Sim Is Building a New Way to Create Software

For the last twenty years, companies have solved problems by buying software.

Need a CRM? Buy one. Customer support? Another subscription. Recruiting, analytics, marketing automation, project management, expense reporting, internal tools. Each new workflow usually meant another application, another database, another integration, and another place where company knowledge became trapped.

The result is familiar inside almost every modern business: dozens of SaaS products, each solving one part of the company’s work and rarely understanding what is happening everywhere else.

AI creates the possibility of a very different model.

Instead of buying a new application every time a workflow changes, a company can increasingly build software around the way it actually works. Agents can reason over company data, connect to existing systems, execute processes, and adapt as those processes evolve.

Sim is building the workspace where that software gets created and operated.

From Workflows to Software

Sim started as a visual way to build AI agent workflows.

Its canvas let developers connect models, tools, and integrations together, then deploy the resulting workflow as an API, chat interface, scheduled job, or automation. The early insight was that agent frameworks gave developers flexibility but required significant code, while simpler visual tools often became restrictive as workflows grew more sophisticated. Sim wanted to sit between those extremes.

The product has expanded considerably from there.

Today, Sim describes itself as an AI workspace for building, deploying, and managing agents. A workspace contains not just workflows, but the resources those workflows operate on: tables, files, knowledge bases, integrations, and shared context. A team can describe a system in natural language and have Sim scaffold it, open the workflow visually to understand and refine the logic, or use the API and SDK when code makes more sense.

That evolution matters because the interesting unit is no longer an individual automation.

It is the system around it.

Imagine an internal recruiting application that knows the company’s candidates, interview notes, hiring criteria, and open roles. Or a support system that understands customers, product documentation, prior conversations, and escalation rules. Or an operations application built around the exact processes of one company rather than the average needs of thousands of customers.

Historically, building those applications required a software team.

Increasingly, Sim can make them something a technical team composes.

A Different Software Stack

This starts to change the economics of SaaS.

Most companies buy software because building and maintaining their own applications has historically been too expensive. A purpose-built CRM might fit the business better, but Salesforce already exists. A custom recruiting platform might reflect how the company actually hires, but buying an ATS is easier than employing engineers to maintain one.

Agents and code-generation systems begin to weaken that tradeoff.

If the cost of building software falls dramatically, companies can become more opinionated about what they use. Instead of forcing a workflow into the structure of an existing SaaS product, they can increasingly build the workflow itself.

Sim adds another important piece: the data and processes can live together.

Its workspace model lets workflows operate over shared tables, files, knowledge bases, and integrations rather than treating every agent as a standalone automation. One workflow can produce information that another consumes. Agents can work from the same underlying context. The company begins to accumulate a software layer built around how it actually operates.

That is a much more ambitious idea than automating tasks.

It points toward companies building more of their own software.

Built by People Who Started Early

Emir Karabeg and Waleed Latif came to Sim with unusually relevant backgrounds for this moment.

Emir studied Data Science and Cognitive Science at UC Berkeley, where he worked on natural-language processing and machine translation. Before Sim, he built WorkNinja, an AI learning platform that reached more than 300,000 monthly active users and was experimenting with chain-of-thought applications at the end of 2022, unusually early in the current generative-AI cycle.

Waleed studied computer science and cognitive science at Berkeley and worked on neurotechnology research before joining Amazon, where he built core backend infrastructure for Ring. His background combines an understanding of intelligent systems with experience building infrastructure that has to operate reliably at enormous scale.

The company itself started with a practical experiment. Emir and Waleed wanted to rebuild their previous company’s sales and marketing operation entirely through a visual framework. If they could reproduce the workflows of a real business using composable blocks, they believed the same approach could eventually be used much more broadly. That experiment became the foundation of Sim.

They have been thinking about the same underlying question from the beginning: how much software can a company build for itself once the cost of creating it collapses?

What Happens to SaaS?

We don’t think SaaS disappears.

There will always be categories where a standardized product makes more sense than something custom. Companies will continue buying software where shared functionality, compliance, networks, or specialized infrastructure matter.

But the boundary between buying and building is moving.

Today, a company might subscribe to seventy-five different applications because each workflow requires its own system. In the future, many of those workflows could be assembled from agents, shared data, and integrations inside a common environment.

That could produce a very different enterprise stack. Fewer isolated applications. More software built around the company’s own data and processes. Agents that can work across functions because they operate from common context rather than hopping between disconnected tools.

Sim has the opportunity to become the place where that software lives.

The company already supports more than 1,000 integrations and every major LLM, and its current product spans agent building, shared context, deployment, monitoring, and production operation. More than 100,000 developers are using the platform.

The long-term question is whether a workspace like Sim can become an alternative application layer for the enterprise.

We think it can.

Why We Invested

We invested in Sim because we think AI changes more than how software gets built. It changes what software companies need to buy in the first place.

Emir and Waleed were early to that shift. Emir had already built a large AI-native product before ChatGPT made the category obvious. Waleed brought the infrastructure experience required to turn experimental systems into reliable software. Together, they started by trying to recreate the workflows of a company and ended up building a platform for anyone else to do the same.

The product has continued moving in that direction. What began as a canvas for agent workflows is becoming a workspace where a company’s processes, data, agents, and applications can live together.

For decades, the default answer to a new business problem was to buy another piece of software.

Sim is betting that increasingly, the better answer will be to build your own.