The Second Screen Is a Graveyard: Why Your Client Calls Keep Going Sideways
Every cross-border seller I know has the same dirty secret: the most expensive hour of their week isn’t spent in Amazon Seller Central or tweaking a TikTok Ads campaign. It’s the client call — the weekly sync with the brand owner, the investor update, the tense negotiation with the 3PL partner who just lost another pallet. We all walk into those calls with a browser full of tabs: the P&L, the inventory forecast, the last email thread, the ad account dashboard. And we all lose. Not because we don’t know the numbers, but because we can’t find the right number in the moment. The question comes, we fumble, we say “let me circle back,” and the other side hears “I don’t actually know.” That gap between what you know and what you can surface under pressure is where margin dies. It’s also exactly the problem Savvy — a new open-source Mac app from maker Jaume Alavedra — is trying to kill. And while it’s pitched at consultants and account managers, the mechanics matter more to anyone running a cross-border operation, where the documents are scattered across time zones, the stakes are higher, and the client on the other end is often working from a completely different cultural playbook.
I don’t care about another meeting summarizer. We have enough tools that turn a 45-minute conversation into a tidy bulleted list that nobody reads. The market is saturated with AI notetakers that join your call, transcribe everything, and then email you a summary that lands in the same folder as the last eleven summaries you ignored. That’s archival, not assistive. It helps you build a record of what happened, but it does nothing to help you win the conversation you’re currently losing. The insight buried in this launch is different: the problem isn’t that you lack information. It’s that your information is disorganized, and your tools are passive. Savvy tries to flip that dynamic by making the AI an active participant — one that speaks up only when it has something genuinely useful to say, and only when it can cite its source.
The Problem: Your Knowledge Base Is a Museum, Not a Weapon
Let me paint the scenario that every Amazon FBA operator with a private-label brand will recognize. You’re on a call with a supplier in Shenzhen about a 20% cost increase on your hero SKU. You’ve got the contract open in one tab, the email where they quoted the original price in another, and a spreadsheet with your landed cost breakdown in a third. The supplier says, “We never agreed to a fixed price for Q4.” You know that’s wrong — you have the email — but you spend 30 seconds scrolling, losing eye contact and momentum, and by the time you find it, the psychological advantage is gone. They saw you search. They know you weren’t sure. That’s the moment the negotiation shifts.
Savvy’s approach is to preempt that fumble. You point it at a folder of documents — contracts, briefs, price lists, email exports — and it builds what the maker calls a “versioned brief per client.” During the call, it listens. It stays quiet, which is the correct default. Then it speaks up for exactly three reasons: the other side asked something your brief already answers, someone crossed a red line you set, or you pressed the Advice button. Every card it surfaces cites the document it came from, so you can verify before you commit. That’s not a summarizer. That’s a second set of eyes that never blinks and never forgets which version of the agreement was current.
For a cross-border operator, this is the difference between walking into a negotiation with a binder and walking in with a legal team. The cultural friction alone — the way “we’ll discuss later” means different things in different business cultures — makes real-time, sourced recall a competitive weapon. When you can instantly pull up the exact clause from the exact contract that says “MOQ adjustments require 60 days written notice,” you’re not just winning an argument. You’re changing the entire tenor of the relationship. You’re signaling that you are the one who keeps the records, and that has outsized power in any cross-border dynamic where one party assumes the other isn’t paying attention.
How Savvy Actually Works — and Why the Architecture Matters
The technical choices here are more interesting than the feature list. Savvy runs locally on your Mac — Apple Silicon, macOS 13 or later. The files, indexes, and transcripts stay on your machine. Only excerpts and the audio stream leave it, which is a meaningful privacy posture for anyone dealing with supplier contracts or unannounced product plans. It’s also open source, MIT-licensed, with the code available on GitHub. That’s a double-edged sword, and I’ll get to the downside in a moment, but for now, let’s appreciate what the local-first architecture buys you.
The privacy angle is real. If you’re a DTC operator running a brand on Shopify and Amazon, your documents contain your entire moat: your supplier costs, your ad metrics, your launch calendar. The idea of piping that through a cloud-based meeting assistant that’s training its models on your data should keep you up at night. Savvy’s local-first design means the sensitive stuff — the index, the transcripts — never touches a third-party server. The audio stream does leave, presumably for transcription, but the extracted knowledge stays local. For a seller who’s had a supplier relationship go sideways and doesn’t want that history floating around in some AI vendor’s training set, that’s a meaningful differentiator.
The “versioned brief” concept is also smarter than it sounds. This isn’t a static knowledge base. It’s a living document that tracks changes over time. When your client sends a new contract revision, or your supplier updates their price list, Savvy rebuilds the brief. The “versioned” part matters because in cross-border deals, the current version is often the entire battleground. I’ve seen deals fall apart because one party was operating off a superseded email while the other was referencing a newer one. A tool that automatically tracks which document version is live — and can cite it during a call — eliminates an entire class of dispute.
### Why Amazon Sellers Should Care More Than Shopify Ones
Here’s a contrarian take: this tool is more valuable for Amazon FBA operators than for Shopify DTC brands, even though the Shopify crowd is usually the early adopter. Why? Because Amazon sellers live and die by documents that are adversarial by nature. Your supplier agreements, your brand registry filings, your appeal letters to Seller Performance — these are all documents where the other side is trying to find a discrepancy. When Amazon suspends your listing and you’re on a call with an appeals specialist, you need to cite your invoice, your certificate of analysis, your letter of authorization — instantly, accurately, and without hesitation. That’s not a “nice to have.” That’s the difference between getting reinstated in 48 hours and losing a month of sales.
Shopify sellers, by contrast, have fewer of these high-stakes document-dependent calls. Their conversations are more about marketing strategy and creative performance, where the source of truth is a dashboard, not a contract. The real-time citation feature matters less when the answer is “our ROAS dropped because of the iOS update” versus “per section 4.2 of our agreement, you cannot renegotiate freight terms mid-season.” If you’re an Amazon seller with a complex supply chain and a history of disputes, this tool is aimed squarely at your pain. If you’re a Shopify seller who mostly talks about creative testing, you might find it overkill.
What Cross-Border Sellers Can Borrow (Even If You Never Install It)
Let’s step back from the tool itself. Even if you decide Savvy isn’t for you — and I’ll get to the reasons it might not be — there are three operational lessons here that any cross-border operator should steal immediately.
First, pre-load your briefs before high-stakes calls. The habit of assembling a folder of relevant documents before the conversation, not during it, is the core discipline. Most of us walk into calls with a messy browser and hope. Savvy forces you to be deliberate: point it at a folder, let it build the brief, and then trust it during the call. You can replicate that discipline with a simple folder structure and a note-taking app, even if you don’t use the AI. The act of curating your documents in advance is itself the competitive advantage.
Second, define your red lines explicitly. Savvy’s concept of “red lines” — conditions that trigger an alert during the call — is a brilliant negotiation hack. Before any call with a supplier or a marketplace rep, write down your non-negotiables. Is it the MOQ? The payment terms? The shipping window? If you can articulate those in advance, you’re less likely to get talked out of them in the heat of the moment. The tool enforces that discipline, but the discipline itself is the skill.
Third, demand citations. The fact that every Savvy card cites its source document is a reminder that AI output without provenance is worthless in a negotiation. When you use any AI tool — whether it’s Helium 10 for keyword research or Klaviyo for email copy — you should always ask: where did this come from? Can I verify it? In cross-border deals, where the other side will challenge every claim, unverified AI output is a liability. Savvy’s citation-first design is the right model for any AI tool you deploy in a commercial context.
Where the Math Breaks: My Honest Skepticism
Now let me be the skeptic in the room. Savvy is a 1.0 open-source tool from an individual maker. That comes with real limitations, and pretending otherwise would be malpractice for a blogger who’s seen too many promising tools die on the vine.
First, the local-only constraint is a scaling problem. Running on your Mac is fine for a solo consultant with one client folder. But a cross-border operation with multiple brands, dozens of suppliers, and a distributed team needs collaboration. Can your VA in Manila access the same brief? Can your operations manager in Shenzhen update the red lines? As far as I can tell from the launch page, the answer is no. This is a single-operator tool, and that caps its utility for anyone running a team. You’d need to export briefs and share them manually, which reintroduces the version-control problem the tool is trying to solve.
Second, the audio stream leaves the device. The maker is transparent that “only excerpts and the audio stream leave it.” That’s a compromise. For transcription, the audio has to go somewhere. If you’re discussing unannounced product specs or pricing strategies with a supplier, you’re trusting the transcription service’s data handling. It’s better than sending everything to the cloud, but it’s not zero-risk. For sellers dealing with truly sensitive IP — a new formula, a proprietary sourcing strategy — this might be a dealbreaker.
Third, the “three reasons to speak up” model is only as good as your brief. If you point Savvy at a folder of disorganized documents, it will build a disorganized brief. Garbage in, garbage out, as always. The tool doesn’t magically understand your business context. You have to curate the input folder carefully, which again requires the discipline most of us lack. The maker’s own framing — “the answer was usually sitting in a document I had already read twice” — is both the insight and the limitation. The tool assumes you’ve already read the documents. It’s not doing the reading for you; it’s just helping you recall faster.
Finally, there’s the maintenance burden of open source. MIT-licensed software from a solo maker is a gamble. Will it be updated for the next macOS? Will it support new transcription models? Will the creator lose interest in six months? For a business-critical tool, that’s a real risk. You’re not just adopting software; you’re adopting a dependency on one person’s hobby. I’d want to see a more active contributor base or a commercial entity behind it before I’d trust it with my core negotiation workflows.
### Where the Math Breaks: The Real Cost of “Free”
Let me dig into the open-source angle a bit more, because “free” is never free for an operator. The MIT license means you can use it, modify it, and even sell it — but it also means there’s no support, no SLA, and no guarantee of future development. For a solo seller, that’s fine. For a brand doing $10M a year in revenue, relying on a hobbyist’s GitHub repo for your negotiation prep is a risk you shouldn’t take.
The counterargument is that open source gives you control. If the tool breaks, you can fix it — if you have the engineering talent. Most cross-border sellers don’t. They’re marketers and operators, not Swift developers. So the practical reality is that Savvy is a great tool for a specific persona: the technical founder who’s comfortable in a terminal and has a manageable number of client relationships. For the broader seller population, it’s a proof of concept — a compelling demonstration of what real-time, cited AI assistance could look like, even if the implementation isn’t ready for enterprise-grade deployment.
That’s not a dismissal. It’s a positioning. The tool is valuable as a template for what the next generation of meeting assistance should be. The question is whether a commercial product will emerge that takes this architecture and makes it accessible to non-technical operators. If I were a betting man, I’d say yes — and I’d be watching the AI meeting assistant space closely for someone to build a hosted version of this exact concept.
What I’d Watch / Test Next
Here’s what I’d actually do this week, as a cross-border operator, in response to this launch. I wouldn’t install Savvy on my main machine and trust it with my entire client portfolio on day one. But I’d run a disciplined experiment.
First, take one high-stakes recurring call — the weekly supplier sync with your most difficult partner — and build a document folder for it. Export your relevant emails, contracts, and price lists into a single directory. If you’re technical, install Savvy and point it at that folder. If you’re not, replicate the workflow manually: create a “brief” document that summarizes the key facts, the red lines, and the open questions, and keep it open on your second screen during the call. The goal isn’t the tool; it’s the habit of walking in prepared.
Second, define your red lines for that specific relationship. Write down the three things that, if they happen, should immediately change your posture. Share them with your team if you have one. Use them as the framework for how you listen during the call. This is a skill that transfers regardless of what software you use.
Third, watch the Product Hunt page for forks and derivatives. Open-source projects that gain traction often spawn commercial spin-offs. If you see someone building a hosted version with team collaboration, that’s the one to watch. For now, treat Savvy as a signal — a sign that the market is moving from passive transcription to active, cited assistance. The tool itself might not be your final answer, but the direction it points is where the industry is heading.
Finally, test the citation feature on a low-stakes call with a friendly partner. See if the sourced cards actually help you respond faster. If they do, you’ll know the concept works — and you’ll be primed to adopt the commercial version when it arrives. If they don’t, you’ve lost an hour of setup time, which is a cheap price for clarity on a tool that could reshape how you handle every negotiation going forward.
The second screen doesn’t have to be a graveyard. It can be a weapon — if you load it right. Savvy is a reminder that the problem isn’t the AI. It’s the discipline. And that, at least, is free.
