A great place to start the Sticklight story is by sharing how we even got here in the first place

🐣 2023: A Startup journey begins

Our journey started in early 2023. Aviv and Matan, both engineers at a predictive analytics startup, discussed a shared vision of AI's future, a few months after ChatGPT's groundbreaking release. Everything was new then, but clearly we had crossed a point of no return.

That year, we experimented with early public LLMs, trained and operated various models, implemented agents, and built a small open-source project for structured model outputs when this was still unsolved.

We understood how LLMs work and their applications. We also wanted to start our founder journey. In late 2023, we both left and began our entrepreneurship adventure.

Soon after, Tom and later Gilad joined as our founding engineering team - good friends and colleagues from our previous company.

🗺️ Before July 2024: Ideation & Pivots

Like any startup journey, we pivoted several times before landing on Sticklight.

Our previous attempt, Vendi.ai, eventually led to Sticklight's creation. Vendi was a PaaS for high-throughput LLM workloads. We operated under the then-controversial idea that various models would exist for different tasks at different price points.

The challenge was: "How do I use the right models, maximize value, and minimize cost?" After 6 months, we accepted this offering wouldn't succeed.

While the concept made sense to people, users struggled to switch models without tools to evaluate quality impacts in production. With models performance improving rapidly and falling prices, this problem wasn't pressing enough.

💡  July 2024: The Birth Of Sticklight

We did learn something very valuable, though, from working with these early adopters of AI. They had no F**ing clue how AI was actually affecting their product.

Going deeper, we discovered something interesting. Existing "Analytics for AI" products were built for developers - using "traces" and "spans" to track LLM performance and cost in production.

We assumed product analytics platforms would address this better, but instead found tools requiring clear events like "user did action X on date Y" - pretty much useless for evaluating AI workflows.

This realization hit us hard, and within a month we completely committed to a new mission:

We are going to help you make sense of AI in your product

Our first step was mapping what companies were doing. After dozens of calls and meetings, we identified two clear patterns:

• Product managers and engineers painfully reviewing agent runs manually, forced by messy LLM prompts to sift through long text chunks for tiny insights into what works.

• Others took the "build it yourself" approach, creating in-house tools for stakeholders to ask basic questions about their data. These projects typically fragment, start and stop with changing ownership, and miss the company's actual needs.

We didn't know the exact solution, but the gap was clear: companies had no way to measure their AI's impact. Sticklight would become the missing analytics layer that makes AI effects visible.

To validate our assumptions, we spoke with over 100 teams building AI products across enterprises, mid-sized companies, and startups.

The feedback was clear: enterprises recognized this as a significant future challenge, though they're currently focused on scaling AI. Meanwhile, startups and mid-sized companies already manage complex AI workflows and actively seek analytics to improve quality.

🧪 October 2024: Designing The Solution

Based on our research, we've decided to focus on younger startups as our initial target. These early adopters are deeply immersed in the AI ecosystem and eager to maximize their initial traction. This means quicker feedback cycles for us and faster value realization for customers. As the product and market mature, we'll expand to larger, more established customers.

Together with selected users, we began designing Sticklight.

We learned Figma and created product mockups to establish a clear vision of the workflow and core components. Working in 2-week cycles, we alternated between product iteration and feedback gathering.

After a few weeks, the user journey became clear:

1. Users send data

2. Sticklight converts data into meaningful views (conversations/users/agent runs)

3. User formulates a question

4. User enriches data with dimensions that help answer the question

5. User gains insights through charts

Our next goal was to build the platform and begin onboarding users.

💸 November 2024: Raising Capital

Thanks to high conviction from our early collaborators and market validation, we raised our first funding round of $2.2 Million in just 3 weeks.

For the following three months, we started working day and night to make the platform a reality.

🚀 Feb 2025: MVP & First Customers

In early January 2025, we had a working prototype and began dogfooding our own system. Despite many bugs, the core user journey was now possible. We tested our hypothesis on our own AI agents and generated data across several use cases to verify if the solution was robust enough for different scenarios.

Today we're onboarding beta customers, gathering feedback, fixing bugs, and preparing for launch. We see customers light up when they grasp the possibilities opening before them, and we're deeply excited about every new user and use-case being built on the platform 🤩