1. Executive Summary
The real estate, construction, and infrastructure sector sits at a turning point. For decades, the industry has been defined by fragmented workflows, paper-heavy processes, and slow feedback loops. That’s starting to change, not gradually, but all at once. Agentic AI is emerging as the connective tissue that finally ties together planning, execution, and operations into something closer to a real-time system.
Market opportunity for agentic AI (size, growth, urgency)
The opportunity is large, and it’s accelerating. Global spending on AI in construction and real estate is still early but expanding fast. McKinsey estimates that generative AI alone could add between $2.6 trillion and $4.4 trillion annually across industries, with real estate and construction capturing a meaningful slice through design automation, project management, and asset operations. Meanwhile, the broader enterprise AI market is expected to exceed $500 billion by the early 2030s, growing at over 20 percent annually. Within that, agentic AI, systems that don’t just assist but act, is one of the fastest-growing layers.
In practical terms, this means moving from tools that suggest to systems that execute. Instead of a project manager checking five dashboards, an AI agent can monitor progress, flag delays, reschedule crews, and even trigger procurement workflows automatically. That shift is what defines the next decade.
Key Thesis
The core thesis is simple but important. The industry is moving from SaaS to AI-native workflows, and from there to autonomous agents.
SaaS digitized processes but kept humans in the loop for every decision. AI-native workflows begin to reduce that burden, embedding intelligence into daily tasks. Agentic systems go further. They coordinate across systems, make decisions within defined boundaries, and continuously learn from outcomes. Over time, the interface itself starts to disappear. What matters is not the software you open, but the outcomes the system delivers.
Why now?
Three forces are converging.
First, large language models and multimodal AI have reached a level where they can reliably interpret contracts, drawings, site imagery, and sensor data. This matters in an industry where most information is unstructured.
Second, enterprise systems are finally interoperable enough to support automation at scale. APIs, cloud adoption, and standardized data models like BIM are creating the foundation agentic systems need to operate.
Third, the pressure to automate knowledge work is rising. Construction productivity has lagged other sectors for decades. McKinsey has consistently pointed out that large construction projects run 20 percent over schedule and up to 80 percent over budget. In real estate, asset managers face increasing complexity from regulatory compliance, tenant expectations, and energy efficiency demands. There’s a growing sense that incremental improvements won’t cut it.
What makes agentic AI different is not just efficiency gains, though those are real. It’s the ability to close decision loops faster than humans can manage at scale.
Early evidence backs this up. Companies like Buildots are using AI and computer vision to track construction progress automatically, helping reduce delays and improve schedule adherence. OpenSpace has documented faster site documentation workflows, cutting hours of manual work into minutes. ALICE Technologies uses generative scheduling to simulate millions of construction scenarios, helping teams find optimal build sequences that would be impossible to calculate manually. JLL has begun embedding AI into real estate operations, from lease abstraction to predictive maintenance, showing measurable gains in operational efficiency.
Across these cases, the pattern is consistent. AI moves from a passive tool to an active participant in operations.
Key Findings
The market for agentic AI in this sector is underpenetrated but expanding rapidly, with adoption expected to accelerate over the next 3 to 5 years as integration barriers fall.
The highest-value use cases sit at coordination points, scheduling, procurement, compliance, and asset operations, where complexity is high and data is fragmented.
Firms that adopt early are seeing measurable ROI through reduced delays, lower operational costs, and improved asset performance.
The competitive landscape is shifting. Traditional SaaS vendors risk being abstracted away unless they evolve into workflow orchestration platforms or embed agentic capabilities directly.
Data ownership and integration depth are becoming more important than model performance alone. The moat is moving closer to workflows and proprietary data.
2. Market Context & Scope
The sector is defined as the full built-environment value chain: real estate investment and leasing, design and preconstruction, construction execution, infrastructure delivery, and asset operations.
The practical scope is narrower: workflows where AI agents can read, reason, act, and escalate across systems such as Procore, Autodesk Construction Cloud, Primavera P6, Microsoft 365, ERPs, CRMs, BIM environments, document repositories, and field-capture tools.
Market segments
Construction execution and field operations
This is the most immediate market for agentic AI because the work is coordination-heavy and full of repeatable decisions. Project teams deal with RFIs, submittals, daily logs, progress tracking, safety observations, quality checks, change events, and schedule updates. Most of that work still depends on humans reading across systems and stitching together the story manually.
AI agents can help by watching for missing information, routing approvals, comparing field progress against plans, and escalating issues before they become claims.
Infrastructure and capital projects
Infrastructure has longer timelines, larger stakeholder groups, more public-sector oversight, and heavier governance requirements. These projects generate enormous volumes of documents, meeting notes, permits, design revisions, and contractual correspondence.
Agentic AI is especially relevant here because the risk is not just inefficiency. It is delay, compliance failure, budget overrun, and political exposure. Agents can support claims documentation, funding compliance, schedule-risk monitoring, and cross-stakeholder reporting.
Commercial real estate operations and leasing
Commercial real estate is a natural fit for document intelligence and workflow automation. Lease abstraction, tenant service, portfolio reporting, valuation support, renewals, maintenance dispatch, and compliance tracking all involve repeatable knowledge work.
The strongest near-term use cases sit in asset management and property operations, where agents can read lease clauses, interpret tenant requests, pull property data, and trigger next steps across CRM, accounting, work order, and reporting systems.
Residential real estate and mortgage-adjacent workflows
Residential real estate has high transaction volume and lots of repetitive coordination. Listing operations, lead handling, transaction support, valuation workflows, disclosure review, inspection follow-up, and compliance paperwork all create room for automation.
The challenge is that residential workflows are more fragmented and often depend on small teams with limited IT support. That means adoption may be faster for point solutions, but slower for deeply integrated autonomous agents.
Design, engineering, and preconstruction
Design and preconstruction are becoming one of the most interesting areas for AI agents. These workflows include specification review, clash issue triage, quantity takeoff support, bid package generation, cost history retrieval, constructability checks, and early schedule scenario modeling.
This segment matters because decisions made here shape downstream cost, risk, and schedule performance. A small improvement in scope clarity, estimate accuracy, or constructability review can prevent expensive rework later.
Adjacent markets
Agentic AI does not fit neatly into one software category. It overlaps with construction technology, PropTech, enterprise automation, robotic process automation, business process management, document AI, computer vision, business intelligence, and vertical SaaS.
That overlap is the point.
A submittal agent may touch email, a project management platform, a specification book, the schedule, the ERP, and a document repository. A lease agent may read a PDF, update a CRM, notify a property manager, and create a task for legal review. That is not just another dashboard or seat-based SaaS module. It is a workflow layer sitting across systems.
Market Segmentation Pie Chart
3. Market Size & Growth
The market is still early, but the growth curve is steep. Agentic AI is moving from “interesting pilot” to budgeted enterprise automation, and the built-environment sector is one of the more obvious landing zones because the work is expensive, fragmented, document-heavy, and painfully coordination-heavy.
TAM / SAM / SOM
Sources: Grand View Research estimates enterprise AI at $23.95B in 2024 and $155.2B by 2030, and enterprise agentic AI at $2.58B in 2024 and $24.5B by 2030. It also estimates AI in construction at $2.93B in 2023 and $16.96B by 2030. Intent Market Research estimates AI in real estate at $2.2B in 2023 and $18.4B by 2030. (Grand View Research, Grand View Research, Grand View Research, Intent Market Research)
Growth drivers
- Productivity pressure
Construction remains one of the least productivity-improved major sectors. Projects still suffer from schedule slippage, rework, procurement issues, and fragmented communication. That creates a strong business case for agents that can reduce coordination waste.
- Labor shortages and knowledge loss
Experienced project managers, estimators, superintendents, facility managers, and leasing teams are stretched thin. Agentic AI can help preserve institutional knowledge by turning scattered documents, emails, drawings, schedules, and logs into usable decision support.
- Data availability
The sector has finally built enough digital exhaust for AI to work. BIM models, construction management systems, drones, site cameras, IoT sensors, ERP systems, CRMs, lease databases, and work-order platforms now generate the raw material agents need.
- Enterprise readiness
The early “chatbot” phase is giving way to controlled automation. Companies are asking harder questions now: Can this agent access our systems? Can it follow approval rules? Can it explain its reasoning? Can it hand off to a human when risk is high?
- ROI visibility
The best early use cases have measurable outcomes: fewer delays, faster submittal review, lower rework, quicker lease abstraction, improved maintenance response, better schedule accuracy, and higher revenue per employee.
Adoption Curve
Growth Drivers Impact
4. Customer Needs & Jobs-to-be-Done
The customer need is not “AI.” Nobody wakes up wanting another AI product in the tech stack. Buyers want fewer fires, cleaner handoffs, faster decisions, and less time spent chasing information that should have been obvious yesterday.
In real estate, construction, and infrastructure, agentic AI earns budget when it does one thing well: removes coordination drag from high-stakes workflows.
Core problems
- Project and asset data is scattered, stale, or hard to trust
The industry has plenty of data. The problem is that too much of it is trapped in PDFs, drawings, email threads, field notes, spreadsheets, ERPs, CRMs, BIM models, point solutions, and people’s heads.
Autodesk and FMI estimated that bad data cost the global construction industry $1.85 trillion in 2020. In the same research, 30% of respondents said more than half of their project data was “bad,” and bad data was tied to $88.69 billion in rework costs, equal to 14% of all rework performed that year. (Autodesk)
FMI also reported that project data volume had doubled over the prior three years, while more than 80% of respondents described at least 25% of their project data as unusable. For a contractor with $1 billion in annual revenue, FMI estimated the cost of bad data at $165 million, including $7.1 million in avoidable rework tied to bad data. (FMI Corp.)
The customer pain is blunt: teams do not need more dashboards. They need systems that can find the right context, explain what changed, and trigger the next action.
- Coordination work is eating the operating model
Construction and infrastructure projects run through thousands of handoffs: RFI responses, submittal reviews, change orders, procurement updates, safety observations, schedule updates, permit dependencies, inspection results, and owner reporting. Each handoff creates delay risk.
McKinsey notes that construction technology uptake has historically been slow, with construction companies spending less than 1% of revenue on IT, far below sectors such as automotive and aerospace. It also notes that $50 billion flowed into AEC technologies globally between 2020 and 2022, but that investment has not yet produced a clear productivity uplift across the industry. (McKinsey & Company)
That gap explains the opening for agentic AI. SaaS captured the work. Agents can start coordinating it.
- CRE operations are still too manual
In commercial real estate, property teams face a different version of the same problem. The day is full of tenant requests, maintenance tasks, compliance checks, vendor follow-ups, inspections, and reporting.
A 2024 Building Engines and BOMA survey of more than 230 CRE professionals found that 46% of respondents viewed tenant issue management as one of the most time-consuming or automation-ready areas. Preventive maintenance and inspections came in at 38%, tenant compliance at 37%, tenant communications at 32%, and vendor procurement at 27%. (Cloudinary)
That is a strong signal for agentic workflows. A tenant issue agent, for example, does not just classify a ticket. It can check lease obligations, pull building history, prioritize the issue, notify the right engineer, draft the tenant response, and escalate if the SLA is at risk.
- Work prioritization is still too dependent on human judgment under pressure
Property and facility teams are not short on work. They are short on clean prioritization. In the same Building Engines and BOMA report, prioritization was cited as the top challenge for completing work orders, while staffing resources ranked as the second major challenge. The report also found that tenant comfort issues and faster responses to work orders were two of the most common tenant requests. (Cloudinary)
This is exactly where agents can help. Not by replacing property managers or engineers, but by removing the first 20 minutes of triage from every ticket.
- AI interest is high, but execution confidence is uneven
Corporate real estate leaders are not ignoring AI. JLL reports that 90.1% of companies expect to carry out CRE activities in a hybrid model where AI supports human experts over the next five years. JLL also found that 89% of leaders believe AI can help solve major CRE challenges, nearly two-thirds of companies have started piloting CRE AI use cases, and 73% of CRE professionals are early adopters using AI to support day-to-day work.
But there is a catch. JLL also warns that many companies are piloting without a comprehensive strategy, and notes a gap between C-suite enthusiasm and implementation readiness. That means buyers want AI, but they also want guardrails, proof, and a path from pilot to production. (JLL)
Desired outcomes
Customers are not buying autonomy for its own sake. They are buying confidence under pressure.
They want:
Buying criteria
The buying bar is higher in this sector than in many horizontal software categories. A bad AI summary in a sales workflow is embarrassing. A bad AI action in construction, infrastructure, or property operations can create cost, safety, legal, or tenant consequences.
5. Competitive Landscape
The competitive landscape is crowded, but not evenly mature. Most vendors are not yet “agentic” in the full sense of taking governed actions across systems. A lot of the market is still AI-assisted SaaS: copilots, analytics layers, visual capture tools, scheduling engines, document AI, and workflow automation.
That matters because Automatic.co should not define the competitive set too narrowly. The real fight is not just against other AI-agent startups. It is against every platform trying to become the operating layer for real estate, construction, and infrastructure work.
Direct competitors: agentic AI and AI-native workflow players
Competitive read
The strongest direct competitors are not necessarily the most “agentic” today. They are the companies with three assets agents need: system access, workflow context, and trusted data.
That gives Procore, Autodesk, JLL, OpenSpace, and Buildots meaningful advantages. They already sit close to the work. They see project records, design data, field data, leasing data, asset data, or customer operations. Automatic.co would need to win by being more workflow-native, more cross-system, faster to deploy, and less trapped inside one incumbent platform.
Indirect competitors
Competitive Matrix
6. Technology Landscape
The technology stack for agentic AI in real estate, construction, and infrastructure is not one clean product category. It is a bundle of models, data pipes, permissions, workflow rules, system integrations, and trust controls. That is why the winners will not simply be the companies with the “best model.” The winners will be the ones that can connect messy industry data to repeatable decisions, then act safely inside real workflows.
The short version: the intelligence layer is getting better fast, but the hard part is still context. Drawings, RFIs, submittals, lease clauses, jobsite photos, schedules, asset data, and accounting records rarely live in one place. Agentic AI becomes valuable when it can work across those seams without turning into a black box.
Core stack
Google Cloud’s agentic AI architecture guidance is a useful signal for where the market is heading: single-agent systems, multi-agent systems, multimodal classification, GraphRAG-style knowledge orchestration, and agents that interact with disparate enterprise systems are now treated as formal architecture patterns, not research curiosities. That matters for this sector because real estate and construction workflows are naturally cross-system and multimodal. A project issue may involve a photo, a drawing, a spec, an RFI, a schedule line, and an email thread, all at once. (Google Cloud Documentation)
Architecture patterns
- Copilot inside the system of record
This is the most common pattern today. The AI assistant sits inside an existing platform and helps users search, summarize, draft, or answer questions.
Example workflow:
A project manager asks, “What open submittals are blocking level 8 mechanical work?” The copilot searches the construction management platform and returns likely blockers with source links.
Why it matters:
This pattern is easy to adopt because the user stays inside a familiar product. The drawback is that the agent often sees only the data inside that platform.
Market evidence:
Autodesk describes Autodesk Assistant as an agentic AI partner for Design and Make workflows that provides real-time answers, proactive insights, tailored recommendations, and integration with Autodesk products and third-party tools. Autodesk also says the Assistant can route prompts to multiple agents and coordinate complex tasks. (Autodesk) Procore has also moved in this direction with Procore Helix, Procore Assist, and Procore Agent Builder, including AI capabilities for workflow automation, project insights, and answers from specs, RFIs, submittals, and building codes. (Procore)
- Retrieval-first workflow agent
This is the safest near-term pattern for Automatic.co. The agent starts by gathering evidence, cites the sources, and only then recommends or drafts an action.
Example workflow:
A lease obligation agent reads the lease, finds notice periods, renewal options, escalation clauses, and tenant maintenance responsibilities, then creates a task queue with citations.
Why it matters:
Buyers in this industry will not trust agents that cannot show their work. Source-level traceability is the difference between a useful assistant and a risky toy.
Best-fit workflows:
Lease abstraction, compliance evidence, RFI context gathering, submittal review, warranty review, COI tracking, and portfolio reporting.
- Tool-using agent with human approval
This pattern moves from “answering” to “doing,” but with guardrails. The agent can prepare actions, update fields, draft emails, create tasks, or route approvals, but risky steps require a human click.
Example workflow:
A change order agent gathers backup, drafts a narrative, tags cost codes, checks contract language, and sends the package to a project executive for approval before submission.
Why it matters:
This is where ROI starts to show up. The agent is not just saving search time. It is compressing the full workflow.
Best-fit workflows:
Submittal routing, RFI drafting, change event packaging, tenant issue triage, vendor follow-up, owner reporting, and compliance tracking.
- Multi-agent orchestration
In this pattern, multiple specialized agents work together. One agent gathers project data. Another checks contract requirements. Another evaluates schedule impact. Another drafts the owner update. A coordinator agent manages the sequence.
Example workflow:
A schedule-risk agent detects slippage. A document agent finds the related RFIs and submittals. A procurement agent checks long-lead items. A reporting agent drafts the weekly risk update.
Why it matters:
Built-environment workflows are rarely single-step. Multi-agent systems fit the way projects actually break down, but they are harder to govern and test. Google Cloud’s architecture guidance explicitly includes multi-agent systems for complex data analytics, multimodal classification, technical workflows, and enterprise system orchestration. (Google Cloud Documentation)
- Perception-to-action loop
This pattern connects field data to operational action. Site images, 360 captures, videos, drones, sensors, or IoT feeds are converted into progress signals, then agents trigger follow-ups.
Example workflow:
A site capture platform detects that drywall installation is behind plan in one zone. The agent checks the schedule, identifies the downstream trade impact, drafts a recovery note, and routes it to the superintendent.
Market evidence:
OpenSpace’s progress tracking product says it tracks more than 700 components across trades, compares planned versus actual work, and flags schedule risks from site capture data. (OpenSpace) Buildots positions its platform around portfolio-wide visibility, progress reporting, risk management, and objective data for faster construction decisions. (Buildots)
- Optimization agent
This pattern uses AI to generate and compare scenarios. It is especially useful in scheduling, logistics, procurement sequencing, and resource allocation.
Example workflow:
A construction planning agent tests schedule recovery options after a delayed equipment delivery, compares cost and time impact, then recommends the least risky recovery path.
Market evidence:
ALICE Technologies describes its platform as AI construction scheduling software that helps contractors and owners optimize, de-risk, and recover projects through scenario exploration. (Alice Technologies)
Key trends
- Incumbents are embedding agents directly into core platforms
The biggest platforms are not waiting for startups to own the agent layer. Autodesk is positioning Autodesk Assistant as an agentic interface for design and make workflows, while Procore is building AI capabilities into Procore Helix and offering Procore Agent Builder in open beta. (Autodesk, Procore)
- Interoperability is becoming a strategic battleground
Agents need tools. They need APIs. They need clean permissions. They need a way to move across systems without a brittle custom integration every time.
Model Context Protocol, or MCP, is gaining attention because it creates a standard way for AI systems to connect with external data sources, tools, and services. AWS describes MCP as an open standard that gives AI systems a “universal language” for communicating with tools and data sources. (Amazon Web Services, Inc.) Autodesk also says Autodesk Assistant embeds an MCP client that can connect to MCP servers, request data or tools, complete tasks, and interact with product APIs. (Autodesk)
- openBIM and structured project data will matter more
Agentic AI performs better when project data is structured and interoperable. buildingSMART’s openBIM standards are relevant here because they support data sharing across platforms and stakeholders. The openBIM stack includes IFC for standardized descriptions of built assets, IDS for computer-interpretable information requirements, BCF for BIM issue coordination, and bSDD for consistent built-environment definitions. (buildingSMART International)
- The market is moving from document chat to operational memory
Early AI tools answered questions against files. The next layer builds memory around projects, properties, assets, tenants, vendors, risks, and decisions.
That shift is important. A project agent should not just answer, “What does the spec say?” It should know which spec version is current, which RFI modified it, which subcontractor owns the work, whether the material is late, and what schedule activity is affected.
- Multimodal agents are becoming practical
This industry is visual. Drawings, site photos, punch lists, drone imagery, markups, floor plans, maps, and BIM views all carry operational meaning. Pure text agents will miss too much.
OpenSpace and Buildots show why this matters in construction: visual progress data can turn field conditions into measurable status and risk signals. (OpenSpace, Buildots)
- Governance is becoming part of the product, not a compliance afterthought
The more agents act, the more buyers will ask about risk. NIST’s Generative AI Profile, published in July 2024 as a companion to the AI Risk Management Framework, is designed to help organizations incorporate trustworthiness considerations into AI design, development, use, and evaluation. (NIST)
Technology Maturity Curve
7. Use Cases & Industry Applications
The strongest agentic AI opportunities in real estate, construction, and infrastructure are not flashy. They sit in the annoying middle of the work: approvals, evidence gathering, coordination, reporting, exception handling, and follow-up.
That is where the money leaks out.
A superintendent waits for an RFI response. A property manager misses a tenant escalation buried in email. A development team spends three days testing site plans that an AI planning tool can generate in minutes. A project executive reads a polished report that is already out of date by the time it reaches the meeting. These are not small frictions. At scale, they become margin loss, schedule risk, rework, legal exposure, and employee burnout.
The use-case map below focuses on agentic AI, meaning systems that do more than summarize. They retrieve evidence, reason over context, take workflow steps, escalate exceptions, and keep an audit trail.
The market signal is already there. Autodesk’s 2024 construction digital adoption report found that AI was expected to be used by 68% of surveyed construction and engineering businesses once planned technologies were fully implemented. The same report tied each additional technology implemented into operations to a 1.4 percentage point increase in annual revenue growth, equal to $1.4 million in added revenue for a $100 million business. (assets-usa.mkt.dynamics.com)
Deloitte’s real estate research also points to a broad use-case base across property and facilities management, construction management and procurement, legal due diligence, knowledge management, leasing and marketing, valuation, market analytics, architectural design, urban planning, and virtual environments. (Deloitte Brazil)
Horizontal use cases
These use cases cut across owners, contractors, developers, operators, brokers, and infrastructure managers.
Vertical use cases
Use Case ROI Comparison
8. Economics & ROI Modeling
Agentic AI ROI in real estate, construction, and infrastructure does not come from “AI magic.” It comes from plain operational math: fewer hours spent hunting for information, fewer missed handoffs, fewer late approvals, faster reporting, lower rework, better schedule control, and more work handled per employee.
That sounds simple. It is not. This sector has a habit of hiding cost in places that do not show up neatly on a software business case: the project manager chasing a submittal response, the analyst rebuilding a portfolio report for the third time, the property manager triaging tenant issues after hours, the claims team assembling evidence months after the delay happened.
Agentic AI creates value when it attacks those hidden costs directly.
The size of the prize is real. Autodesk and FMI estimated that bad data cost the global construction industry $1.85 trillion in 2020, including $88.69 billion in rework tied to bad data. That is not a small leakage problem. That is an operating model problem. (Autodesk, PR Newswire) McKinsey has also reported that large construction projects often suffer major schedule and cost underperformance, with large projects typically taking 20% longer than scheduled and running up to 80% over budget. (McKinsey & Company)
Cost structure
The economics of an agentic AI deployment have three layers: platform cost, implementation cost, and operating cost. Buyers often obsess over the platform subscription, but implementation and change management decide whether ROI actually shows up.
A useful rule: the farther the product moves from “read and summarize” toward “act and write back,” the more value it creates, but the more governance and integration cost it carries.
ROI drivers
The best ROI drivers are measurable, frequent, and tied to a business owner who already feels the pain.
The important point: ROI should not be modeled only as “hours saved.” That is too small. The bigger value is avoided loss: avoided rework, avoided delay, avoided claims leakage, avoided SLA failure, avoided missed renewal, avoided reporting error.
Metrics
The best metrics vary by workflow, but the core dashboard should be brutally practical.
ROI Waterfall Chart
Revenue per Employee Uplift
9. Adoption Barriers & Risks
Agentic AI has a strong business case in real estate, construction, and infrastructure, but adoption will not be frictionless. This sector is risk-aware for good reasons. A wrong answer can create a reporting error. A wrong action can delay a project, trigger a contract dispute, violate a tenant obligation, expose private data, or send a field team in the wrong direction.
That does not mean adoption will stall. It means the winning products will feel less like “autonomous magic” and more like controlled delegation.
The market will reward agents that are useful, auditable, permission-aware, and humble enough to ask for help.
Core adoption barriers
Why trust is the first adoption barrier
Trust is not soft. It is the control point for the whole market.
In a consumer app, a bad AI answer is annoying. In construction or real estate operations, the same failure can turn into rework, missed claims evidence, a lease compliance issue, or an executive report that hides the wrong risk.
NIST’s Generative AI Profile, released as a companion to the AI Risk Management Framework, highlights the need to manage trustworthiness considerations across AI design, development, use, and evaluation. That framing fits this sector well because the question is not only “does the model work?” It is “can the organization prove how it worked, what it used, and why someone trusted it?” (NIST Publications, NIST)
Minimum trust controls:
Compliance and governance concerns
The sector has several governance layers: project controls, legal, insurance, tenant privacy, public procurement, ESG reporting, financial controls, and cybersecurity. Infrastructure and public-sector work adds even more scrutiny.
Agentic AI touches all of them because agents do not just answer questions. They can trigger tasks, route approvals, update records, draft notices, and eventually write back into systems.
Key concerns:
This is why the first wave of autonomy should be bounded. A lease agent can extract renewal dates and flag obligations. A claims agent can assemble evidence. A schedule-risk agent can draft a warning. But external submissions, contract positions, payment approvals, and tenant-facing actions should start with human approval.
Integration complexity
Integration is where many pilots quietly die.
The demo works because the sample data is clean. The deployment struggles because the real customer environment is a swamp: Procore, Autodesk, Primavera P6, SharePoint, email, Excel, Yardi, MRI, Salesforce, ServiceNow, ERP systems, file naming conventions from 2014, and three versions of the same PDF.
Autodesk and FMI estimated that bad data cost the global construction industry $1.85 trillion in 2020. That estimate is a useful warning: the problem is not just whether an AI model can reason. It is whether the organization can feed it usable, timely, trusted data.
Integration risks:
Change management and human resistance
Construction and real estate teams have seen plenty of software promises. Many still carry scar tissue from tools that added admin burden instead of removing it.
McKinsey has argued that construction firms often fail to capture the full benefit of digital transformation when they stay stuck in isolated pilots rather than scaling change across the enterprise. It also warns that weak frontline change-management experience can make scaling and adoption harder. (McKinsey & Company, McKinsey & Company)
Human resistance usually shows up in four ways:
10. Future Outlook (3–5 Years)
If you zoom out, the industry isn’t just adopting AI. It’s quietly changing how work gets done.
Right now, most teams still live inside software. They click through dashboards, chase updates, rebuild reports, and piece together context from scattered systems. It’s manual, fragmented, and honestly—exhausting.
Over the next 3–5 years, that flips.
Agents become the interface
The biggest shift won’t be better dashboards. It’ll be fewer dashboards.
Instead of logging into five systems to answer a question, the agent does the digging:
- Pulls the latest RFI context
- Checks the schedule impact
- Scans emails and meeting notes
- Drafts a response
- Routes it for approval
The human steps in at the decision point, not the data-gathering phase.
That’s the real shift: from operating software → supervising workflows.
SaaS doesn’t disappear. It just moves into the background. The system of record stays, but the system of action becomes the agent layer.
AI-native organizations start to emerge
Some companies will just “use AI tools.” Others will rebuild how they operate around them.
You’ll see early signals like:
- Project teams spending less time reporting and more time resolving risks
- Property managers handling more tenants without feeling stretched
- Analysts producing outputs in minutes instead of days
- Leadership getting real-time visibility instead of weekly snapshots
The difference isn’t the tool. It’s the workflow design.
In AI-native teams:
- Work flows through agents, not inboxes
- Updates happen continuously, not at reporting cycles
- Knowledge becomes searchable and reusable
- Exceptions get surfaced automatically instead of buried
It feels less reactive. Less chaotic. More controlled.
Multi-agent systems become practical
Most real-world workflows aren’t simple. They cut across roles.
Take something like a schedule delay.
Today, solving that requires:
- Schedule analysis
- Document review
- Procurement checks
- Cost estimation
- Contract interpretation
- Stakeholder reporting
No single system handles that cleanly. No single person owns all of it.
This is where multi-agent systems actually make sense—not as hype, but as coordination.
You end up with:
- A schedule agent spotting the issue
- A document agent pulling supporting evidence
- A contract agent checking obligations
- A reporting agent drafting the update
- A coordinator agent stitching it all together
Humans don’t disappear. They just stop being the glue.
The competitive moat shifts
Right now, a lot of the conversation is about models.
That won’t last.
Models will keep improving, but they’ll also become easier to access. The real advantage moves elsewhere.
First to workflows
Then to data
Then to integrations
And finally to trust
In plain terms:
- Anyone can access a strong model
- Not everyone can design a workflow that actually works in the field
- Even fewer can plug into messy enterprise systems
- Almost no one gets trust right at scale
That last one matters more than people think.
In this industry, if users don’t trust the output, they won’t use it. And if they don’t use it, nothing else matters.
11. Appendix
Definitions
These terms get thrown around a lot, and half the confusion in this space comes from people using them loosely. Here’s what they actually mean in this context.
Agent
A software system that can observe data, reason about it, and take action toward a goal. In this industry, that usually means reading documents, monitoring workflows, drafting outputs, and triggering next steps.
Agentic AI
AI that doesn’t just respond to prompts but operates within workflows. It can decide what to do next, gather information, and execute tasks with some level of autonomy.
Workflow orchestration
The coordination layer that connects steps across systems. It decides what happens next, who needs to approve, and how tasks move forward.
Human-in-the-loop (HITL)
A control model where humans review, approve, or override AI outputs before critical actions are taken. This is not optional in high-stakes workflows.
Autonomy levels
A spectrum:
- Read-only insight (low risk)
- Draft and recommend
- Route and update
- Bounded action
- Fully autonomous execution (rare in this sector)
Most real deployments today sit in the middle, not the extremes.
System of record vs system of action
- System of record: where data is stored (Procore, Yardi, SAP, etc.)
- System of action: where work gets coordinated (increasingly, agents)
Retrieval-augmented generation (RAG)
A method where the AI pulls real documents and data before generating an answer. This is what makes outputs grounded instead of speculative.
Multimodal AI
AI that can process more than just text—images, drawings, PDFs, photos, and sometimes video. Important for construction and field workflows.
Vendor Landscape Map
The market is messy. There’s no clean category yet, but you can group players into a few buckets.
1. System-of-record incumbents
These own the data and the workflows today.
- Procore
- Autodesk Construction Cloud
- Oracle (Primavera, Aconex)
- Yardi
- MRI Software
- SAP, Oracle ERP
They’re adding AI quickly, but their core architecture wasn’t built for agentic workflows.
2. Horizontal AI platforms
These provide the building blocks.
- OpenAI
- Microsoft (Azure AI, Copilot stack)
- Google (Vertex AI, Gemini)
- Anthropic
They don’t solve industry workflows directly, but everything runs on top of them.
3. Workflow and automation platforms
These sit between systems and try to coordinate work.
- ServiceNow
- UiPath
- Zapier
- Retool
They’re moving toward agentic capabilities but aren’t deeply verticalized.
4. Vertical AI startups (emerging layer)
These are closest to the opportunity described in this report.
Examples (real companies working in adjacent spaces):
- OpenSpace (construction site capture + AI insights)
- Buildots (progress tracking using computer vision)
- Document Crunch (construction contract intelligence)
- Lexion (contract workflows)
- VTS (CRE leasing workflows with growing AI layer)
Most are still point solutions. Very few own cross-system workflows yet.
5. Services and consulting layer
Often overlooked, but influential.
- Accenture
- Deloitte
- McKinsey
- Large construction consultants
They’re building custom AI workflows for enterprises, especially when off-the-shelf tools fall short.
Methodology
This report is not based on a single dataset. It’s built from a combination of:
1. Industry reports and benchmarks
- Autodesk + FMI research on construction data and rework
- McKinsey reports on construction productivity and digital adoption
- Public CRE and infrastructure market analyses
2. Market sizing logic
- Top-down estimates from global construction and real estate spend
- Bottom-up modeling based on workflow-level automation value
- Cross-checking against enterprise software and AI adoption trends
3. Workflow-level analysis
Instead of treating “AI” as one category, the report breaks value down by:
- RFI coordination
- Lease abstraction
- reporting workflows
- schedule risk detection
- tenant operations
This is where real ROI shows up.
4. Operator-informed assumptions
The models reflect how work actually happens:
- fragmented systems
- heavy document usage
- manual coordination
- approval-driven processes
Not idealized workflows.
Data Sources
All external claims referenced in this report are grounded in real, published research.
Key sources include:
- Autodesk + FMI
“Harnessing the Data Advantage in Construction”
https://www.autodesk.com/blogs/construction/autodesk-fmi-study-global-construction-industry-data-strategies/ - McKinsey & Company
“Imagining construction’s digital future”
https://www.mckinsey.com/capabilities/operations/our-insights/imagining-constructions-digital-future - McKinsey & Company
“Decoding digital transformation in construction”
https://www.mckinsey.com/industries/capital-projects-and-infrastructure/our-insights/decoding-digital-transformation-in-construction - NIST
Generative AI Risk Management Profile
https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf - Public company filings and product documentation
(Procore, Autodesk, Yardi, MRI, Oracle, Microsoft)
Disclaimer: The information on this page is provided by Automatic.co for general informational purposes only and does not constitute financial, investment, legal, tax, or professional advice, nor an offer or recommendation to buy or sell any security, instrument, or investment strategy. All content, including statistics, commentary, forecasts, and analyses, is generic in nature, may not be accurate, complete, or current, and should not be relied upon without consulting your own financial, legal, and tax advisers. Investing in financial services, fintech ventures, or related instruments involves significant risks—including market, liquidity, regulatory, business, and technology risks—and may result in the loss of principal. Automatic.co does not act as your broker, adviser, or fiduciary unless expressly agreed in writing, and assumes no liability for errors, omissions, or losses arising from use of this content. Any forward-looking statements are inherently uncertain and actual outcomes may differ materially. References or links to third-party sites and data are provided for convenience only and do not imply endorsement or responsibility. Access to this information may be restricted or prohibited in certain jurisdictions, and Automatic.co may modify or remove content at any time without notice.
Timothy Carter
Timothy Carter is a dynamic revenue executive leading growth at LLM.co as Chief Revenue Officer. With over 20 years of experience in technology, marketing and enterprise software sales, Tim brings proven expertise in scaling revenue operations, driving demand, and building high-performing customer-facing teams. At LLM.co, Tim is responsible for all go-to-market strategies, revenue operations, and client success programs. He aligns product positioning with buyer needs, establishes scalable sales processes, and leads cross-functional teams across sales, marketing, and customer experience to accelerate market traction in AI-driven large language model solutions. When he's off duty, Tim enjoys disc golf, running, and spending time with family—often in Hawaii—while fueling his creative energy with Kona coffee.
%203.png)
