The engineer that never leaves.

The model is no longer the bottleneck

Something fundamental has changed in enterprise AI.

For the past two years, the conversation has been about models. Bigger models, smarter models, lower costs, better benchmarks. That race matters and it isn't over. But inside real companies, something else became true: the model is no longer the bottleneck. Deployment is. You can see the whole industry reaching this conclusion at once, from AI labs hiring forward deployed engineers to enterprises and systems integrators reorganizing around the same problem. Building intelligence stopped being the hard part. Getting it to work inside a business is.

A $60 million Series B

Today we're announcing a $60 million Series B led by Energize Capital, with Insight Partners increasing its investment and new strategic investors including Home Depot Ventures and United Rentals joining us. This brings our total funding to nearly $100 million, following a year in which our revenue grew 5x. The round matters. What we believe matters more, so that's what this post is about.

Deployment is not an engineering problem

Most people treat the deployment gap as an engineering problem. It isn't. Deployment is hard not because models are difficult to access, but because every business has its own operating logic, and most of that logic was never encoded in software. Having spent years working alongside distributors, manufacturers, logistics operators, field service companies, and healthcare systems, we've seen where it actually lives.

Companies already know how their business works: how they price a deal, why one customer gets an exception and another doesn't, how dispatch changes when two technicians call out, which compliance rule can never be broken. That knowledge is scattered across experienced people, operating habits, documents, and systems that have accumulated decades of decisions. Deploying AI means learning that logic, turning it into software, and keeping it current as the business changes.

The engagement ends

That's what forward deployed engineers actually do. The title makes it sound like an integration job, but the real work is learning how a business operates and translating that understanding into software that can run inside it. And for decades, that model had a structural flaw: the engagement ends. People learn the business, software gets built, the team leaves, the business changes, and the cycle starts over. The software remained. The understanding walked out the door. Hiring more people to close that gap might scale for a few more years. I don't believe it scales for decades.

The engineer that never leaves

So we built Lily to do the job itself. Lily learns how a business works, builds the software the business needs, deploys it inside the systems already running the business, and then stays. That last part is the whole point. As prices change, policies shift, systems get replaced, and operators correct it, Lily keeps the software aligned with how the business actually runs. A human forward deployed engineer eventually hands over the system and leaves. Lily remains inside the operation.

This is a different goal than most AI products have. A coding agent makes a developer more productive. A copilot makes an analyst more productive. Lily isn't built to make an individual more productive. It's built to make the business itself more capable.

The judgment was never in the software

One thing surprised us along the way. Across every deployment, the limiting factor was rarely the software companies already owned. Their systems recorded the business faithfully: every order, every ticket, every route. What those systems couldn't do was use the judgment surrounding the work, the pricing decisions, customer exceptions, and operational calls that experienced people make all day. Once we saw that pattern in one industry, we started seeing it everywhere.

Live today, not in pilots

Lily is live today inside some of the largest operators in construction, industrial distribution, logistics, field service, and healthcare. At one national distributor, it identifies and pursues opportunities across the entire customer base, generating more than $200 million in new annual sales that no commercial team could realistically cover manually. At an industrial supplier, quotes that took hours now take seconds. Field technicians who spent fifteen minutes diagnosing equipment now get answers in under ten seconds, and cost to serve dropped 98 percent. Logistics teams that replanned routes in fifteen minutes now do it in three. None of these are pilots. They run every day.

The intelligence system behind Lily

Behind Lily is InstaBrain, the intelligence system that decides how work should be done. It isn't tied to any one model. It routes each task to the model and environment best suited to it, checks the work before it ships, and improves those decisions over time, with one objective: the highest intelligence per parameter and intelligence per watt for every piece of work.

  quote a 4,812-line rfq ───▶ frontier-lg ───▶ shared cloud    ✓
  replan 42 routes       ───▶ small-fast  ───▶ edge · small dc ✓
  redact patient phi     ───▶ on-prem-md  ───▶ on-premise      ✓

That intelligence runs where the work runs, in the cloud, on premise, or at the edge through our Small Data Center, built with NVIDIA technology and informed by our work with Google DeepMind, keeping customer data private while lowering latency and cost.

Making businesses more intelligent

This is why we believe enterprise AI is entering a different phase. The first phase was about making models more intelligent. The next is about making businesses more intelligent. It changes where value is created, how software is built, and what enterprise software ultimately becomes.

We started in the physical economy because it's where operations are most complex and deployment is hardest. What we've learned there applies well beyond it. We'll use this round to make Lily more capable, expand into new industries, keep building the Small Data Center, and grow our teams in New York, San Francisco, London, and Toronto.

For decades, enterprise software asked businesses to adapt to it. We believe the next generation will adapt to the business instead. Every company has spent decades learning how it works. That knowledge shouldn't disappear into meetings, spreadsheets, and the heads of experienced employees. It should become software that stays with the company and improves alongside it.

That's the future we're building with Lily.

If your business runs on hard-won operational judgment, we'd love to show you what Lily can do. And if you want to build this with us, we're hiring.