commerce – something that the earnings call also called out several times.

Based on Pinterest stats reported on and aggregated by Sprout Social in 2026, almost all Pinterest searches start from a product outcome rather than a brand. Unlike other product discovery engines, where a consumer might start with a particular product in mind, this creates an interaction surface that’s potentially far better-primed for advertising that aligns with subjective and semantic search.

By way of comparison – as cited by Bill Ready during the earnings call – around 40 billion Pinterest searches each month (50% of its overall search volume) are commercial or product-driven, while ChatGPT only has 2% of queries across its own vast userbase that are anchored in product search or shopping.

Practically speaking, The Interline does not expect apparel or footwear brands to build their own image generation models. But if readers take one conclusion away from this week’s analysis, it should be that owned data and exclusive context is likely to prove more valuable to the long-term success of AI initiatives than an army of consultants.