AI Is Cutting Construction Costs 17-20%. That Changes Everything About Development Feasibility.
Robots walked through a 31-story apartment tower in Seattle.
Shot 360-degree photos of every floor. Flagged water leaks, exposed beams, fire extinguishers in wrong positions. Checked images against blueprints.
The inspection that would have taken weeks of site walks took days. Skanska saved 40 hours per week.
That's AI on a construction site today.
But that's just the inspection layer. What happens when AI runs the entire build?
Suffolk and MIT built a model and applied it to a real project: 180K SF San Francisco apartment tower finished in 2024.
Estimated savings if AI had run the whole build:
17-20% construction cost reduction
22-25% schedule acceleration
5-6 percentage points added to unlevered IRR
1-2 points added to yield on cost
Suffolk calls that difference between projects getting built and projects staying on the drawing board.
But here's what property investors need to understand: Cost cuts that make development viable also add competing supply. The same efficiency that makes your deal pencil makes everyone else's deal pencil too.
THE PROBLEM THAT AI SOLVES (75% OF PROJECTS FAIL THEIR OWN BUDGETS)
Three-quarters of construction projects run late or over budget.
Labor inefficiency wastes $30B-$40B annually across the industry.
That's not a rounding error. That's the difference between a deal working and not working.
The typical cost overrun:A $50M multifamily project planned for 18 months and $50M budget runs 22 months and $57M actual cost.
7% schedule overrun. 14% budget overrun.
Where does it go?
Labor inefficiency. Rework. Permitting delays. Scheduling conflicts. Supply chain gaps. Design changes caught mid-build.
The economic impact:
A $50M project with 14% cost overrun loses ~$2M in margin. That $2M spread across pro forma equity eats returns.
On a 20% equity raise ($10M), that's 20% of equity returns vaporized by cost overruns.
That's why developers stop building when cost pressures rise.
AI addresses this structural problem.
WHERE SUFFOLK AND MIT FOUND THE MONEY (SEVEN SPECIFIC COST CATEGORIES)
The white paper identified six places where AI adds value:
1. Design Automation
AI checks designs against building codes before they go to permitting. Catches errors before expensive rework.
Example: A designer places a window 6 inches from an electrical panel. AI flags it instantly. Manual review catches it in construction (cost: $15K rework).
2. Offsite Manufacturing
AI optimizes prefabrication batching. Instead of building custom walls one at a time, AI identifies patterns and manufactures in batches.
Reduces on-site labor. Compresses schedule. Improves quality.
3. Permitting Acceleration
The killer: Permitting runs through 20,000+ independent jurisdictions. Each has unique codes and timelines.
AI reads jurisdiction codes, auto-generates permit applications, flags missing documentation before submission.
Result: Permitting timeline compressed 6-8 weeks on average projects.
4. Schedule Optimization
AI models task dependencies and resource constraints. Identifies critical path inefficiencies humans miss.
Example: Two crews waiting for the same equipment (cost: 5 days of idle labor). AI reschedules task sequence to eliminate wait.
5. Skilled Labor Allocation
AI forecasts labor needs by trade weeks in advance. Prevents overstaffing and understaffing.
Also identifies tasks that can shift to apprentices (lower cost) vs. journeyworkers (higher skill requirement).
6. Subcontracting Efficiency
AI analyzes subcontractor performance across your portfolio. Flags underperformers before contracts renew.
Also models cost breakdowns across subs and identifies where to consolidate vs. split work.
7. Supply Chain Management
AI forecasts material delivery timelines. Identifies potential shortages months ahead.
Also models alternative suppliers and long-lead-time strategies (order early, negotiate discounts).
THE IRR UPLIFT (5-6 POINTS IS GAME-CHANGING)
Let's model this against a real deal:
Base case (no AI):
Acquisition: $50M
Hard costs: $35M (estimated, typically runs over to $40M)
Soft costs: $5M
Equity: $15M
Projected unlevered IRR: 12%
AI case (17-20% construction cost savings):
Acquisition: $50M
Hard costs: $28-29M (17-20% savings on $35M estimate)
Soft costs: $4.5M (some savings in design/permitting)
Equity: $15M (same capital required)
Total project cost: $82-83.5M vs. $90M base case
Unlevered IRR: 17-18% (5-6 points uplift)
That difference is massive. Projects that don't pencil at 12% IRR absolutely pencil at 17%.
The yield on cost impact:
Base case NOI $5.5M / $50M cost basis = 11% yield on cost.
AI case NOI $5.5M / $42-43M cost basis = 12.8-13.1% yield on cost (1-2 points).
Lower basis. Same NOI. Higher yield. Better underwriting. Lenders happier.
WHO'S ALREADY USING IT (AND HOW TO GET ACCESS)
Large developers building their own AI systems (Lennar, Toll Brothers, Meritage).
But smaller operators and GCs don't need to build systems. They're getting it through existing software:
Autodesk. AI embedded in design and BIM software. Checks designs against codes automatically.
Procore. AI in scheduling and budget tracking. Forecasts cost overruns weeks ahead.
Both platforms already in use by contractors. AI layers on top of existing workflows.
Smaller contractors are already adopting. Not all bells and whistles, but dipping toes in the water.
Why? Because preconstruction tools (evaluating opportunities, analyzing bids, speeding up estimating) are where AI delivers immediate value.
Those same preconstruction processes set renovation budgets. Worth asking your GC what they're running before you sign the next construction contract.
THE DOUBLE-EDGED SWORD (COST CUTS = MORE COMPETING SUPPLY)
This is the part developers and investors need to think through.
A 17-20% cost reduction makes marginal projects viable.
A project that didn't pencil at $50M hard costs now pencils at $42M hard costs.
From a development company perspective: Great. More deals flow through underwriting.
From an existing operator perspective: Problem. That marginal project is now a competing supply pipeline.
The supply headwind:
In a market with weak fundamentals (oversupply, flat rents, occupancy pressure), cost-cutting AI doesn't help developers. It doesn't fix the market.
But cost-cutting AI does mean: Projects that sat on the shelf become viable. More units hit market. Occupancy pressure increases. Rent growth stays suppressed longer.
The timing dynamic:
Markets with tight supply and strong rent growth: AI cost cuts = developer upside. More building helps everyone.
Markets with weak supply absorption (San Marcos, Huntsville, Austin secondary submarkets): AI cost cuts = headwind. More cheaper supply hits already-soft market.
Operators holding existing product when this happens face extended absorption periods and pricing pressure.
WHAT YOU CAN ACT ON TODAY (BEFORE THE NEXT BID)
The 17-20% savings figure is a first-pass model of one completed project. Nobody has banked those savings in real time yet.
But what operators can actually implement:
Step 1: Ask Your GC the Right Questions
Before signing the next construction contract:
What design automation software are you using?
Do you run cost forecasting on all projects?
What's your typical schedule variance across portfolio?
Have you implemented offsite prefabrication? At what scale?
Their answers tell you whether they're capturing cost efficiencies or not.
Step 2: Require Cost Transparency
Demand itemized budget breakdowns by cost category (design, permitting, labor, materials, subs).
Ask GC to identify where AI could improve outcomes (permitting delays, rework patterns, labor inefficiency).
Good GCs have this analysis. Mediocre ones don't.
Step 3: Baseline Your Own Data
Track your own cost overruns across projects. Identify patterns (permitting delays, change orders, labor productivity).
Use that data to evaluate whether new GC relationships actually improve outcomes or just promise to.
Step 4: Model Competing Supply
When analyzing acquisition opportunities, ask: How much cheaper could a new build be with current AI cost-cutting tech?
If your market is already supplied with new units, that cheaper competing supply matters.
THE HONEST ASSESSMENT (RESULTS DEPEND ON IMPLEMENTATION)
Construction technology consultant Erin Khan (former Suffolk executive) said it clearly: "Results depend on how teams implement the tools and share data."
MIT's research is directional, not validated. The industry needs stronger evidence base.
Translation: 17-20% cost cuts are possible. But they're not automatic.
Implementation matters. Contractor competency matters. Data sharing across subs matters.
A mediocre GC using Procore won't suddenly become efficient. A smart GC using Procore + rigorous processes gets the efficiency.
What's realistic:
Permitting timeline savings: 6-8 weeks (validated across projects)
Design rework reduction: 10-15% (depends on how early AI catches errors)
Schedule variance reduction: 8-12% (depends on complexity)
Overall cost reduction: 10-15% realistic (17-20% is best-case scenario)
10-15% is still 3-4 points of IRR uplift. That changes deal economics.
But don't assume every GC is capturing it.
THE STRATEGIC IMPLICATIONS (THREE SCENARIOS)
Scenario 1: You're a developer
Cost-cutting AI is now a competitive requirement. GCs using it can underbid GCs not using it. This is already happening.
Start requiring Procore + Autodesk + active cost forecasting from GCs. Otherwise you're leaving 10-15% on the table.
Scenario 2: You're an operator buying existing product
Cost cuts that make new development viable are your supply headwind.
In soft markets (oversupply, weak absorption), factor AI-powered competing supply into underwriting.
In tight markets (constrained supply, rent growth), new building via AI is actually good—it fills gap you can't fill.
Scenario 3: You're considering a value-add deal
Ask the architect/GC what AI tools they're using for renovation budget and schedule.
Renovation budgets are more reliable than new construction budgets. But they're also where AI can tighten scope (lighter upgrades, phased approaches).
Same preconstruction principles apply.
THE BOTTOM LINE
AI construction cost cuts are real. 17-20% is aspirational but 10-15% is realistic.
That's 3-4 points of IRR. That's 0.5-1.5 points of yield on cost. That's projects that didn't pencil now penciling.
But cost cuts don't fix fundamentals. They make marginal projects viable. In oversupplied markets, that's another unit competing for your tenants.
What operators can do today: Ask their GC what tools they're using. Demand itemized cost breakdowns. Track your own overruns. Model competing supply based on realistic cost-reduction scenarios.
The contractors and operators moving first on this are already getting efficiency. By 2027, it becomes baseline expectation.
LIVE Q&A TRAINING WITH JUSTIN THIS WEEK! 6PM PST
🙏🏼 Thanks for reading!
Join our Facebook Group here!
Click here to join our WhatsApp Community.
Here's how I can help:
Book a strategy call with Justin and his team to get "eureka moment" clarity about where you're at and where you want to go with real estate investing and plan.
Get investing tools and learning by starting with The Multifamily Schooled Courses.
—Justin Brennan
Comments