Shackleton, a New York company selling what it calls an agentic operating system for development and construction management, emerged from stealth on 4 November with a $5 million seed round led by Battery Ventures. Brick & Mortar Ventures participated, alongside Trust Ventures, Correlation Ventures, Motivate Partners, Thayer Investment Partners, Act One Ventures, Brainstorm Ventures and Silence VC, with the release noting others not named. The release does not state a valuation or a total raised to date, and a search for a reported figure did not surface one.
$5 million is a small round by the standards of this sector’s recent weeks. What makes it worth reading closely is who wrote the cheque, and what the founders did before they wrote software.
Battery Has Now Funded Two Construction Operating Systems in Five Weeks
On 30 September, Battery Ventures led Miter’s $40 million Series B. Miter’s own boilerplate calls its product an “AI-native operating system”; Shackleton’s calls its product an agentic operating system. The two are not competing for the same workflow — Miter starts at payroll and builds outward through HR, time tracking and job costing, the money side of the back office, while Shackleton is pointed at project administration: the correspondence, the logs, the approvals.
But the pattern is the same bet placed twice, five weeks apart, by the same firm: that a contractor’s software should be a single system of record with agents working inside it, rather than a best-of-breed stack with integrations between the parts. Michael Brown, a general partner at Battery, is quoted in the release saying “Construction is one of the largest and least digitized sectors in the economy” — the standard thesis sentence, and the one the firm is now backing at two different layers of the same building.
The Founders Ran a Contractor First
The three co-founders — Michael Ragheb, the chief executive, with Hassan Yasin and Olexandr Kaltsev — previously grew Ernest, a technology-enabled commercial general contractor, to $35 million in revenue across the southeastern United States, per the release. Ragheb’s account of the problem is a contractor’s rather than a vendor’s: “I was spending more time moving data between systems than running my jobs at Ernest.” He puts the cost in margin terms in the release: “Every hour a project manager spends chasing status across disconnected tools is margin lost.”
That background is the most load-bearing detail in the announcement, and it is also the hardest for a reader to check. What can be checked is that the product’s scope reads like it was specified by someone who had done the job. The platform covers site selection, preconstruction, project and field management, finance, communications and project turnover. Its agents, per the release, draft RFIs, follow up on overdue submittals, flag budget variances before payment requests are submitted for approval, and turn field notes into daily reports. The company’s site says sheets stay readable and annotatable with no signal, reconciling when the crew is back in range. Its integrations page lists four systems, all on the accounting and ERP side: QuickBooks Online, QuickBooks Desktop, Sage 300 CRE and Sage Intacct.
The Approval Gate Is in the Release, Not Added Afterwards
The release states that human review and approval workflows remain part of the process. That sentence is becoming the standard disclosure in this category rather than a differentiator, and this publication has noted the pattern twice in recent weeks: a Sydney builder’s software opened itself to AI agents and put a human in front of the money scope, and CMiC made permissions the headline feature of the MCP server under its ERP.
The detail that matters in Shackleton’s version is where the gate sits. An agent that flags a budget variance before a payment request is even submitted for approval is doing work well upstream of the irreversible step, which is the cheap place to put automation. An agent that chases an overdue submittal is sending correspondence in a contractor’s name, which is not reversible in the same way a draft is. The release does not break down which actions require approval and which do not, and that distinction is the one a buyer should ask about.
$1 Billion Is a Throughput Figure
Shackleton says the platform, which launched in June 2026, now manages more than $1 billion in project value across more than 15 states. This is the company’s figure and is reported as such. It is worth being precise about what it measures: project value running through the software, not revenue, not valuation, and not a count of customers. For a product five months after launch it is a throughput claim, and throughput is the right thing for a system-of-record pitch to be measuring.
The named customers are the unusual part. The release identifies three multi-location brands — Red Lobster, Dog Haus and Pepper Lunch — alongside a target market of midsize general contractors in multifamily, industrial and commercial work. Andi Cramer, senior director of construction and rebuild at Red Lobster, is quoted in the release saying that consolidating communication and project tracking on Shackleton has kept the chain’s rebrand on track and on schedule. Ragheb says in the release that multi-location customers have finished buildouts up to 35% ahead of schedule and as much as 18% under budget; those are the company’s figures, offered as ranges and without a baseline or methodology in the release, and this publication found no independent verification of them.
Restaurant rollout programmes are a sensible wedge for this product. A chain rebranding its estate is running the same small project over and over with the same approvals and the same submittals, which is the shape of work where agent automation has the least ambiguity to deal with — and the shape least representative of the midsize general contractors the company says it is aiming at.
