IrisGo Review: The AI Assistant That Learns by Watching You Work

The short version: This IrisGo review looks at a desktop AI assistant that turns a task you demonstrate once into a reusable workflow. It launched for solopreneurs on 7 October 2026, took #1 Product of the Day on Product Hunt, and is free while in beta on Windows 11 and Apple Silicon Macs. The real question is not whether the AI works but whether your repetitive work is routine-shaped enough to replay. It is a promising free experiment, not a tool to reorganize your business around yet.

Most automation tools ask you to build the workflow before you can save any time. You map the triggers, connect the apps, define every edge case. For the long tail of small, annoying routines that eat a freelancer’s week, that setup cost is usually bigger than the payoff, so the routine stays manual forever.

IrisGo flips the order. You do the task once while it watches, and it turns that demonstration into a workflow it can replay. The company calls it Watch & Learn, and it is the whole pitch. Everything else — the daily brief, the email triage, the to-do radar — is scaffolding around that one idea.

IrisGo review: what it actually does

IrisGo is a desktop app, not a website. You install it on Windows 11 (64-bit) or an Apple Silicon Mac, and it runs alongside your work. Per the maker’s launch notes on Product Hunt, the solopreneur package ships five things:

• Watch & Learn: demonstrate a task once; IrisGo records the pattern and offers it back as a reusable workflow.
• Daily Brief: a morning summary of priorities drawn from email, calendar, and to-dos.
• Email Triage: sorting the inbox into what needs an urgent reply and what can wait.
• My Radar: one to-do list that keeps updating itself as you finish things.
• Meeting Summary: turns calls into notes with action items added for you.

The installation screen on irisgo.ai/solopreneur states three things plainly: it is free during the beta, it sees your data and workflows only with your consent, and updates install automatically. That is refreshingly direct for a beta sign-up page.

The judgment that decides everything: is your work routine-shaped?

Here is the honest way to evaluate IrisGo, and it has nothing to do with how clever the AI is. Watch & Learn treats your demonstration as a pattern, not a recording. When a Product Hunt commenter asked how replay handles a different number of files each time — downloading and sorting a stack of resumes, in their example — the maker answered that it processes each item one by one and uses visual checks to know when the list is empty, and that you can pass natural-language tweaks at replay time, like a custom naming format. In other words, the input varies but the skeleton is fixed.

That is the entire buying logic in one sentence: the more of your repetitive work follows a fixed skeleton with varying inputs, the more IrisGo is worth to you. Renaming and filing client files, logging invoices, building the same weekly status email from Friday’s notes, prepping a standard client onboarding folder — these are routine-shaped. Negotiating a scope change, designing a proposal, deciding which lead deserves a follow-up — these are not, and no demonstration-based tool will touch them.

This is also where IrisGo parts company with the automation tools most freelancers already know. Zapier and Make are connector-based: you define triggers, actions, and conditions, and they execute faithfully. They shine when the pipeline is structured — a form submission that creates a CRM record, a payment that triggers an invoice. IrisGo is demonstration-based: it shines where there is no clean connector, no API, just a messy sequence of clicks across apps you will never integrate. The two overlap less than they appear to.

Where the numbers stand

The company behind IrisGo is a startup co-founded by Jeffrey Lai, a former Apple engineer who worked on the Chinese-language version of Siri (the name Iris is Siri spelled backwards). In May 2026 it raised $2.8 million in seed funding led by Andrew Ng’s AI Fund, with Nvidia and Google also backing the company, TechCrunch reported (a second write-up at imp.news confirms the $2.8 million figure and the Nvidia/Google backing). It has also signed a deal with Acer to preinstall the software on AI PCs, which is the clearest signal of where the company thinks this category goes.

On 7 October 2026, the solopreneur launch took #1 Product of the Day on Product Hunt with roughly 443 votes and 38 comments (per a launch tracker; Product Hunt numbers move, so treat these as approximate). The comments thread is unusually substantive for a launch day — makers answering technical questions about dynamic replays, Linux plans (on the roadmap, after macOS and Windows stabilize), and whether Linear, Jira, ClickUp, or Asana tasks will join the Radar. For an IrisGo review written a week after launch, that thread is the most valuable source available.

One comment worth flagging: a user asked directly how much it costs and said they could not find a pricing page. There is no pricing page. Free during the beta is the only price statement the company has made, on its own site. That is fine for trying it, but it means you cannot do a real cost comparison against paid tools yet — anything about future pricing would be speculation.

IrisGo vs Zapier vs Make: which solves your problem?

The table above is this IrisGo review’s synthesis from the company’s own documentation and launch materials, not from hands-on testing — nobody outside the company has had this tool long enough for an independent review to exist, and I am not going to pretend otherwise.

Notice what the table implies for solopreneurs choosing their automation stack: these tools are complements, not substitutes. A freelancer could reasonably run Make for the structured stuff — new lead in, invoice out — and IrisGo for the demonstrated stuff — the weekly folder cleanup, the recurring report assembled from three apps. The decision is per-workflow, not per-person.

How much admin time could a tool like this reclaim? Run your own numbers

The marketing claim behind every automation tool is “hours saved.” The honest version depends on your week, your rate, and how much of your admin work is actually routine-shaped. This calculator estimates the reclaimable share — the number is our calculation from your inputs, not a vendor promise.

The formula, plainly: reclaimable hours per week equals your admin hours multiplied by the share that follows a repeatable pattern. Multiply by weeks worked for the yearly figure, then by your hourly value for the dollar estimate. This is our calculation — it assumes the tool handles the pattern reliably, which is exactly the part a beta has not yet proven at scale. Use it as a ceiling, not a forecast.

The privacy trade, honestly

There is no way to soften this: an assistant that learns by watching you work must watch you work. The installation page says it sees your data and workflows at your consent, and the company says most processing happens on-device, with cloud processing only when you explicitly authorize it and with end-to-end encryption. That architecture is genuinely better for privacy than a cloud-first competitor — your screen recordings are not the product.

But on-device does not mean no-risk. You are giving a beta application from a $2.8-million-seed startup broad visibility into your desktop: your inbox, your files, your meeting transcripts. If you handle client data under confidentiality agreements, or work in anything touching regulated information, read the terms before installing and probably wait for a stable release with an audited privacy policy. A sensible AI usage policy for a small business would treat any screen-watching beta as experimental software until proven otherwise. That is the privacy verdict of this IrisGo review in one line. The risks of unvetted AI tools inside a business apply doubly to one that can see everything on your screen.

Who should install the beta, and who should wait

Install it this week if: you run Windows 11 or an Apple Silicon Mac, you do several hours a week of repetitive desktop routines with fixed skeletons, and you can tolerate beta roughness — missed steps, wrong guesses, workflows that need re-demonstrating.

Wait if: you are on Linux or Intel Macs, you handle client data under strict confidentiality, your machine is old or slow (on-device processing is the cost of that privacy), or you need pricing certainty before adopting a tool — because there is none yet. If the meeting summaries are what attract you, note the consent laws around AI meeting recording apply to any tool that captures calls, this one included.

One more honest note from this IrisGo review about where this fits in a freelancer’s toolkit: IrisGo does not replace the judgment layer of your work. AI agents are good at execution and bad at decisions, and a demonstration-based assistant is the purest form of that split — it copies how you did it, including the parts of your routine that were never efficient in the first place. Automate a bad routine and you get a faster bad routine.

What we cannot confirm yet

Because this is a week-old public beta, the list of unknowns is long, and it would be dishonest to shrink it:

• Post-beta pricing. Free during the beta is confirmed; everything else is unconfirmed. A Product Hunt commenter already asked and found no pricing page.
• Reliability at scale. The maker’s answers on dynamic replays are promising, but they are the maker’s answers. No independent, hands-on testing exists yet.
• The Acer preinstall timeline. The deal is signed, per TechCrunch, but there is no public date for when IrisGo ships preinstalled.
• How “on-device” holds up. The hybrid architecture means some tasks go to the cloud; the boundary between the two is defined by the company’s own terms, which can change.
• Whether the category survives. A $2.8M seed is small for building an “AI operating system.” The Acer deal and Andrew Ng’s backing help, but this is an early-stage bet, not an established product.

And the practical beta advice: have a plan for when an AI tool you depend on breaks — doubly true for one in public beta. The time-saving claims around AI tools deserve the same skepticism here as everywhere else: measure the reclaimed hours yourself after two weeks, or you are just believing a launch page.

Is IrisGo really free?

Yes, during the public beta — that is the company’s only pricing statement, on its own site. No paid tiers have been announced, and there is no pricing page. Expect paid plans eventually; nothing about them is confirmed.

Does IrisGo send my screen data to the cloud?

Mostly no, by design. The company says much of the processing happens on-device, with cloud processing only when you explicitly authorize it, using end-to-end encryption. It is a hybrid architecture, not a fully local one — read the current terms for the exact boundary.

What computers does it run on?

Windows 11 (64-bit) and Apple Silicon Macs. Linux is on the longer-term roadmap after macOS and Windows stabilize, per the maker’s answers on Product Hunt.

How is it different from Zapier or Make?

Zapier and Make are connector-based: you build rules and they execute them. IrisGo is demonstration-based: you show it a routine once and it replays the pattern. It is strongest on messy, long-tail desktop tasks with no API connector — exactly where the connector tools struggle.

What happens when the beta ends?

Unconfirmed. The free-during-beta offer will presumably convert to paid tiers, but the company has announced nothing. The Acer preinstall deal suggests the long-term business model may involve bundled distribution, not just subscriptions.

Who should not install it?

Linux users, anyone handling client data under strict confidentiality, anyone on weak hardware that cannot comfortably run on-device AI, and anyone who needs pricing certainty before committing to a tool. A beta that watches your screen is experimental software — treat it that way.

Shurah is the founder of AI Tools Daily, tracking pricing, licensing and policy changes across AI tools so readers can make decisions without wading through marketing claims themselves.