A note from Karthik Kamalakannan, Founder and CEO of Skcript.

I want to tell you why we are starting Skcript Applied Labs, and why I am putting my own name and time behind it.

Over the last year, almost every company I talk to has asked me the same thing in a different way. They have tried AI. They bought the licences. They ran a few experiments. Somewhere there is a pilot that looked great in a meeting. Then it stalled, and the team quietly went back to doing the work by hand. Each time I hear this, I feel the same pull. These are good teams with real problems, and the technology to help them is already here. It just never reached their desk.

I started Skcript in 2013 because I believe good technology should not belong only to the biggest companies. Our mission is to bring simplicity to the world's work through design and technology. For me that has always meant one thing: getting useful technology into as many companies as we can, and making it simple enough that people actually use it. AI is the biggest chance I have seen in my career to do that. I don't want it to stay stuck in demos. So we are building a team to get it into real work.

What SAL is

Skcript Applied Labs, or SAL for short, is a small team of senior engineers at Skcript. You give us one task that eats your team's week. In 10 working days, you see AI doing that task on your own data, with before and after numbers, and a fixed price to take it live.

We are not selling a product. We build inside your tools, and you own everything we build: the code, the tests and the numbers.

We have been building software since 2013. We run AI features in production in our own products, FeatureOS and SupportWire. For a global bank, we automated payroll for more than 10,000 people across more than 25 systems, and it went live in 3 months. SAL takes that habit of shipping and points it at one problem: getting AI to work in your business.

Why pilots stall

In my experience, a pilot rarely stalls because the model is weak. It stalls in the gap between a demo and a system a team can rely on every day. Nobody measured the starting numbers. Nobody tested it on real cases. Nobody connected it to the tools people actually use.

That gap is the work I want us to do. It is not glamorous. It is careful engineering, honest measurement and patience. It is also where all the value is.

The market has changed

This year the big AI labs started running their own deployment teams. Anthropic, with Blackstone and Hellman & Friedman, launched Ode in July. OpenAI has its own deployment arm too. They put engineers inside large companies to make their models work in real processes.

I think this is good news. It tells every board that the hard part of AI is not the model. It is the engineering.

I recently watched an interview with Eddie Siegel, Ode's CTO. A lot of what he said matched what we have learned the hard way. Start from a business case you can measure in money. Write tests before you trust anything. Don't build a demo you will throw away. Don't wait for a data clean-up before you start. I agree with all of it.

Where we differ is the starting point. A company owned by a model maker will naturally start from its own model. That is a fair choice for them. We start from your numbers instead. We don't sell a model, and we are not owned by one.

How we work

We start with money, not AI. On day one we measure how long the task takes today, what it costs and how often it goes wrong. Then we write down the one money number the pilot should move. On day ten we report against it. If the number doesn't move, we tell you plainly.

We test every model on your own cases. We build tests from 100 to 200 of your real cases. Then we run Claude, OpenAI and open models through the same tests and show you the scores side by side, with the cost per task. You pick with the facts in front of you.

The pilot is not a demo. What we build in 10 days is the first working piece of the real system. It runs on your data, inside your tools. Nothing is thrown away.

Your data doesn't need to be perfect. The most common reason I hear for waiting is "our data isn't ready." For this kind of work, a set of real examples is enough to start.

We work inside your rules. We use your systems and follow your security rules. Data stays in the UAE or Saudi Arabia where required.

We keep it honest after launch. Models change fast. When a better one ships, we run it through your tests. We switch only if it wins.

Why a small team, and why me

Big programmes create a lot of meetings. Small teams with clear numbers create working software. SAL is small on purpose. The engineers on your first call are the people who build it.

I will be on that first call too. I want to hear about your task from you, understand what it costs your team, and be straight with you about whether AI is the right answer. I stay involved after that. You will always know who is accountable, and it is me.

It also means we can work with companies that big programmes are not built for. A mid-sized firm in Dubai or Riyadh with one painful workflow should not need a year-long contract to find out whether AI helps.

What it costs

The 10-day pilot is a fixed USD 7,500, about AED 27,500. You get:

  • A business case with a money number
  • Tests built from your own real cases
  • A model scorecard comparing Claude, OpenAI and open models, with cost per task
  • A working first piece that carries on to production
  • A fixed price for the full rollout

If you start the rollout within 30 days, the pilot fee counts toward it. If the pilot shows the task isn't worth doing, you still keep the numbers and the tests.

If you already tried AI and it never went live, start with a 5-day pilot check-up. We test what you have on your real work and tell you why it stalled.

Where we are starting

We are starting with companies in the UAE and across the Gulf. That is where I see the most appetite and the most stalled pilots. We work Gulf hours, and we meet in person for the sessions that matter.

The work we like best is high volume and heavy on text or documents. Invoices and purchase orders. Supplier and customer emails. Contracts, claims and onboarding files. Internal requests.

Let's talk

You don't need an AI strategy deck to start. You need one task that costs your team real time, and ten days.

If you have that task in mind, I would genuinely love to hear about it. Start a chat with us right here on the site, or book a call through the Applied Labs page. I read every message myself, and I will be on the first call. Let's get your team some of its week back.

Karthik