We built an AI product. That changes what we can tell you about AI.
everTest turns plain-English test cases into production Playwright and Selenium code, in bulk, with human QA in the loop. SWStrategies designed and built it, and our founder Shachindra Agarwal is its co-founder and president. It has paying customers today.
Hundreds of testsin weeks, not quarters
Test cases written in plain English become real automation code at bulk scale.
~75%lower cost to author a test
Measured against writing the same test by hand. Automation stopped being worth it for most teams because authoring and then maintaining it cost more than the bugs it caught.
You own the codeno platform, no lock-in
Standard Playwright or Selenium in your repository. If everTest disappeared tomorrow, your suite would still run.
Ten years in the making, and we mean that literally
everTest is not something we started when generative AI arrived. It is the third attempt at a problem we have been circling since 2016.
About ten years ago we built ZenVerify — a record-and-playback test automation product. The economics were deliberate: a junior QA engineer records the ordinary path by simply using the application, and a senior engineer writes a few lines of JavaScript only where a case is genuinely hard. The expensive person touches a fraction of the work. We released it as a product and licensed it to a Big 4 accounting firm.
They used it for KYC, and that became the 800-bot project on our case studies page — data pulled continuously from more than 800 registry and regulator sites, after two previous attempts had been abandoned because maintaining hand-written automation cost more than the data was worth. It went to production in six months and cut the cost of building and maintaining those bots by 90%.
But it never made test automation mainstream, and that is the thing we actually wanted. Recording still breaks when the application changes. Someone still has to write the hard cases. For most teams the maintenance curve still bent the wrong way, and the suite still rotted. We have been thinking about that problem for a decade.
Generative AI opened the door. Not because it writes code quickly — that was never the bottleneck — but because it can read a test case written in plain English, work out what it means against a real application, and regenerate the automation when the application moves. That is the part recording could never do. everTest is what we built once that became possible.
The line, in one place
ZenVerify
Record and playback, built by SWStrategies and released as a product. Cost came from splitting the work by skill. Licensed to a Big 4 accounting firm.
800 bots in production
A KYC project two previous attempts had failed to make affordable. Six months to production, 90% cheaper to build and maintain.
everTest
Plain English in, production Playwright and Selenium out, with human QA in the loop. The same goal, finally reachable.
That is why we are not nervous about AI and not credulous about it either. We have been trying to solve this specific problem since long before it was fashionable, and we know exactly which part of it the models actually solved.
Who runs it
A separate company with its own founders, not a side project with a product name on it.
Shachindra Agarwal
Co-founder & president, everTest · Founder & CTO, SWStrategies
The same person who would lead your SWStrategies engagement. That overlap is the point: what we learn running an AI product in production goes straight into client work, and what we learn on hard client systems goes straight into the product.
Pulkit Agarwal
Co-founder, everTest · strategy and operations
Runs strategy and operations at everTest, after fifteen years in global health. He is not part of SWStrategies engagements — he is named here because everTest is a real company with real founders, which is the claim this page is making.
Why a services client should care
Every firm you are talking to says they do AI. Here is the difference you can actually check.
We run AI in production, not in a deck
An AI system that has to behave correctly against real applications, every day, with paying customers watching, teaches you things a pilot never will — about cost per call, failure modes, evaluation, and which parts of the industry pitch are not true yet.
We know what AI engineering really costs
We track model spend, latency and accuracy on our own product as an operating expense. When we estimate an AI feature for you, the numbers come from a running system rather than from a vendor’s pricing page.
Your team gets the smaller, faster version
Building with AI tooling is why a team of four senior people now delivers what used to take fifteen. That is the main reason our costs have come down, and we pass it on rather than billing for seats.
If testing is your problem
You can engage everTest directly as a product, or have us fold it into a wider engagement.
Go straight to everTest
You have an application, a backlog of manual test cases and a QA team that cannot keep up. everTest will automate a first batch free so you can judge the output before you commit to anything.
Bring us the whole problem
Testing is rarely the only thing wrong. If the release process, the architecture or the delivery model is also in the way, that is a SWStrategies engagement — and everTest becomes one part of it.
Want to see it run against your application?
We will take a handful of your real test cases, automate them, and show you the code. No charge, and you keep whatever we produce.