Stackpoint

A widely cited MIT study found that most corporate AI pilots deliver no measurable return. The exact figure has been challenged, but the broader pattern holds across the 2025–2026 research. Two findings deserve more attention than the headline number. 1. Most failures trace to the engineering around the model rather than the model itself: integration, governance, evaluation, cost control. The model call is a small share of an enterprise-grade agentic system. The surrounding system decides whether it holds up in production. 2. Systems built with an external partner succeeded roughly twice as often as those built internally. The systems that work tend to embed in one workflow, adapt to the operator's context, and expand from a narrow start. That profile fits a vertical AI company built around a real workflow with a design partner inside it. Our latest piece looks at what the research shows (link in comments)

In site feasibility, most of the early work is manual — pulling zoning codes, flood maps, utility records, and engineer notes into a spreadsheet before you can even make a go/no-go call. Dirt does that for you, so the decision takes minutes instead of months. Nick Kouloheras, Senior Land Development Manager at Barron Collier Companies, on what changed for his team: "Workflows that traditionally required months of effort are now being completed in a fraction of the time. Our team is making development decisions faster and with greater confidence." Join our waitlist today → https://godirt.ai/

Excited to see our portfolio company, LoanLight, Inc, selected to present at the ACT Tech Summit. We partnered with Max Klein to tackle a problem hiding in plain sight: non-QM lending is a $100B+ market, yet it still lacks a system of record for underwriting quality and eligibility. That gap creates friction, inconsistency, and unnecessary manual work across the lending process. LoanLight is building the first AI-native underwriting platform designed specifically for non-QM lenders. If you're at ACT, make time to see the team in action.

We launched SurfaceAI with Jason Wallis because we believed AI could transform how multifamily operates, starting with the work buried in documents nobody has time to read. This Thesis Driven deep dive from Brad Hargreaves tells that story. It walks through how SurfaceAI uses computer vision and autonomous agents to take on multifamily's hardest, least-loved workflows: lease files, rent rolls, vendor contracts, property transitions, and financial reviews. With SurfaceAI's Lease Audit AI Agent alone, operators report a 6-7x return relative to software cost. Read the full deep dive below.

We spend most of our time inside enterprise workflows figuring out how to build agentic AI that work. We get this question a lot from operators: why can't we just vibe-code this ourselves? It's a fair question. Prototyping has never been easier. But we've also watched the pattern that follows. An internal team ships a working demo in a weekend, scopes the full build against the demo they saw, and ends up having to rebuild the entire system eighteen months later. The prototype is roughly 10% of the work. The other 90% is integration plumbing, audit trails, auth, error handling, eval frameworks, and the thousand decisions that make software safe, observable, and improvable. That gap is an order of magnitude, and it's invisible in the demo. Before committing real time and money, every operator should ask themselves two questions: - Who is going to use this thing — just me, or other people? - What happens if it's wrong — a shrug, or something real? Plot those as axes and the answer becomes obvious. Vibe coding is genuinely powerful in the bottom-left. It cannot get you to the top-right, because everything around the AI is what determines whether the system holds up. The model call is the same in every quadrant. What changes is everything around it. We wrote a full breakdown of where vibe coding shines, where it breaks, and what becomes required as you move up and to the right. Link in the comments.

Congrats Matthew Kligerman and the DirtAI team!

Stackpoint Co-founder and General Partner Adam Pase and Family Office Exchange Executive Chairman Nathan Hamilton explore how family offices can build new companies from the operating expertise and proprietary data they already have. The shift is from "What should we invest in?" to "What should we build?", turning decades of workflows, domain knowledge, and customer relationships into enterprise-grade software — assets that have always held value but were historically hard to productize. Two of the case studies feature Stackpoint portcos: DirtAI (with Baron Collier Company) in land development, and SurfaceAI (with Coastal Ridge Real Estate) in multifamily operations. Both started with an operating family recognizing a problem in their own business, then partnering early to build the solution. Thanks to Nathan Hamilton for the partnership on this thought leadership piece. Link in comments.

Welcome Mehool Patel to the Stackpoint family as Head of Engineering! The hardest problems in complex industries such as real estate, construction, insurance, and finance sit where AI meets real-world operations: messy data, evolving workflows, multi-source inputs, non-trivial failure modes. Building products that hold up in that environment takes more than good models. It takes the evaluation harnesses, infrastructure, and agentic patterns that separate a demo from a system operators actually trust. We are thrilled to bring on Mehool to lead the charge and shape the technical direction of our AI-native companies from 0→1 through production. Welcome to the team!

We're hiring a Principal Engineer at Stackpoint! This isn't a one-product role. You'll embed with portfolio companies, build 0→1, ship production systems, hand off, repeat. Full-stack: Python/Node + React/NextJS. Production agentic AI and LLM integration. Greenfield every time — no legacy, just build. DM me or apply directly 👇 https://lnkd.in/eHYhqENJ

Welcome Zaheen Rahman to the team as Principal Engineer! Zaheen brings deep experience building AI products at startup speed. At ImagineCreate.ai, he built AI models and generation pipelines from scratch, led a fast-shipping engineering pod using agentic coding, and delivered across the full stack. At Stackpoint, he'll be partnering with founders in our portfolio to launch purpose-built AI products in complex industries where the problems are complex, the data is messy, and the opportunity to change the status quo is 10X! Excited to have Zaheen onboard!