Casselini Inc., Six Months In: Extending the Operating Headroom Without Expanding the Team

The conventional read of a search-fund acquisition is “keep the brand, continue the operation.” Six months into Casselini Inc.’s new chapter — closed December 2025 under Rocketstar’s search-fund acquisition, with KITAGATA installed as President and CEO — the surface looks like that story. The brand is intact: the same five in-house labels, the same Jingumae flagship, the same collaboration cadence with hanky panky, New Era®, HARUTA, and others continuing without interruption.

Underneath, something quietly additive is happening. The brand is what we keep. The operating headroom is what we’re extending.

This essay is an interim report from inside that transition — what becomes possible when AI is treated not as a feature but as the next infrastructure layer, and what a 39-year-old fashion company can attempt when the backbone underneath it is set up to be easy to change.

An occasional piece on this site. KITAGATA also writes a regular Japanese-language column on note.com — the ongoing “Next OS for Specialty Apparel” series — exploring how Japan’s specialty apparel companies are rewriting their underlying systems for the next decade. See note.com/kitagata.

1. The weight starts coming off

The first thing that changes when AI gets wired into operations is not productivity. It’s felt weight.

A monthly budget review that used to take three people two days now takes one person an afternoon. A vendor reconciliation that absorbed the accounting team’s entire month-end now resolves in hours. A weekly sales-and-traffic analysis that depended on someone exporting CSVs from four systems and aligning them in Excel now writes itself.

None of these are dramatic line items. None of them show up in a press release. But collectively, they remove the kind of cognitive load that ages a company’s people. The MD team starts the week thinking about product, not about whether the numbers reconciled. The accounting team has time to look at the story the numbers tell, not just whether they balanced.

What I notice from inside, as the executive responsible for the direction:

• Decisions arrive earlier in the week. When the analysis isn’t bottlenecked, the discussion isn’t either.
• The phrase “wait for the data” has mostly disappeared. The data is already there.
• The team’s energy goes into judgment, not into preparation for judgment. This is the change that matters most.

The clearest evidence isn’t a system rollout or an executive directive. It’s that our merchandising lead — not the AI champion, not the analytics team, but the operating MD — started using the internal AI tools on her own initiative. Her seasonal-planning decks, the ones that used to take two or three days at the front of each buying cycle to assemble, now draft themselves with her direction.

The time that opened up went straight into the conversation that actually drives the business: what more can we do to grow revenue? Instead of into the assembly of the materials that supported the conversation.

That progression — the AI assistant adopted by the people who would benefit, not the people who would champion it — is the surest signal that the integration is real, not theoretical.

The mechanism is unremarkable in 2026: API-connected business data, an internal AI assistant trained on the company’s specific vocabulary and product taxonomy, and a discipline of routing routine analysis through it rather than around it. The implementation is technical. The result is cultural.

2. The headroom expands

The slower change is the one that pays off over longer horizons: bringing the operational backbone into a configuration where every workflow improvement is cheap to try.

The back-office systems that companies of Casselini’s generation grew up on did their job — keeping the catalog, the inventory, the orders, and the wholesale layer aligned through decades of growth. The trade-off built into that generation of architecture is that modification cost is high: a new field, a new workflow, a new report each carries an external quote in the hundreds of thousands of yen and lead times measured in months.

For a company optimizing inside steady patterns, that trade-off was a fair one. For a company actively widening its surface — adding channels, testing product configurations, deepening the loop between merchandising and demand — the calculus shifts in favor of a backbone built to be modified often.

What we’ve been building since the new chapter began is PMS: a Python-based, cloud-native product management system, developed in-house. The architecture:

• Google Cloud Run + PostgreSQL for the application and database layer
• Identity-Aware Proxy (IAP) with company-domain access control
• A small internal team plus an AI co-pilot doing the actual build

Phase by phase, PMS has been taking on the responsibilities that anchor the operation: product master data (1,107 SKUs now native to the new backbone), sample-progress tracking, payment-term logic, and the connection scaffold to NE — the order-management layer that bridges into the EC marketplaces.

Two qualities of a cloud-native, self-built backbone reshape what the company can attempt:

1. Iteration speed expands. A new field, a new workflow, a new report — the question shifts from “what will the vendor quote?” to “how fast can we ship it?” Days, not months. That tempo opens experiments the old tempo simply never allowed.
2. The unit economics of change collapse. Variable hosting cost in the tens of thousands of yen replaces fixed license cost in the millions. The arithmetic is favorable, but the bigger point is qualitative: the cost-per-improvement falls far enough that small ideas are worth trying — which is exactly the soil that operational innovation grows in.

The backbone is not done. We’re at Phase 5.1, with EC integration as the next major step, and Phase 6 (in-store POS via Smaregi) in pilot. But the proof point has already landed: the company can run substantial parts of its operations on infrastructure it owns, at a fraction of the prior cost, with iteration cycles measured in days.

The team building it is small. The AI co-pilot makes that feasible.

What expands here is not just speed or cost — it’s the headroom for what the company can attempt next.

3. Growth without headcount

The third change is the one that matters most for the company’s next chapter: scaling channels without scaling headcount.

The traditional retail-growth playbook in Japanese apparel is linear. Add a store, add a store manager, two staff, a portion of head-office support functions. Add an EC marketplace, add the people to manage that marketplace’s listings, pricing, customer service, returns. Each new channel arrives with a proportional cost in people.

That playbook is not what we’re following.

The PMS we’re building is designed around a different assumption: the company’s product master is the single source of truth, and every channel — owned EC, Rakuten, ZOZO, Amazon, eventually overseas marketplaces — pulls from it via API. The marketing of a product, the pricing of it, the inventory state of it, the post-purchase data from it: all flow through one backbone, and the channels are thin layers on top.

In practice, this means:

• New EC marketplace integration is a configuration task, not a hiring decision. When we add an additional sales channel, the team that manages it is the team that already manages the existing channels — because the master, the pricing logic, and the order flow are the same.
• New physical stores can be added with linear retail cost, not linear back-office cost. The point-of-sale integration (Smaregi, Phase 6) shares the same inventory and loyalty master as EC. A new store does not need its own analyst, its own inventory clerk, or its own dedicated marketing coordinator.
• The marginal cost of a new store has been quietly redefined. We are now in a position to plan store openings that, under the old backbone, would have required hiring three to five additional people in head office to support. Under the new backbone, they require zero or one.

This is what AI-assisted operations and cloud-native infrastructure compound into. Not a smaller company. A more capable one, at the same size.

4. What “inheriting a brand” actually means

Most coverage of search-fund acquisitions treats the succession as the story. New CEO, new board, sometimes new strategy. The brand continues, the operations continue, and the question is whether the new leadership can grow what already existed.

Six months in, my honest read is that the conventional framing under-describes the opportunity. The succession is the starting position; the work is extending the headroom on top of what was inherited. A 39-year-old Tokyo apparel company carries 39 years of accumulated value — the motif vocabulary, the production-partner relationships, the wholesale standing, the brand recognition in the Harajuku-Omotesando corridor and across Japan’s department-store and select-shop network. Succession is the moment to take that as the foundation and ask: what becomes possible when the backbone underneath it is set up for what’s next?

The work of the first six months has been to start drawing that line — to identify what’s worth keeping intact, and where the room for expansion is.

What we keep — because it’s the source of the brand’s value:

• The 39-year motif vocabulary and the collaboration discipline that has produced capsules with Care Bears™, WICKED, FILA, graniph, and many others
• The production-partner relationships built and renewed across decades in India, the Philippines, Vietnam, Thailand, China, and Japan
• The five-label portfolio that gives the company natural coverage from sculptural daily bags to streetwear utility

What we extend — because the headroom is real and the unit economics are now ours:

• The systems backbone, modernized to a cloud-native architecture we own and can change
• The analytical workflow, AI-assisted from the bench-level outward
• The channel architecture, built around a single source of truth that scales linearly in retail/EC, not in head-office headcount

What I tell our team — and what I would offer anyone else inheriting a heritage brand in this market — is this:

The brand is the thing you keep. The operating headroom is the thing you extend. The brand carries 39 years of value forward — what we’re building underneath is the room for the next chapter to take it further.

5. What the next six months look like

The first half of the new chapter has been about getting the foundation in place. The second half is about putting weight on it.

Concretely, the next six months include:

• Completing the EC integration backbone — full API-driven product-master distribution to owned EC, Rakuten, and ZOZO, with Amazon scoped for the cycle after
• POS pilot conversion — moving from the Smaregi trial to production deployment, unifying retail and online loyalty into a single customer view
• AI-operations expansion — extending the internal AI co-pilot’s reach into MD planning, inventory signals, weekly performance attribution, and the longer-horizon forecasting that has historically required external consultants
• AEO/GEO maturation — the company’s AI-visibility program, documented in its own baseline elsewhere, reaching a measurable second milestone as the brand’s presence in generative search becomes legible to the models that increasingly mediate how customers research what to buy

These are not separate initiatives. They are different surfaces of the same underlying decision: that the next chapter of Casselini Inc. will be built on a backbone the company owns, runs, and can change.

The brand stays. The operating system changes. That is the work.

---

— KITAGATA is President and CEO of Casselini Inc. (株式会社キャセリーニ), and the author of Sell Through: The Five-Step System for Building a High-Margin Apparel Brand (Amazon Kindle, 2026). Previously Vice President, WEGO (2004–2017), and President & CEO, LeSportsac Japan (2018–2025, acquired by Itochu Corporation).

---

About this writing

This essay is part of an occasional English-language journal on modetokio.com — pieces appear when the underlying work suggests one is worth writing, not on a fixed schedule.

For more regular reading, see KITAGATA’s Japanese-language monthly column at note.com/kitagata — the ongoing “Next OS for Specialty Apparel” series, exploring how Japan’s specialty apparel companies are rewriting their underlying systems for the next decade:

• June 2026 — AI を業務に組み込めないアパレル企業は取り残される (“Why apparel companies that can’t embed AI will be left behind”)
• July 2026 — アパレルは「特権」ではない (“Apparel is no longer a ‘privilege’”)
• August 2026 — キャセリーニ承継後 6 ヶ月 — 経営の伸び代を広げる (the Japanese-language counterpart to this piece)

The English essays on modetokio.com and the Japanese column on note.com run in parallel rather than as translations — each is shaped for its audience.