$273B yearly loss vs 13.5% AI adoption in construction

Funding Surges as Deals Consolidate

PermitFlow's $54 million Series B in November, led by Accel, pushed the New York startup's valuation above $500 million and capped a year where construction technology funding hit $2.2 billion across 199 rounds, a one-fifth increase over 2024 even as deal count dropped by nearly a fifth. Data tracked through December shows fewer rounds, larger checks, and a clear concentration around companies applying agentic AI to the industry's least digitized workflows: preconstruction, permitting, procurement, and the financial backbone of billing, accounts receivable, and cash-flow management.

Investors are not betting on incremental ERP add-ons. They are funding AI-native platforms built from the ground up on agentic AI (software that doesn't just surface data but acts on it) to automate the core financial and administrative workflows where legacy systems have failed to deliver. The construction software market, long dominated by Sage, Viewpoint, and CMiC, now faces competitors that replace the workflows those systems were built to manage.

Attentive.AI, based in Wilmington, Delaware, raised $30.5 million in a Series B led by Insight Partners, Business Wire reported; its Beam AI takeoff product serves more than 1,100 companies, delivering fully automated quantity takeoffs across major trades through a human-in-the-loop model that eliminates manual measurement work. "While AI has transformed other industries, preconstruction — where every job can be won or lost — still primarily relies on manual takeoffs and spreadsheets," the company said in its November announcement. San Francisco-based Kojo secured a $10 million Series C extension from Wesco International, the Pittsburgh-based distribution and logistics firm, merging Kojo's materials and inventory management AI with Wesco's distribution network to eliminate manual data entry and email-based tracking while deploying AI agents for procurement, scheduling, and distributor follow-ups. ConCntric, a Greenbrae, California preconstruction platform, raised a $10 million Series A led by 53 Stations, GlobeNewswire reported, to expand Amplify, its agentic AI that completes tasks on behalf of users rather than merely surfacing insights. "Preconstruction is the most important and least digitized phase of the project lifecycle," said Dell'Orto, ConCntric's founder and CEO. Unlimited Industries, an AI-native construction company in San Francisco, closed a $12 million seed round co-led by Andreessen Horowitz and CIV. Planera, in Pleasanton, California, added $8 million to build scheduling tools tailored for data center construction. As vital infrastructure, industrial facilities, and data centers enter aggressive rollout phases across the U.S., contractors face pressure to deliver more projects faster, Construction Dive reported.

Why Cash Flow Still Bleeds

The construction industry bleeds an estimated $273 billion annually from payment delays and cash-flow crunches — a figure that hasn't budged despite decades of ERP investment. Eight in ten firms cite cash flow as a top challenge. The structural reasons are stubborn: project-based billing cycles, multi-tier payment chains from owner to general contractor to subcontractor, retainage withholds that sit for months, and lien rights that demand active management. Generic accounts-receivable advice doesn't map to this environment; it was built for recurring SaaS subscriptions, not progress billings where the concrete subcontractor won't get paid until the owner releases the third draw.

Legacy automation hasn't solved it. Rules-based systems send pre-written emails at 30, 45, and 60 days past due: no intelligence, no adaptation, no ability to handle anything outside the script. Robotic Process Automation adds cross-system navigation but follows brittle scripts; a bot can't reason about a debtor's situation or handle an unexpected response during a phone call. Most legacy AR platforms operate at this level. The gap is execution: "automated" billing software captures and routes data but stops cold when a retainage calculation conflicts with the contract or a change order hasn't reached accounting yet. A human must review even minor exceptions.

Agentic AI operates at a fundamentally different level. Each agent has a goal (recover the account), a toolkit (email, phone, SMS, payment links, dispute resolution), and the autonomy to decide how to use them. It reasons about context, adapts in real time, and handles the full spectrum of debtor interactions including natural-language phone conversations. The technology ingests unstructured inputs (scanned PDFs, emailed invoices, portal uploads) and maps line items to the correct project, cost codes, and vendor accounts without human sorting. It then orchestrates the full workflow: data capture, validation, matching, exception detection, approval routing, and payment scheduling. Over time it learns: flagging the vendor that repeatedly miscodes invoices, suggesting process improvements, and prioritizing discount-eligible payables to optimize working capital.

The results are measurable. Adaptive, launched in February 2023, now serves more than 280 construction companies managing over $1.4 billion in project volume. One mid-sized contractor processing 500 invoices monthly cut data-entry errors by roughly 92 percent and reclaimed 20 hours of staff time each week. Itemize's deployment at Arguijo Corporation delivered a 50-percent-plus reduction in initial invoice processing time, eliminated PDF conversion steps, and enabled two-way purchase-order matching that lets coders dispute vendor errors instantly. EY's 2024 Global Construction and Engineering Report found companies using AI-driven billing systems saw a 37 percent drop in overdue invoices and a 25 percent improvement in Days Sales Outstanding. Thoughtful.ai's 2025 data puts the DSO reduction at up to 20 percent for construction firms.

Collections show the sharpest contrast. AgentCollect's platform processes up to 85,000 accounts per day, recovering approximately half of placed accounts within the first 20 days — traditional agencies typically recover 15 to 20 percent over six months. Roughly 90 percent of disputes resolve without human involvement. Attorney-mode emails hit 70 percent open rates versus 20 percent for generic collection letters. Stuut's agentic AI has processed over $2 billion in B2B invoices and now partners with Fiserv to embed that capability into Commerce Hub and SnapPay, extending order-to-cash automation into working-capital management for enterprise clients.

Adoption is accelerating. A 2025 Deloitte poll found 13.5 percent of finance organizations already using agentic AI, with another third developing or planning it. Gartner's 2026 CFO survey ranks AI agent integration as the top technology priority for 54 percent of finance leaders — ahead of ERP modernization and data analytics. The deployment model is phased: phase one requires human approval on every email; phase two lets routine reminders under a set limit fire autonomously; phase three sends everything, with escalations (invoices 90 days overdue or balances above $25,000) routed to a human before the model even sees them. The agent never threatens legal action, credit reporting, or collection agencies. The final notice states only the consequence the company has committed to enforce.

Integration remains the practical hurdle. Agentic AI must mesh with existing ERP, accounting, and project-accounting systems so invoice data flows into the right project codes and cash-flow forecasts. Audit-grade documentation, compliance readiness for public-works bonding, real-time anomaly detection for duplicate or fraudulent invoices, and explainable decision logs are non-negotiable for infrastructure firms. KYRO AI addresses this with SOC 2 certification and an interface designed to minimize the learning curve. The companies moving fastest are those treating agentic AI not as a plug-in but as a controlled-autonomy layer that resolves routine exceptions within defined policies while pulling humans in when judgment, risk, or accountability demands it.

Incumbents Scramble to Bolt On AI

The funding surge didn't just buy startups runway. It put the three legacy ERP pillars — Sage, Viewpoint (now Trimble), and CMiC — on notice. Their combined install base still runs the financial backbone of most North American contractors. But the architecture underneath that install base dates to the 1970s (Viewpoint) or the Oracle-on-prem era (CMiC), and neither was built for agentic AI that can match remittance data to invoices across a multi-tier payment chain in seconds.

Viewpoint's answer has been quiet integration into Trimble's broader portfolio. Since the acquisition, Viewpoint Vista and Spectrum sit alongside Trimble ProjectSight and a handful of field tools. The pitch is ecosystem breadth: one vendor for estimating, field management, and accounting. What's missing is a native AI layer. The BuildAEC analysis is blunt: "Legacy architecture limits how quickly either platform can adopt AI-native features. While both vendors have announced roadmap investments in analytics and automation, neither currently delivers AI-powered functionality (such as ML-driven cost forecasting, AI takeoff, or intelligent bid analysis) as a native, production-ready capability." Neither Viewpoint nor CMiC offers native AI-powered quantity takeoff, a capability that AI-native preconstruction tools now treat as table stakes.

CMiC has moved louder and faster. In April 2025, the Toronto vendor announced NEXUS, branded as "the first AI-powered construction ERP." The rollout is phased: CONNECT 2025 in November marks the official debut, but the feature list is already specific. CMiC has shipped AI-powered API integrations, AI-generated workflows, AI-enhanced process automation, and AI-driven data entry screens. The system uses AI for data summarization, anomaly detection alerts, and predictive analytics. "Ask AI," built on OpenAI's Assistant V2 API with GPT-4, lets users query project data in natural language. Daily Sentiment Analysis scores each project's journal entries on a 1-to-10 scale, flagging crew morale and risk before they show up in cost reports. An AI drawing extraction feature, using an LLM to pull sheet metadata, is next in the queue. Steve Cangiano, CMiC's chief product officer, frames it as a multi-year journey: "Our long-term AI strategy is intentionally designed to leverage the power of AI within our core features and functionality. This approach will include phased rollouts and ongoing improvements to ensure continuous innovation."

Sage took a different lever: the developer platform. In April 2026, the FTSE-listed company unveiled Sage Agent Builder and an AI Gateway at its Sage Future conference in San Francisco. The tools give partners a structured way to design, test, and deploy AI-powered experiences inside Sage workflows: Sage Copilot, Sage Marketplace, and the core ERP lines (Sage Intacct, Sage X3, Sage Active). A unified developer experience across those three products means "build once, integrate more easily, bring solutions to market faster." Sage also introduced usage-based pricing and revenue-sharing models so partners scale costs with adoption. IDC's projection — a tenfold increase in third-party and custom AI agents used by enterprises over five years — underscores the bet: Sage wants to be the platform where those agents live, not the vendor trying to build them all in-house. The Sage CRM roadmap reinforces the pivot. A July 2026 partner advisory council session detailed an enriched semantic schema layered on the existing REST API, giving AI systems a richer map of business data relationships without loosening security controls. The Sage Semantic Engine will replace the Ally assistance layer, aligning CRM with the wider Sage AI portfolio. Beta arrives late September 2026.

The structural problem none of these moves fully solves is total cost of ownership. Implementation for Vista, Spectrum, or CMiC can rival or exceed the annual license fee. Both platforms demand dedicated IT staff; CMiC's Oracle backend all but requires a database administrator. Subcontractors routinely pay for GC-centric modules (owner billing formats, joint-venture accounting, document control) that don't map to sub workflows, forcing expensive customization. The BuildAEC comparison concludes: "For the large segment of contractors in between (specialty contractors, growing subcontractors, and mid-size GCs managing $10M–$200M in annual revenue) the total cost of ownership for either platform is difficult to justify when modern, AI-native platforms can now deliver comparable core functionality with faster onboarding, lower implementation costs, and better mobile-first field tools."

Legacy ERP vendors are no longer ignoring AI. They're embedding it, opening APIs, and courting partner ecosystems. But they're doing it on top of codebases that predate the cloud, let alone the transformer. The question for the next 12 months isn't whether Sage, Trimble, and CMiC will ship AI features — they already are. It's whether contractors buying or renewing in 2026 will accept a bolt-on agent layer on a 20-year-old foundation, or whether the AI-native alternatives — faster to deploy, cheaper to run, built for the payment-chain problems that actually bleed cash — force a platform switch the incumbents can't patch their way out of. That switch requires talent to build, deploy, and run the new tools — and the pipeline is cracking.

The Talent Pipeline Is Cracking

Nearly 1,400 firms responded to the Associated General Contractors of America and National Center for Construction Education and Research's 13th annual Workforce Survey, conducted in late July and early August 2025. Nine in ten firms that are hiring report difficulty finding qualified workers, with virtually identical rates for craft and salaried roles. Almost nine in ten have openings for craft workers; four in five have openings for salaried staff. Nearly half say worker shortages at their own firms or their subcontractors have delayed projects. More than three-quarters report at least one delayed project in the past twelve months.

Worker shortages are the leading cause of project delays, full stop. New immigration enforcement has compounded the problem: more than a quarter of respondents report direct or indirect impacts in the past six months. Five percent had a jobsite or offsite visited by immigration agents. One in ten lost workers to actual or rumored enforcement actions. One in five say subcontractors lost workers. The impact varies sharply by state: three-quarters of Georgia firms affected, just over a third in South Carolina. Only one in ten firms uses the H-2B visa program or other temporary work visas.

The pipeline is leaking at every joint. Nearly half of new hires fail to show up or quit shortly after starting. Two in five potential employees lack required credentials: driver's license, work permit, clean background check. One in three cannot pass a drug test. Nearly a quarter cite unreliable transportation. More than one in five need flexible schedules or remote options, often due to childcare gaps. More than half of firms say available candidates simply lack essential skills or appropriate licenses.

Meanwhile, the AI-native construction tech boom is creating a parallel talent crisis. Startups automating billing, accounts receivable, and cash-flow management need software engineers who understand construction's multi-tier payment chains, retainage mechanics, and lien-rights workflows. They need data scientists who can train models on messy, project-based financial data. They need product managers who have walked a jobsite. The Bureau of Labor Statistics projects strong job growth in computer and mathematical occupations through 2034 driven by AI adoption, but construction tech competes with every other sector for that talent.

Firms are responding. Seven of eight contractors raised base pay, many matching or exceeding the prior year's increases. More than two-fifths boosted training and professional development spending. More than half added digital recruiting (social media, targeted advertising) to reach younger applicants. More than half partnered with high schools, colleges, and career and technical education programs. AGC is urging Congress to at least double funding for high school CTE programs and pass a new Workforce Innovation and Opportunity Act allocating at least half of funds to workforce training. The association also wants a construction-specific temporary work visa program. NCCER President and CEO Boyd Worsham put it plainly: "These are high-paying, high-wage jobs that literally will break a poverty cycle. For some families, this changes generational poverty, and it makes a difference."

The workforce sees AI as ally, not threat. In the AGC/NCCER survey, forty-four percent said AI and robotics will improve job quality and make workers safer and more productive; forty-five percent said they will automate manual, error-prone tasks. Only twelve percent expect net job elimination. That optimism matters; if the industry can't recruit the people to build and run these tools, the fresh capital buys software that sits on shelves.

Salaries are already reflecting the squeeze. General AI talent commands premiums: Databricks lists senior roles at $340,000 to $605,000, Zero G Talent found; Anthropic's staff research engineers range $500,000 to $850,000, Zero G Talent's figures put. Construction tech startups won't match those bands, but they're bidding against them for the same engineers. The firms that win will be the ones that translate "construction domain knowledge" from a nice-to-have into a compensated specialty, and that treat the jobsite not as a recruiting afterthought but as a pipeline. The next funding round won't close the gap. The next graduating class might.

What Comes Next

For construction operators, the next 12 to 24 months center on data readiness. Space AI Research finds that firms with a common data environment and structured project data are 3.4 times more likely to report successful AI deployment. Deloitte's 2026 E&C Outlook urges firms to "institutionalize data governance frameworks, invest in continuous workforce development, build ecosystem partnerships, and embed digital performance metrics throughout project delivery" as cloud-native digital twins and AI agents become standard. The same report projects modest growth in structure investment for 2026, driven by AI-related data center outlays, but warns that "projects in early planning stages could be affected by signs of evolving government policies or economic slowdown." Operators who have not yet mapped their RFI, submittal, and change-order flows into a queryable layer will fall further behind peers who have.

CFOs face a sharper mandate: automate the cash-conversion cycle or watch margins erode. The largest checks in the recent funding wave went to startups that tackle billing exceptions, retainage tracking, and three-way invoice matching. Deloitte notes that "margin pressures from tariffs, volatile material costs, and ongoing supply chain disruptions are intensifying the potential need for firms to rethink traditional operating models" and that "financial agility will be equally integral to strategy." The AGC/NCCER survey of nearly 1,400 firms confirms the labor squeeze (half a million new workers needed in 2026, wages up 4.2% year over year as of August 2025), which means every manual hour spent reconciling pay applications is an hour not spent on higher-value work. Agentic AI that can draft lien waivers, flag overbilling, and predict retainage release dates moves from pilot to line-item budget necessity.

Software developers should expect the platform wars to accelerate. ZACUA VENTURES argues that "in an era where AI can write code, software 'features' are no longer defensible" and that competitive advantage now depends on "owning the Authoring Layer" and "Workflow Substitution." Market forecasts agree on direction; the spread reflects differing definitions of what counts as "construction AI."

Developers building on top of Procore, Autodesk Construction Cloud, or Sage Intacct should watch API changelogs closely; legacy ERP vendors are embedding AI copilots and opening ecosystems to avoid commoditization. The Deloitte 2026 Power and Utilities Outlook signals a parallel track: by 2027, nearly 40% of utility control rooms will use AI, and "performance-based interconnection could increasingly tie queue priority to telemetry and flexibility" in 2026. That utility-side digitization will pull construction data into new compliance and interconnection workflows.

Talent remains the binding constraint. Deloitte projects a potential shortage of over two million skilled craft professionals by 2028 if current trends persist, with 41% of the workforce retiring by 2031 and only 10% under 25. The same report notes "the migration of engineering talent to technology firms (driven by demand for tech-enabled skills) is intensifying competition for skilled workers." Firms that redesign career paths around AI fluency (not just tool training) will recruit and retain the hybrid profiles the next phase requires. The window to act is the next four quarters; after that, the gap between AI-native operators and laggards hardens into a structural cost disadvantage.

Working in AI? Zero G Talent tracks the openings: see every open Databricks role, browse AI jobs, openings at Anthropic, and the people building the field.