Getting here was not straightforward. This post covers what the journey actually looked like, because most product announcements skip the part where things did not work.
EscalationAI is a Zendesk app that scores every support ticket for escalation risk in real time. It reads the ticket content and conversation history, evaluates it across more than 25 signals, and returns a risk score directly inside the Zendesk ticket view.
The signals include things like legal language in the customer’s message, repeated contacts on the same issue, sentiment deterioration across the conversation, manager requests, churn indicators, and response time gaps. The score is accompanied by the specific signals that drove it, so agents know exactly why a ticket was flagged, not just that it was.
The goal is simple. Catch the tickets that are about to escalate before they do, while there is still time to intervene.
Zendesk takes its marketplace seriously. Every app goes through a manual review that checks code quality, security implementation, API handling, data privacy compliance, and adherence to Zendesk’s developer standards.
For EscalationAI, it took multiple submissions to get through.
The first submission was rejected with three specific blockers. The domain whitelist implementation was missing. The short description exceeded the character limit. And the secure settings implementation did not follow Zendesk’s required pattern.
Each blocker was legitimate. None of them were vague. Zendesk’s reviewer was specific about what needed to change and why.
The fixes required a complete rebuild of the authentication layer. The correct pattern for Zendesk apps requires the frontend to retrieve the installation context via the ZAF SDK, send it to the backend, and resolve the API key from there. A shared authentication library was built, a new endpoint was added to handle installation resolution, and a setup page was created to manage the configuration flow.
That submission passed review. EscalationAI is now approved by Zendesk and the public listing is being processed.
EscalationAI was built to align with ISO/IEC 42001:2023, the international standard for AI management systems. This was a deliberate design decision, not an afterthought.
In practice it means:
Every AI decision is logged. Every risk score has an audit trail. The signals that drove the score are always surfaced to the agent, not hidden inside a black box. Human agents remain in control at every point. The app assists judgment, it does not replace it.
For support teams operating under compliance requirements or with leadership that wants accountability over AI-assisted decisions, this matters. Most AI tools in the Zendesk ecosystem do not address governance at all.
The public listing on the Zendesk Marketplace is being processed and will go live shortly. This is a manual process handled by the Zendesk Partner team.
In the meantime, early access is open. If you manage a Zendesk support team and want EscalationAI installed before the public listing opens, reach out directly through the product page. Setup takes under 30 minutes and does not require changes to your existing Zendesk configuration.
Early access users get direct input into what gets built next and access before the general Zendesk user base.
Once the public listing is live on the Zendesk Marketplace, EscalationAI will be discoverable by any Zendesk customer searching for escalation management tools. The app will continue to be actively developed with improvements informed by real usage feedback.
If you are running a support team on Zendesk and escalations are a recurring problem, now is the right time to get in. Early access is open and the public listing is coming shortly.
]]>You run a search. Forty results come up. You open the first profile, copy the name, switch to your spreadsheet, paste it, go back, copy the title, paste that, go back again for the company. Forty times. Sometimes more. By the end of it you have a messy spreadsheet, a few missing fields, and about ninety minutes of your day gone.
LinkedIn does not give you an export button. That is a deliberate product decision on their part. But it creates a real problem for anyone who prospects at scale.
It is a Chrome extension. You run your LinkedIn search the same way you always do. When you are ready, you click Extract. The extension pulls every result on the page and downloads it as a clean CSV. Names, titles, companies, profile URLs. Under ten seconds.
That is the whole product. One problem, one solution.
It works on standard LinkedIn search and on Sales Navigator. No scraping in the background, no data leaving your machine to a third-party server. Everything happens locally in your browser.
Twenty-two years in service delivery taught me one thing about tools: the ones that actually get used are the ones that fit exactly where the pain is. Not adjacent to it. Not a workaround for it. Right where it hurts.
The pain here is specific. SDRs, recruiters, and founders who prospect on LinkedIn spend a disproportionate amount of time on data entry that should not exist. The research is valuable. The copy-pasting is not.
I built SalesScraper Pro because I wanted a tool I would have used on day one of any sales or recruiting role. No subscription to forget about. No complex setup. Just install, click, get your data.
SalesScraper Pro goes live on Product Hunt on Wednesday June 10 at 1:31 PM IST.
If you know anyone in sales, recruiting, or business development who prospects on LinkedIn, I would genuinely appreciate a share. An upvote on launch day takes thirty seconds and makes a real difference for a solo-built product.
You can try the extension today at sspro.vijit.in. There is a sixty-second demo video on the page that shows exactly how it works.
The pricing is straightforward. $49 one-time or $2,999 for lifetime access. No subscription. Seven-day refund policy if it does not work for your use case.
If you do try it and run into any issues, email me directly at support@vijit.in. I read every message.
It took about six hours to fix everything properly. Not because the changes were hard, but because one of the three blockers was pointing at something more structural — the way secure settings were being handled in the frontend.
Here is what happened, what was wrong, and what I changed.
1. Missing domainWhitelist in manifest.json
When a Zendesk app uses secure parameters, you must declare which external domains the app calls. I had api_key marked as secure: true but no domainWhitelist defined. The fix is one line:
"domainWhitelist": ["escalationai.vijit.in"]
Simple. But it forces you to think about something important: every domain your app touches needs to be declared. If you are calling multiple services, all of them go here. It is a useful audit of your own app’s surface area.
2. short_description over 80 characters
The submitted file had 93 characters. Limit is 80. The version I thought I had submitted had 55. Turns out the file on disk was different from what I had been working on locally. The fix was trimming and resubmitting the correct file.
Lesson: always verify the exact file you are packaging, not the one open in your editor.
3. Secure settings not coded correctly in index.html
This was the real one.
The original code was fetching the API key on the frontend using client.metadata():
const meta = await client.metadata();
const apiKey = meta.settings.api_key;
Zendesk’s reviewer flagged this as incorrect. And they were right.
Here is why: when a parameter has secure: true in manifest.json, Zendesk intentionally prevents the frontend from reading it. The whole point of marking something secure is that it never touches the browser. If your frontend is trying to read a secure setting, you have designed the auth flow backwards.
The correct pattern is: the frontend does not handle credentials at all. Instead, it sends an installationId to your backend. Your backend resolves the actual API key from there.
In practice, this meant:
// Frontend: send installationId, not the key itself
const context = await client.context();
const installationId = context.installationId;
fetch('https://escalation-ai-backend.vercel.app/api/score', {
headers: {
'X-Installation-Id': installationId
}
});
The backend then looks up which API key maps to that installationId in Redis. The key never leaves the server.
This required a new Redis mapping (installation:{id} to api_key) and a new endpoint (/api/resolve-installation) that customers call once during setup to create that mapping.
It also required a setup page at escalationai.vijit.in/setup so customers can link their Zendesk installation to their API key without it ever passing through the browser during normal use.
The Zendesk Marketplace review is not just a checkbox exercise. The three blockers they caught were all legitimate architectural concerns:
Getting flagged on the third one was useful. The app is more secure now than it was before submission.
EscalationAI v1.1 is submitted and under review. Changes in this version:
domainWhitelist added to manifestshort_description corrected to 78 charactersclient.context() installationId patternlib-auth.js) that resolves installationId to api_key via Redis/api/resolve-installation endpoint for one-time setup linkingThe ISO 42001 compliance features — audit trail, explainable AI, human override logging, bias monitoring — are all unchanged. The auth fix did not touch any of that.
If you are building a Zendesk Marketplace app, the documentation on secure settings is worth reading before you write a single line of frontend code. The constraint shapes the architecture in a good direction.
EscalationAI is a Zendesk sidebar app that scores every ticket for escalation risk in real time. Free 7-day trial at escalationai.vijit.in.
]]>That is not a process failure. That is a tooling gap.
EscalationAI was built to close that gap. It runs inside Zendesk, scores every ticket for escalation risk in real time, and gives agents the context they need to act before a situation spirals. It is now submitted to the Zendesk Marketplace for review and available today as a private app for teams who want early access.
The app reads the content of a Zendesk ticket and returns a risk score along with the signals that drove it. Those signals include things like legal language, manager requests, repeated contacts on the same issue, sentiment deterioration across the conversation, and churn indicators.
The score appears directly inside the Zendesk ticket view. No switching tabs, no exporting to a spreadsheet, no separate dashboard to check. Agents see the risk score while they are working the ticket, which is the only place it is useful.
Every score comes with an explanation. Not just a number, but the specific reasons behind it. That matters for two reasons. First, agents can act on it immediately without guessing. Second, team leads have an audit trail for every AI-assisted decision, which is increasingly important as companies face scrutiny over how AI is used in customer-facing operations.
ISO/IEC 42001:2023 is the international standard for AI management systems. It covers how AI tools are governed, monitored, and held accountable within an organization.
Most AI tools in the customer support space do not address this at all. They generate outputs, and you trust them or you do not. There is no transparency, no audit trail, no framework for oversight.
EscalationAI was built from the start to align with ISO 42001 principles. Every AI decision is logged. Every score is explainable. Human agents remain in control. The app assists judgment, it does not replace it.
This is not a marketing claim. It is a design decision that runs through every part of how the app works. If your organization is subject to compliance requirements, data governance policies, or simply has a leadership team that wants to understand how AI is being used in operations, EscalationAI is built with that accountability in mind.
Building a Zendesk app is straightforward. Getting it listed on the Zendesk Marketplace is a different matter. The review process is thorough. Zendesk checks the app against their security requirements, reviews the code, tests the functionality, and verifies that the app meets their standards for performance and reliability.
EscalationAI has been submitted and is currently under review. That process takes time and that is fine. The review exists for good reason. Zendesk customers trust that apps in the marketplace have been vetted, and EscalationAI will meet that bar.
While the marketplace listing is pending, the app is available as a private Zendesk app. Private apps work identically to marketplace apps. The only difference is the installation method. Your Zendesk admin installs it directly using the app package, rather than through the marketplace interface. Full instructions are on the product page.
EscalationAI is built for support teams that handle volume. If your agents are managing 50 or more tickets a day each, the probability of a high-risk ticket slipping through without being flagged is not theoretical. It happens regularly. The cost is a churned customer, a legal complaint, or a manager escalation that consumes hours of time that could have been prevented with a 30-second intervention.
It is particularly useful for:
If your team is smaller and every agent already knows every customer personally, you probably do not need it yet. Come back when volume becomes a problem.
Once the marketplace listing is approved, EscalationAI will be available for any Zendesk customer to install directly. Pricing, tier details, and a free trial option will be available at that point.
Until then, early access is open. If you manage a Zendesk support team and want to evaluate EscalationAI before the public listing, reach out through the product page. Setup takes under 30 minutes and does not require any changes to your existing Zendesk configuration.
EscalationAI is built by Vijit Labs. If you have questions about how it works, the ISO 42001 alignment, or the private app installation process, the contact details are on the product page.
]]>Open LinkedIn. Find a prospect. Copy their name. Switch to your CRM. Paste it. Go back to LinkedIn. Copy their job title. Switch tabs again. Paste. Repeat this 50 times a day, and you’ve burned anywhere from 45 minutes to an hour. On data entry. Not selling.
That is not a productivity problem. That is a tooling problem.
SalesScraper Pro is a Chrome extension that pulls prospect data directly from LinkedIn profiles and exports it in a clean, usable format, without you touching a spreadsheet manually.
You visit a LinkedIn profile. The extension reads the page. You get the data you actually need: name, job title, company, location, and more, ready to copy or export as a CSV.
No scraping scripts. No Python setup. No API keys. Install it from the Chrome Web Store, and it works.
Sales reps and SDRs who build their own lists and are tired of copying data between tabs. If your quota depends on volume, every minute of manual data entry is a minute not spent on conversations that close deals.
Founders doing their own prospecting. Especially in the early stage when there is no dedicated sales hire. You know who you want to reach. You just need a faster path from “found them on LinkedIn” to “in my outreach tool.”
Recruiters and growth operators who research people on LinkedIn regularly and need to move data somewhere without building a workflow from scratch.
If you are already on LinkedIn Sales Navigator, SalesScraper Pro works alongside it. Free or Premium account works too.
There are other scraping tools. A few are expensive. Several break after LinkedIn updates its page structure. Most were built for technical users.
SalesScraper Pro is built for people who just want the data, fast.
Three things that matter:
It runs in your browser. No server involved, no third-party storing your search data. The extension processes what you are viewing on your screen.
The output is clean. You do not get a raw dump of every field on the page. You get the fields that actually go into a CRM or outreach tool, formatted consistently.
No learning curve. Install it. Open a LinkedIn profile. Done. No dashboard to configure, no onboarding flow to sit through.
Manual prospecting at scale is where sales productivity quietly dies. Most reps do not notice how much time they are spending because it is spread across 50 small actions throughout the day.
The math is straightforward. If you spend 45 minutes on data entry daily across a 250-day sales year, that is over 187 hours. More than four full 40-hour weeks spent on work that does not require your expertise and does not move a single deal forward.
SalesScraper Pro does not close deals for you. But it gives back the time you are currently losing to work that a tool should handle.
SalesScraper Pro is available at sspro.vijit.in and in the Chrome Web Store.
Install it, run it on a profile you already had open, and you will see immediately whether it fits how you work.
If you manage a sales team and want to evaluate it across multiple reps, reach out via the contact page. Team licensing is available.
SalesScraper Pro is the first product out of Vijit Labs. The goal with every product here is simple: one problem, solved well, with no setup friction. If you try it and have feedback, use the feedback option inside the extension. This product is actively developed and your input shapes what gets built next.
]]>