decision model

Train your own data to unlock what your decision models can do

Sqwish Labs is asking a smart question: why settle for a generic decision model when you can train it on your own data?

3 min readTechCrunch
Train your own data to unlock what your decision models can do

For years, decision models have been treated like black boxes, powerful, yes, but fundamentally opaque and disconnected from the specific context of your work. Sqwish Labs is challenging that assumption with a straightforward proposition: let users train their own data to unlock what those models can actually do. This is a practical shift, not a theoretical one, and it signals a maturing of the AI tools that many of us have already started weaving into daily workflows. We think Sqwish Labs has landed on something that could matter more than the next incremental feature release.

The insight here is almost deceptively simple. A decision model trained on generic data is like a consultant who knows the industry but has never seen your spreadsheets. It can give you a framework, but it can't tell you where your own patterns break. By letting users feed in their proprietary data, sales figures, inventory churn, client history, Sqwish Labs is effectively turning the model into a tool that understands your specific constraints rather than just applying general rules. This is a direct contrast to the approach we've seen in other corners of the industry, where AI agents are reshaping investment banking workflows by automating entire deal processes. Both paths aim to reduce friction, but Sqwish Labs is betting that the real unlock comes from personalization, not just automation.

What makes this worth watching is that it lowers the barrier to entry for teams that don't have dedicated data scientists. You don't need to build a model from scratch or negotiate with a vendor to customize a proprietary system. You bring your own data, you train the model, and you get outputs that reflect your reality. That's a fundamentally different value proposition from the one offered by the major cloud players, who have been pouring capital into foundational models, Amazon and Google combine for nearly 30% stake in Anthropic after IPO, to build ever-larger general-purpose systems. Sqwish Labs is going the other direction: smaller, more specific, and more immediately actionable for the person staring at a spreadsheet on a Tuesday afternoon.

Here's the concrete takeaway: the next wave of productivity gains in data work won't come from bigger models. It will come from models that are willing to listen to your data. If Sqwish Labs executes on this, the question shifts from "what can AI do?" to "what can AI do with what I already know?" That is a much more interesting question, and it's one that every team relying on decision models should be asking right now. The specific detail to watch is how well the training process scales, can a non-technical user actually feed in a messy dataset and get reliable results, or does the promise still require a layer of expertise? That answer will determine whether this wrinkle becomes a standard feature or just a clever footnote.

From TechCrunch

Sqwish Labs has a new wrinkle on the decision model: Train it with your own data to max its capabilities.

Read the original at TechCrunch