Why We Embed Engineers Instead of Selling Software

Why didn't we build Soren as a software product?

Because every business we talked to runs a little differently.

Most AI products are point solutions. They pick one industry, one workflow, and build a clean version of it. That works well for the companies whose process matches the product.

For a lot of small and mid-market businesses, it doesn't match. So they ask for customizations. And a product team serving thousands of customers can't answer every ticket that starts with "we do it a bit differently here."

The average company already runs about 305 SaaS apps.¹ Usually the problem is that the tools don't fit together, or don't fit the way the team actually works. MuleSoft's 2025 benchmark found that only 29% of the average enterprise's applications are integrated.²

So we built Soren the other way around. Instead of asking a business to change how it works to fit our product, we build for how the business already works.

What does an off-the-shelf tool miss?

How your business actually runs, especially the handoffs between systems. Think about a mid-market law firm.

They might already have AI tools for billing, for intake, for accounting. Each one is good at its own job. But the firm's flow doesn't look like the flow those tools were designed for. Matters move between systems in an order that's specific to them. The judgment their people have built up over years lives in the handoffs, not in any one tool.

What they actually need isn't a fourth tool. It's AI that sits on top of the systems they already use and works the way their team works.

That's the part I think gets missed. We don't want to rip and replace. We want to build on what people are already used to.

You can see this beyond our own work. MIT's 2025 study of enterprise AI found that tools built with outside partners reached deployment about 67% of the time, compared with about 33% for tools companies built on their own.³ The researchers are careful to say that's a correlation, not proof. But it matches what we see: the hard part is fitting AI into real work, and that takes people who are close to the work.

What do our engineers actually do in the first weeks?

They listen first. You get the whole engineering team, not one person assigned to your account.

In the first two weeks, on site or virtual, our engineers meet the team and sit down one-on-one with the people doing the work. They learn how it actually runs, not how the process doc says it runs. By the end of week two there's a clear scope and goal, and we're already building in the background.

Then it's iteration. We build working versions fast, test them internally, and put them in front of the people who'll use them. The first workflow is usually something with clear, measurable value, and it's typically live within two to three weeks of getting access to the systems it needs. You can read more about how we work.

The client owns everything we build. And we don't hand it off and leave. Our engineers stay to monitor, support and improve it.

What's the honest downside of this model?

It takes a lot of time. Our engineers spend weeks inside each client's business. It also asks more of the client up front: their team has to make time for interviews and testing, and it costs more at the start than a tool that happens to fit.

We're okay with that. Our own team runs on AI every day, which lets us stay lean and still give each client the time they need. And every engagement makes the next one faster. When we've solved a similar problem before, we're not starting from zero, because we already know what worked. What we build for a client stays theirs.

Why are more AI companies embedding engineers now?

Because the industry is learning that AI only pays off once someone fits it to real work.

Forward-deployed engineers work directly inside a customer's business, and the role has become one of the most talked-about in AI. Job postings for it grew more than 800% between January and September 2025, according to Indeed data.⁴ Andreessen Horowitz called it the hottest job in startups, arguing that enterprise AI needs deep integration and context to work.⁵

I don't think that's a coincidence. Frontier intelligence keeps getting better. What's still missing for most businesses is someone who can connect it to how work actually gets done. That's the deployment gap, and it's why we built Soren this way.

When should you just buy the vertical tool?

If a vertical AI tool fits your process almost exactly, and you're fine handling a few manual steps around it, buy the tool. It'll be faster and cheaper, and that's the right call.

We'd rather tell someone to buy the tool than sell them something they don't need. For a broader comparison, we wrote about AI-native firms versus traditional consulting.

AI isn't something most businesses can just buy. It's something you build into how your business already works.