Latch is built for accountants buried in click-heavy close work. Instead of writing prompts, you do a task on screen and talk it through, including the parts no SOP captures: the three-day matching window, the rate you always override. Latch asks about the exceptions, then builds a sidekick that runs the task the same way every time, right in the tools you already use. When something doesn't fit, it stops and asks you. Every step is logged for review.
Hi Product Hunt! I’m Stefan, co-founder and CEO of Latch.
AI has come with a lot of big promises, and yet many of us still spend hours pointing and clicking: pulling reports, moving data between systems, checking one number against another. That kind of work looks easy to automate until you dig in and find the nuance underneath: the exceptions, the edge cases, the judgment calls nobody wrote down.
I’ve built and sold two companies, and in both, some of our most capable people lost hours every week to exactly this kind of work. I started Latch to capture the nuance and free people to focus on the work that really matters.
You show Latch how the work gets done, and it helps you see which parts are truly repeatable and where the gray areas are. The repeatable parts become sidekicks that run the same steps every time and stop to ask you when they hit a gap they can’t fill or a decision they shouldn’t make alone. The goal is to hand the tedium to a machine, so people can spend their time on the judgment calls and new ideas that move things forward.
We started with accounting teams, because few processes have more nuance than the month-end close. The capture tool is where the magic happens. As you work, it records three streams: your narration, your screen, and your clicks and keystrokes. That’s how you get a working sidekick without ever writing a prompt or opening a coding agent.
Try our free plan today. No credit card required.
One question for you: if you had one recurring task you could hand over that would be done right, how impactful would it be for you?
Congratulations to the Latch team! I’ve done manual data entry for my accounting coursework, so I get why this matters. This seems like a great tool for improving efficiency, and I would love to try it!
Hi everyone, I'm Sam, head of product and customer success at Latch.
Handing a task to automation is a big ask, especially when it touches your books. So we built a safe place to try Latch first: a sample invoicing sidekick, ready in the app.
It works through a week of orders for a fictional coffee roaster and turns them into invoices. Along the way, it hits a couple of messy situations, the kind real orders always have, and stops to ask you how to handle them. That's the best way to get a feel for working with a sidekick: watching it follow the steps, then checking in with you when a decision is yours to make.
When you try it out, you'll see an Excel sheet with the demo data ready for download, so all you need is access to Excel. No accounting system to connect, no real data to share, and it runs in about a minute.
When you're ready, teach Latch one of your own tasks. If you get stuck, reply here and I'll help.
What would you want to see a sidekick do before trusting it with your real work?
Fantastic! Since the time of RPA, we've all been waiting for a simple system that could just watch, listen, and then intelligently reproduce workflows. And it's finally here!
Congratulations on the launch! I'm really excited about this problem! Reconciling books and closing has always been filled with Excel macros and a whole bunch of custom workarounds that don't handle nuance or exceptions very well, so this looks incredibly promising.
See it learn from someone doing a demo and emphasizing exceptions to take into account made the idea make sense. Seems like a super great tool.
Really like the point about all the little exceptions that make automation difficult. That's something many teams can relate to. Great work, Stefan, and congrats on the launch! 👏
I recently saw a prerelease end-to-end demonstration of the workstream capture and playback functionality of Latch. Super slick and really intuitive! This platform will be such a boost for folks who have complicated, repeating parts of their jobs.
Hi Product Hunt! I'm Emilie, the designer for Latch.
When we started building out Latch, we kept hearing the same worry: "Is AI going to replace me?" We designed around the opposite idea: that you're the expert. Latch learns from you, takes the tedious parts off your plate, and leaves you with the judgement calls.
That shaped a lot of decisions:
You decide what gets recorded. Latch only learns from sessions you initiate.
No surprises before the first run. Latch lists every step a sidekick will take and explains each permission it asks you. You approve only the apps it needs, and you can do a test run before it touches real data.
When something doesn't match what it learned, a sidekick stops and asks, always explaining the problem in plain language.
Everything is visible and every run leaves a step-by-step log you can track.
Under the hood, Latch is SOC 2 compliant, and data is encrypted in transit and at rest.
What would make you comfortable handing a task to a Latch sidekick?
Hey Product Hunt!
I'm Jared, co-founder and CTO. Building Latch was a serious technical challenge - it turns out if you take a 3 hour video and hand it to an LLM, it gets completely wrapped around the axle. If it even manages to take it all in, it can't tell the signal from the noise.
Solving that took some real smarts - but we landed on a multi-agent architecture that intelligently reads between the lines of what you're actually doing to build deep context around how you do the work you want Latch to do, and to make it reusable via MCP and our knowledge graph.
At the end of all of that, what you get is actually a deterministic sidekick to help get stuff done. Not another agent that's going to make your judgement calls for you.
Happy to go deep on architecture in the comments! What would you want to know before trusting automation with your books?
In my previous company we spent a ton of time building accounting system integrations, and even after that, many operations required multi-step manual workflows to move CSVs and reports around from one system to another. End of month close and remediating credit balances was a real pain. I'm excited to see some practical application of AI for day to day tasks in the accounting work, where accuracy and auditability are paramount but a lot of the work to achieve that is very repetitive!
Congrats on the launch! Stopping to ask instead of guessing is the right call for close work. Curious how a sidekick handles the underlying tool changing, e.g. Xero renaming a field or moving a step mid-quarter. Does the capture need to be redone, or does it notice the drift and flag it?
Glad it pauses when something doesn’t fit. For close work, a clear question is much more useful than a confident guess. Good luck, Stefan and team!
when latch hits an exception it has not seen before and stops to ask, does your answer get saved for the next run?
Congrats on the launch! We're an early-stage startup using Stripe + Xero, but haven't managed to get everything syncing properly. Ideally, we'd love to automate our bookkeeping so every transaction in Stripe Treasury gets recorded and categorized correctly in Xero, without having to manually review everything. Could Latch help us set up and automate a workflow like this?
Wait, this is actually interesting idea of just showing Latch what you're doing instead of explaining every little thing through prompts makes so much sense for accounting workflows
Love it congrats on the launch 👏
About Latch.ai on Product Hunt
“Automate complex accounting work without a single prompt”
Latch.ai launched on Product Hunt on November 10th, 2026 and earned 242 upvotes and 47 comments, placing #4 on the daily leaderboard. Latch is built for accountants buried in click-heavy close work. Instead of writing prompts, you do a task on screen and talk it through, including the parts no SOP captures: the three-day matching window, the rate you always override. Latch asks about the exceptions, then builds a sidekick that runs the task the same way every time, right in the tools you already use. When something doesn't fit, it stops and asks you. Every step is logged for review.
Latch.ai was featured in Productivity (665.5k followers), Artificial Intelligence (483.6k followers) and Accounting (1.9k followers) on Product Hunt. Together, these topics include over 319.1k products, making this a competitive space to launch in.
Who hunted Latch.ai?
Latch.ai was hunted by KP. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
Want to see how Latch.ai stacked up against nearby launches in real time? Check out the live launch dashboard for upvote speed charts, proximity comparisons, and more analytics.
KP
Hi Product Hunt! I’m Stefan, co-founder and CEO of Latch.
AI has come with a lot of big promises, and yet many of us still spend hours pointing and clicking: pulling reports, moving data between systems, checking one number against another. That kind of work looks easy to automate until you dig in and find the nuance underneath: the exceptions, the edge cases, the judgment calls nobody wrote down.
I’ve built and sold two companies, and in both, some of our most capable people lost hours every week to exactly this kind of work. I started Latch to capture the nuance and free people to focus on the work that really matters.
You show Latch how the work gets done, and it helps you see which parts are truly repeatable and where the gray areas are. The repeatable parts become sidekicks that run the same steps every time and stop to ask you when they hit a gap they can’t fill or a decision they shouldn’t make alone. The goal is to hand the tedium to a machine, so people can spend their time on the judgment calls and new ideas that move things forward.
We started with accounting teams, because few processes have more nuance than the month-end close. The capture tool is where the magic happens. As you work, it records three streams: your narration, your screen, and your clicks and keystrokes. That’s how you get a working sidekick without ever writing a prompt or opening a coding agent.
Try our free plan today. No credit card required.
One question for you: if you had one recurring task you could hand over that would be done right, how impactful would it be for you?