Questions, answered.

KiKi is an AI category manager for fashion retail. Every morning it reads your catalogue at variant level — every SKU, size and colour — and then makes and executes the in-season decisions that follow: reordering winners before they run dry, deferring or cancelling POs stacking on top of slow movers, setting markdown depth and timing, moving stock between locations, and recovering demand lost to stockouts. It does not hand you a dashboard and wish you luck; it makes the call and owns the outcome.

Those tools are copilots. They forecast, they surface a number, and then they hand the decision back to your team — which means the work still lands on a merchandiser who has a few hundred other things to do. KiKi is an autopilot: it makes the call, executes it through your connected systems, and answers for the margin it puts back. The difference is not forecast accuracy, it is who does the work.

Sidekick is a generalist assistant inside the Shopify admin — good at answering questions and running admin tasks across any kind of store. KiKi is fashion-native and sits above it: it understands size curves, seasonal drops, returns behaviour and markdown cadence, and it operates at variant level continuously rather than when asked. They are not competing for the same job.

Two to three months. The first phase is human-in-the-loop: KiKi proposes decisions, your team approves or rejects them, and it learns your policies, guardrails and risk appetite from those choices. Once the approval rate stabilises, it moves to running 60–70% of seasonal SKUs autonomously inside the limits you set. The remaining 30–40% — range bets, brand calls, vendor negotiations — stay with your team by design.

No. You set the policies, the approval gates and the limits, and KiKi acts inside them. Every decision it makes is logged with its reasoning, so you can audit any action after the fact. You can pull any category, any style, or the whole system back to manual at any time without unwinding anything.

You register, we come back within two working days with what we need — usually read-only access to your Shopify data, or an export if you prefer. We then run the same read KiKi runs daily and hand you the findings: where margin went in markdown, what cash your slow movers are trapping, and the demand you lost to stockouts. It is free, it takes days rather than quarters, and there is no obligation on either side.

KiKi reads sell-through, stock positions, product and variant data, and returns. During the audit and the human-in-the-loop phase, access is read-only — nothing is written to your store. Write access is only enabled later, deliberately, for the specific actions you have approved, and always inside the policies you set. See our Privacy Policy for how enquiry data is handled.

Shopify is the core — Admin, Stocky, Flow, Markets and B2B. KiKi also connects across the wider stack: forecasting and planning tools such as Inventory Planner, Cogsy and Prediko; marketing and lifecycle platforms including Klaviyo, Postscript and Attentive; feeds and merchandising via Meta, Google and TikTok, Feedonomics, Search & Discovery and Rebuy; and returns and reviews through Loop, Returnly, Yotpo and Okendo.

On outcomes, not seats. There is a setup fee, a monthly retainer, and a share of the margin and dead-stock cash KiKi recovers. We are deliberately exposed to whether it works: if it does not put money back, the outcome-based component does not pay. The free audit is the right place to start — it shows you the size of the opportunity before any commitment.

Fashion and footwear brands with enough seasonal SKUs that no human can watch them all daily — in practice, from a few hundred variants upward. Below that, a good merchandiser with a spreadsheet is genuinely competitive. Above it, the volume of daily variant-level decisions exceeds what a team can cover, and that gap is where the margin leaks.

Four steps. We map how you trade today — systems, processes, calendar and guardrails — in about a week. We connect Shopify and your wider stack at variant level. We wire up the systems that execute: reorders, pricing, collection sort and feeds. Then we turn it on in human-in-the-loop mode. Eight to twelve weeks from start to 60–70% autonomy.

Every decision is logged with its reasoning, so a bad call is visible and reversible rather than buried. Policies and approval gates cap what it can do unsupervised — you decide the markdown floor, the reorder ceiling and which categories it may touch at all. During the human-in-the-loop phase nothing executes without your approval, which is precisely when its judgement is least proven.