I replaced our support queue with a script tag
Most in-app guidance tools were built before AI existed. You spend weeks writing flows, maintaining them every time your UI changes, and praying users actually follow the steps.
I wanted something that required zero maintenance and actually understood what the user was trying to do.
Ozkour is an AI assistant that sees your user's screen in real time. It highlights clickable elements, guides them step by step, and adapts to whatever state the interface is in, no predefined flows, no manual updates.
You add one script tag. It works in 3 minutes.
Also : it's GDPR compliant, cookieless, and never stores personal data or IPs.
If you run a web app and your support team is answering the same "how do I..." questions every week, this is for you.
Happy to answer questions and brutal feedback is welcome 😌
the one script tag install is the right move for adoption. biggest barrier for any embedded tool is the integration effort - if its more than 5 minutes, most solo devs wont bother during evaluation.
real question though: whats the latency like for the screen reading + AI response cycle? if a user clicks help and waits 3+ seconds for the AI to understand their screen state, it feels broken. sub-second context awareness is what would make this feel magical vs just another chatbot.
also thinking about your go-to-market: the best distribution channel for this might be dev tool directories and saas boilerplate communities. the people building new saas products with nextjs templates are exactly who would drop in a support widget on day one. catch them at project setup and youre embedded before they ever consider alternatives.
thx for you reply, love the idea of going into more tech adjacent communities for distribution! For the latency issue, we're in the lower 2s bracket without sacrificing quality, we could go lower but at some point it's a balancing act. Funny to think an actual support ticket can takes up to 24h to get resolved but 3s is too much in our case :D
The "sees your user's screen in real time" part is what makes this different from the typical Intercom/Chameleon approach. Those tools are all about pre-built flows that break the moment you ship a UI change. Having the AI understand the current state of the interface is a much better architecture for fast-moving teams.
One thing I'd be curious about — how does it handle complex multi-step workflows where the user needs to jump between different pages? Like if someone's trying to set up an integration that involves going to Settings > API Keys > Create Token > paste into another app. Does Ozkour maintain context across page navigations, or does each page load reset the conversation?
Also, the GDPR/cookieless angle is smart positioning. A lot of EU-based SaaS companies won't even look at tools that store user data on third-party servers. If you can genuinely deliver the guidance without any PII touching your infrastructure, that's a real differentiator worth highlighting more prominently on the landing page.
hey! yes ozkour handles page jumping like a champ, only case we're not solving yet is multi app flows. The GDPR thing is real, we're based in France and to have any success in this market we had to build it as transparent and light as possible without sacrificing any useful context.
Really interesting approach — zero maintenance is the key insight here.
We hit the same pattern from a different angle: instead of guiding users through the app, we eliminated the support questions before they happened. Connected fulfillment webhooks to proactive emails so customers always knew their order status without asking.
Result: 40 support tickets/month → 8. The "how do I track my order?" questions went to zero.
The common thread seems to be: most support load comes from users lacking information at the right moment, not from the product being hard to use. Give them the info proactively and the queue disappears.
Curious how Ozkour handles the moment when a user is stuck but doesn't know what question to ask — does the AI detect frustration signals or only responds to explicit actions?
hey, good luck with your product man! we're not tackling the proactive yet but it's a really great suggestion. We're toying with prototypes where we detect erratic mouse patterns that could show frustration, or even the mac's built-in accelerometers to see if a user has hit their desk ahah
Smart move going niche with TripSuite over fighting TripIt. A clean time-based timeline for meetings, deadlines, and commitments hits the real pain for business travelers who want productivity, not just itineraries. Offline-first plus optional AI refinements keep it practical. Freemium with Pro after the first trip is a solid conversion hook once value shows. Laser focus on this audience should win big. Great execution.
Congrats on the launch, looks solid. How are you currently thinking about acquiring early users and gathering feedback?