The operator’s real bottleneck isn’t ads — it’s the 40 minutes a day you lose retyping things
Cross-border sellers love to talk about stack consolidation as if it were a finance problem: too many SaaS subscriptions, too many logins, too many invoices to reconcile at quarter-end. That’s the small version of the problem. The bigger version is cognitive overhead. Every time you alt-tab from Amazon Seller Central into a meeting notetaker, then into a dictation app, then into an AI assistant to draft a supplier email, you’re paying a tax in context-switching that no P&L line item captures. That’s why I paid attention to Koi, a Mac-only “ghost on the edge of your screen” from maker Irakli that collapses dictation, screen-aware AI assistance, and call recording into three keyboard shortcuts. It is not a seller tool. It is not even close to a seller tool. But the design pattern underneath it — one hotkey, one local-first data store, one bill — is exactly the pattern I wish more e-commerce SaaS founders would steal.
Here’s the thesis: cross-border operators don’t need another dashboard. They need fewer surfaces. The average seven-figure Amazon FBA brand I talk to is running somewhere between eleven and nineteen paid tools, and the marginal tool almost never earns its seat. What earns its seat is anything that removes a manual keystroke from a workflow you repeat fifty times a day — supplier follow-ups, listing copy revisions, compliance notes from a call with a freight forwarder, a quick reply to a TikTok Shop affiliate. Koi is aimed at a general Mac productivity buyer, but its architecture is a useful mirror for anyone building or buying tools for sellers. Let me walk through what it actually does, where it sits against the incumbents, what a seller should borrow from it, and where I think it stumbles.
What Koi actually solves, stripped of the launch-page gloss
The maker’s own framing is refreshingly honest: he was paying for a dictation app, a meeting notetaker, and an AI assistant, and jumping between all three all day. So he built one Mac app that does all of it. That’s the whole pitch, and it’s a good one because it’s a real workflow complaint rather than a feature list dressed up as a vision.
Concretely, there are three shortcuts. Fn + ⌥ gives you dictation: you talk, and your words type into any app, running locally on your Mac, offline, unlimited. Fn + ⌘ is the “talk to Koi” mode, where the assistant sees your screen, points to the right button, draws things out, and rewrites your text in place. Fn + ⌃ records any call — no bot joins the meeting — and produces notes and to-dos afterward. Everything lands as pages on your Mac, and the same app bundles a calendar, an expense tracker, and a “Shift wheel” that lets you drag a picture, hold Shift, and crop or compress it.
That last cluster is where I’d push back hardest, and I’ll get to it. But the core three shortcuts map onto three genuinely high-frequency seller tasks. Dictation into any app means you can dictate a supplier email inside Gmail, a listing bullet inside Amazon Seller Central, or a customer reply inside Shopify’s admin without ever leaving the tab. Screen-aware assistance means you can say “rewrite this product title for the German marketplace” while looking at the listing. Call recording with no bot joining means you can capture a negotiation with a Shenzhen supplier without the awkward “Koi Notetaker has joined the meeting” notification that spooks half of overseas vendors.
Why Amazon sellers should care more than Shopify ones
Shopify operators live in a browser and a handful of apps that already have decent native AI — Shopify Magic writes product descriptions, and Klaviyo’s AI handles a lot of email drafting. The marginal value of a screen-aware assistant is real but incremental.
Amazon sellers are different. Seller Central is a hostile interface. Bulk listing edits happen in flat files. Case logs with Seller Support are painful to search. A screen-aware assistant that can look at a flat-file error report and tell you which column is broken, or read a policy warning and draft an appeal, is worth materially more than the same assistant pointed at a Shopify product page. If you sell on both, the Amazon side is where a tool like this earns its keep first.
How it stacks up against the incumbents you’re probably already paying for
The honest comparison set is not “other AI assistants.” It’s the specific tools Koi is trying to replace, plus the platform-native options that come free with subscriptions you already have.
On dictation, the obvious incumbent is Wispr Flow, which has become the default Mac dictation layer for a lot of operators I know, and on the Windows side there’s Superwhisper and the built-in macOS dictation that most people forget exists. Koi’s differentiator here is that dictation runs locally and offline and is unlimited — no per-minute metering, no cloud round-trip. That matters if you’re dictating supplier terms or pricing strategy and would rather not ship your voice to a third party.
On meeting notes, the incumbents are Granola, Otter.ai, and Fireflies.ai. Almost all of them work by joining your call as a bot or by capturing system audio. Koi’s pitch — “record any call, no bot joins” — is a genuine differentiator for cross-border operators, because a bot joining a call with a factory rep in Guangzhou or a freight forwarder in Rotterdam is a trust problem, not just a UX problem. Whether the recording quality holds up on a WeChat or WhatsApp call is not disclosed, and that’s the first thing I’d test.
On the AI assistant layer, the comparison is ChatGPT, Claude, and the growing pile of screen-aware desktop agents. Koi’s edge is that it’s ambient — it lives at the edge of your screen and can see what you see without you copy-pasting context. Its weakness is the same as every ambient assistant: an allowance on “talking to Koi,” which means the thing you’ll use most is the thing that’s metered.
Pricing is the loudest part of the launch. There’s a 3-day trial that requires a card, and a Lifetime deal from $59 while early spots last. Against a stack of Wispr Flow plus Granola plus a ChatGPT subscription, $59 lifetime is aggressive — almost suspiciously so. I’d want to see what “while the early spots last” actually means in practice before I’d build a workflow on it.
What cross-border sellers should borrow from this — even if they never install it
This is the section I actually care about, because most of you reading this are not going to switch your dictation app this week. But there are three design lessons here that map directly onto how you should be building your own internal ops.
Lesson one: local-first is a compliance feature, not a privacy slogan
Koi keeps dictation on your Mac and stores everything as pages locally. The maker is explicit that when you talk to Koi, your voice and screen go to the AI provider, but dictation stays on-device. For a cross-border seller, that distinction is not abstract. If you’re dictating margin floors, supplier cost sheets, or ad spend targets, you do not want that audio sitting in a vendor’s cloud with unclear retention. The same logic applies to how you store your own ops data: if your sourcing notes and landed-cost models live in a shared Google Drive that every VA and agency has access to, you have a compliance problem dressed up as convenience. Local-first, or at least access-scoped, is the right default for anything touching cost or contract terms.
Lesson two: the three-hotkey model is the correct UX for repetitive ops
Most seller tools fail because they require you to go to them. You have to open Helium 10, or log into Jungle Scout, or navigate to the Klaviyo flow builder. Koi’s insight is that the highest-value tools come to you, in the app you’re already in, via a shortcut you can build muscle memory around. If you’re evaluating any new tool this quarter — a repricer, a review automation layer, a returns portal — ask one question: does it meet me where I already am, or does it demand a new tab? The ones that demand a new tab almost always get abandoned by month three.
Lesson three: bundling adjacent chores beats best-of-breed for small teams
The calendar, expense tracker, and Shift wheel image tool are, on their own, mediocre versions of tools you already have. But for a two-person DTC team that doesn’t want to pay for a separate expense app, the bundling is rational. The lesson for sellers is not “buy bundles.” It’s that the threshold for a tool earning a seat on a small team is lower if it removes an entire category of subscription, even at the cost of being worse at any single job. That’s the same logic behind why so many sellers end up on all-in-one platforms like Shopify plus a single fulfillment partner rather than a dozen point solutions.
Where my judgment says it falls short
I want to like Koi more than I do, and the reasons I don’t are specific.
Mac only. This is the biggest one. A huge share of cross-border ops teams run Windows, especially in China, Vietnam, and Eastern Europe where the sourcing and VA side of the business actually sits. A Mac-only tool is a tool for the founder, not the team. If you’re an Amazon seller with a sourcing agent in Shenzhen on a Windows laptop, Koi does nothing for them, which means the workflow you build around it is founder-shaped and doesn’t scale.
The allowance on “talking to Koi.” Dictation and recording are unlimited, but the screen-aware assistant — the most valuable part — is metered. That’s a classic freemium trap: the thing you’ll lean on hardest is the thing that will bill you. For an operator doing fifty listing rewrites a week, that allowance will get hit fast, and the pricing page doesn’t disclose what happens after.
The card-required trial. A 3-day trial that requires a card is a conversion tactic, not a trust signal. For a tool that wants to live at the OS level and see your screen, I’d want a longer no-card trial before I let it watch me work.
The data flow on the assistant mode. The maker is upfront that voice and screen go to the AI provider when you talk to Koi. That’s honest, but for a seller handling supplier contracts or customer PII, it’s a real consideration. Dictation staying local is good; the assistant mode is a different risk profile, and the launch page doesn’t name the AI provider.
The bundle dilution. The calendar, expense tracker, and image cropper feel like scope creep. They make the app harder to explain and harder to trust for the core job. I’d rather see a tighter tool that does three things excellently than a Swiss Army knife where the blade you actually need is metered.
Where the math breaks
Run the numbers against your own stack. If you’re currently paying for a dictation app at roughly $10–15/month, a notetaker at $15–20/month, and an AI assistant at $20/month, you’re at $45–55/month, or roughly $540–660/year. A $59 lifetime deal looks like a no-brainer. But the lifetime deal is “while early spots last,” which means it’s a customer-acquisition price, not a sustainable one — and the metered assistant mode means your effective cost could climb. The math only breaks in your favor if you actually cancel the three incumbents and don’t replace them with something else. Most operators I know don’t cancel; they stack. That’s the trap.
What I’d watch / test next
If you want to pressure-test Koi against your own workflow this week, here’s what I’d do. First, install it on the Mac you actually work on — not a test machine — and run the 3-day trial against a real task: dictate a supplier follow-up inside Gmail, then use the screen-aware mode to rewrite a listing title while looking at Seller Central. Second, test the call recording on a real call with an overseas vendor over WeChat or WhatsApp, not Zoom, because that’s where the “no bot joins” claim actually earns its keep. Third, read the pricing page carefully and figure out what the assistant allowance actually is before you commit to the lifetime deal. Fourth, and most importantly, decide whether the tool is for you or for your team — because a Mac-only, founder-shaped tool is a personal productivity win and a team-standardization loss. If it’s the former, buy it and cancel two subscriptions. If it’s the latter, wait for the Windows build or the team plan, neither of which is disclosed. And if you’re a SaaS founder reading this, steal the three-hotkey model and the local-first dictation, and skip the calendar.
