Artificial intelligence is rapidly changing the way construction projects are priced, designed, coordinated, and built. Contractors can use AI-assisted estimating platforms to analyze drawings, identify quantities, predict labor requirements, compare subcontractor proposals, and generate preliminary budgets. Architects and engineers can use generative design tools to evaluate layouts, test building performance, identify clashes, and accelerate the production of digital models. Owners can use predictive systems to assess project feasibility, monitor costs, and forecast potential delays.
These tools may improve speed and efficiency, but they also create a difficult legal question: who is responsible when the machine-assisted result is wrong?
An AI-generated estimate may omit an entire category of work. A digital takeoff may misread the scale of a drawing. A predictive model may rely on historical prices that are no longer realistic. A building information model may contain incorrect dimensions, incompatible systems, or structural assumptions that were never reviewed by a licensed professional. A contractor may rely on the model when ordering materials, only to discover that the components do not fit in the field. An owner may approve a project based on an unrealistically low budget and later face millions of dollars in change orders.
Although the technology may be new, the starting point for liability remains familiar. Courts generally do not treat “the AI” as the legally responsible actor. Instead, liability will usually be assigned among the owner, architect, engineer, contractor, construction manager, subcontractor, consultant, model author, software provider, or data vendor according to their contracts, professional duties, warranties, representations, and conduct.
The decisive question is therefore not simply whether AI produced the error. The more important questions are who selected the tool, who supplied the data, who was expected to verify the output, how the output was described, who was permitted to rely on it, and whether the governing contracts clearly allocated the resulting risk.
Artificial intelligence systems do not independently enter construction contracts, hold professional licenses, stamp drawings, submit bids, or direct construction work. A person or company ordinarily makes each of those decisions. When an AI-assisted estimate or model fails, the legal analysis will focus on the conduct of those human and organizational actors.
The National Institute of Standards and Technology’s Artificial Intelligence Risk Management Framework emphasizes that AI risk should be addressed through governance, mapping, measurement, and management. It also stresses accountability, transparency, reliability, explainability, documentation, and human oversight. ¹ These principles are especially important in construction because an inaccurate output can cause more than an administrative inconvenience. It can result in an underbid, a delayed project, defective work, physical damage, or a safety hazard.
The American Institute of Architects has similarly advised architecture and design firms that AI should be used responsibly, transparently, and with continued attention to professional judgment and the protection of public health, safety, and welfare. ² The architect’s professional role does not disappear because a machine generated part of the analysis. An architect or engineer who incorporates AI-assisted work into professional services may still be expected to understand the output sufficiently to evaluate whether it is appropriate for the project.
This does not mean that every AI error automatically establishes negligence. Architects and engineers are generally not guarantors of perfect results, and contractors do not necessarily warrant that every estimate will match the final cost. The legal standard ordinarily asks whether the responsible party performed its contractual obligations and exercised the level of care required under the circumstances. Nevertheless, using an automated tool without reasonable validation may become evidence that the required standard was not met.
Construction disputes involving AI will ordinarily begin with the project contracts rather than with abstract principles of technology law. The owner-contractor agreement, owner-architect agreement, consultant contracts, subcontracts, software licenses, BIM exhibit, and project execution plan may each contain provisions affecting responsibility.
The contracts may identify who is responsible for preparing cost estimates, who must review design information, which documents control in the event of a conflict, and whether a digital model is a contract document. They may also contain limitations of liability, warranty disclaimers, indemnity obligations, consequential-damage waivers, notice requirements, insurance provisions, and restrictions on reliance.
A contract stating that a preliminary estimate is provided only for planning purposes may significantly limit an owner’s ability to treat it as a guaranteed construction price. Conversely, a contractor that expressly guarantees a maximum price may assume greater responsibility for estimating errors, subject to allowances, contingencies, exclusions, and permitted adjustments. An architect who promises only an opinion of probable construction cost is ordinarily in a different position from a consultant retained specifically to produce a detailed quantity takeoff and validated cost estimate.
The same distinction applies to digital models. A model used only for visualization may not be intended for fabrication, quantity takeoff, or field layout. A model identified as sufficiently developed for fabrication may invite a substantially greater level of reliance. Liability can therefore turn on whether the disputed use of the model was authorized and reasonably foreseeable.
Modern BIM agreements recognize the importance of documenting these distinctions. The AIA’s 2022 BIM document system separates major model-sharing and reliance terms from the more detailed BIM execution plan and model element table. Those documents allow project participants to identify model authors, permitted uses, development levels, update procedures, software requirements, data-security measures, and the degree to which particular model versions may be relied upon. ³
ConsensusDocs 301 similarly provides a contractual framework for addressing legal and administrative issues involving BIM, including a BIM execution plan and provisions concerning the participants who contribute information to the model.⁴ These documents illustrate an important principle: reliance should be negotiated before an error occurs, not reconstructed after a dispute begins.
A bad construction estimate can arise in several different ways. The AI system may fail to identify elements shown on the drawings. The drawings themselves may be incomplete. The estimator may upload the wrong version. The system may apply outdated unit prices or labor productivity assumptions. A human user may accept a suggested quantity without reviewing the underlying measurement. The output may also be technically accurate but presented without adequate contingencies, escalation, design-development allowances, or location-specific adjustments.
Determining liability requires separating the source of the error from the party that assumed the estimating risk.
When a contractor uses its own AI estimating software to prepare a lump-sum bid, the contractor will ordinarily face the initial risk of an internal estimating mistake. A contractor is generally expected to evaluate the bid documents, calculate its costs, obtain appropriate subcontractor input, and submit the price it is willing to accept. The fact that an employee relied on an automated takeoff program does not normally shift the resulting loss to the owner.
The contractor may have a claim against the software provider if the program malfunctioned or failed to perform as represented. That claim, however, may be constrained by the software license. Technology agreements frequently disclaim implied warranties, exclude consequential damages, limit remedies to repair or replacement, and cap liability at the amount of fees paid.
The danger of those limitations is demonstrated by M.A. Mortenson Co. v. Timberline Software Corp. In that case, a contractor alleged that defective estimating software caused its construction bid to be approximately $1.95 million lower than intended. The Washington Supreme Court enforced the software license provisions that limited the vendor’s liability.⁶ Although the case did not involve contemporary generative AI, it remains highly instructive. A construction company may suffer an enormous project loss while possessing only a modest contractual remedy against the software vendor whose product contributed to the error.
The lesson is that a contractor cannot assume that the technology vendor will reimburse the full consequences of an underbid. Before adopting an AI estimating platform, the contractor should understand the warranty, limitation-of-liability, data-retention, update, and dispute-resolution provisions in the vendor agreement. The contractor should also determine whether the vendor maintains appropriate technology errors-and-omissions coverage and whether the contract permits meaningful recovery for a system failure.
The analysis changes when the bad estimate results from inaccurate information supplied by the owner or design professional. If the owner issues defective plans or specifications and requires the contractor to price and construct the work in accordance with them, the contractor may have a claim for additional compensation. The longstanding Spearin doctrine recognizes that an owner that furnishes plans and specifications generally impliedly warrants their adequacy for the intended construction. A contractor that follows those requirements may not be responsible for consequences caused solely by defects in the owner-provided design.⁵
AI does not necessarily change that principle. If an owner provides a digital model as a contract document and directs bidders to use its quantities, an error embedded in the model may be treated similarly to an error in conventional plans or specifications. The owner may not be able to avoid responsibility merely by explaining that an algorithm generated the model.
The outcome will depend heavily on the contract. An owner may state that model quantities are informational only and that bidders must perform independent takeoffs. A contractor may agree that the drawings and specifications control over the model. The bid instructions may require bidders to report discrepancies before submitting a price. If the contractor ignores an obvious conflict or uses the model for a purpose the contract expressly prohibits, the contractor’s recovery may be reduced or barred.
Architects, engineers, and construction managers frequently prepare opinions of probable cost during planning and design. These opinions help the owner determine whether the proposed project is financially feasible. They are often based on incomplete drawings and must account for changing material costs, market conditions, project phasing, labor availability, and unknown site conditions.
An estimate exceeding the final cost does not automatically establish liability. Most professional agreements do not guarantee that contractor bids or actual construction costs will match an early budget. The professional may instead promise to exercise reasonable skill and judgment in preparing an opinion based on the information available at the time.
AI may nevertheless affect how that standard is applied. A professional who uses an AI system to prepare an estimate should know the source and limitations of the data. A model trained on national averages may not accurately reflect labor rates, union requirements, soil conditions, material availability, permitting practices, or seasonal constraints in southeastern Michigan. A tool using historical costs may fail to recognize rapid price escalation. A generative system may also present an incomplete assumption as though it were a verified fact.
Professional judgment is therefore required before the output is communicated to the client. The reviewer should determine whether the estimate reflects the actual project scope, design phase, geographic market, procurement method, schedule, escalation period, contingency, and level of uncertainty. An estimate presented as highly reliable when it has not been validated may create a stronger misrepresentation or negligence argument than one clearly identified as a preliminary planning range.
The contract should also explain what happens if the project exceeds the owner’s budget. Some agreements require the architect to modify the design without additional compensation if bids exceed the budget by an agreed percentage, while others allow additional compensation where the excess results from market changes, scope expansion, owner decisions, or circumstances beyond the architect’s control. AI-generated estimates should be integrated into this existing contractual framework rather than treated as an entirely separate service.
Digital-model disputes are often more complicated than estimating disputes because many project participants may contribute information to the same model. The architect may develop the architectural model. Structural and mechanical engineers may contribute discipline-specific models. The contractor may create coordination and sequencing models. Subcontractors and fabricators may add shop-level details. A model manager may combine the separate components into a federated model.
When a defect appears, identifying the party that “created the model” may therefore be insufficient. The more precise question is who authored, controlled, approved, or modified the particular model element that caused the loss.
Suppose an HVAC subcontractor fabricates ductwork based on a coordinated model, but the ductwork conflicts with structural steel in the field. Liability may depend on whether the structural element was modeled at the required level of development, whether the HVAC subcontractor was authorized to rely on its exact dimensions, whether a later structural revision was properly incorporated, and whether the clash-detection process should have identified the conflict.
A model’s level of development is particularly important. A low-development conceptual element may show general size and location without representing fabrication-level accuracy. A more developed element may be intended to support quantity takeoff, coordination, fabrication, installation, or facilities management. Treating every visible object in a model as equally reliable creates substantial risk.
The contract should identify the model author for each important element, the required development level at each project milestone, and the authorized uses of that element. It should also establish who is responsible for coordinating changes, publishing updated versions, archiving superseded models, and notifying participants of revisions.
The AIA BIM framework is designed to document precisely these issues. Its 2022 documents allow participants to distinguish models shared throughout the project from models restricted to the design or construction team. They also allow the parties to decide whether particular model versions may be enumerated as contract documents. ³ The legal significance of a model may therefore differ substantially from one project to another.
If the model is expressly incorporated into the contract documents, a contractor’s reasonable reliance may be stronger. If the model is provided solely for convenience and the contract states that two-dimensional drawings control, reliance on conflicting model geometry may be more difficult to justify. The governing language must be examined carefully rather than assuming that BIM has the same contractual status on every project.
An architect or engineer may face professional-negligence liability when a model defect results from a failure to exercise the ordinary learning, judgment, and skill expected of a similarly situated professional. Michigan’s model civil jury instruction describes professional negligence as failing to do what a professional of ordinary learning, judgment, and skill would do under the same or similar circumstances or doing something such a professional would not do.⁹
The use of AI does not necessarily raise the professional standard to perfection. It may, however, affect what constitutes reasonable practice. If responsible design firms routinely validate AI-generated structural calculations, independently review code compliance, and maintain model version controls, a firm that performs none of those safeguards may have difficulty demonstrating reasonable care.
The professional should also avoid delegating licensed judgment to an unlicensed tool or employee. Michigan law requires licensed architects or professional engineers to prepare plans, specifications, and estimates for covered public works, subject to statutory exceptions. ¹⁰ An AI platform may assist in preparing those materials, but it is not itself a licensed professional capable of assuming statutory responsibility.
Professional liability may also arise from coordination failures. Even if each discipline-specific model is accurate in isolation, the architect, engineer, or BIM coordinator may have undertaken responsibility for integrating the models. If the contract assigns that coordination function, a failure to identify inconsistent geometry, missing information, or incompatible revisions may support a breach-of-contract or professional-negligence claim.
At the same time, the architect should not be held responsible for work outside the agreed scope. A design model may not include temporary bracing, fabrication tolerances, erection sequences, or contractor means and methods unless the professional specifically agreed to provide those services. The model’s apparent detail should not silently expand the professional’s contractual obligations.
Contractors and construction managers may face liability when they use AI to develop takeoffs, schedules, logistics plans, coordination models, or construction means and methods. Their responsibility will depend on the project-delivery method and the services they agreed to perform.
A construction manager providing preconstruction services may be expected to review constructability, identify scope gaps, reconcile estimates, and evaluate cost trends. If it accepts an AI-generated estimate without reviewing major assumptions, it may breach those obligations even though it did not create the software.
A design-builder may face broader exposure because design and construction responsibility are consolidated. The owner may not need to determine whether the defect originated with the design consultant, estimator, or construction team before asserting a claim against the design-builder. The design-builder may then pursue indemnity or contractual remedies against the responsible subcontractor, consultant, or vendor.
Contractors also remain responsible for construction means, methods, techniques, sequences, and procedures to the extent assigned by the contract. A contractor cannot necessarily avoid responsibility for an unsafe or unworkable sequence by stating that an AI scheduling platform recommended it. The contractor’s management team must evaluate whether the recommendation is appropriate for actual site conditions.
The contractor may nevertheless have a valid claim when it reasonably relies on defective owner-provided design information. The Spearin principle remains relevant when the contractor is required to follow the owner’s plans, specifications, or incorporated digital model.⁵ The key distinction is between an error in the owner’s design requirements and an error in the contractor’s independent means, methods, estimating, or coordination work.
Trade contractors increasingly prepare highly detailed models used for coordination and fabrication. Their contracts should define whether they are merely converting design intent into a fabrication model or whether they are performing delegated design.
A subcontractor that accepts delegated design responsibility may be required to retain a licensed engineer, satisfy specified performance criteria, and coordinate the resulting design with other systems. If its AI-assisted model fails to meet those requirements, the subcontractor and its design consultant may face liability.
A subcontractor that is not responsible for design may still be liable for failing to follow the model, disregarding field conditions, or fabricating from an outdated version. Conversely, the subcontractor may have a claim if it used the current authorized model for its intended purpose and the information supplied by another participant was defective.
Model-sharing protocols should extend through every contractual tier. It is not enough for the owner and architect to agree on BIM rules if the subcontractors and fabricators actually using the model never receive or accept those rules. Both the AIA and Consensus Docs approaches contemplate aligning project participants around model responsibilities, reliance, and execution procedures. ³ ⁴
Software vendors may be responsible when a defect in the product causes incorrect calculations, corrupted data, lost model information, or inaccurate output. Potential claims may include breach of express warranty, breach of implied warranty, negligent misrepresentation, product liability, or breach of contract.
In commercial transactions, however, the vendor agreement may sharply restrict those claims. The license may state that the software is provided “as is,” that outputs must be independently verified, that the vendor does not warrant accuracy, and that the customer assumes responsibility for decisions made using the product. It may also exclude lost profits, delay damages, replacement costs, and other consequential losses.
Michigan’s economic-loss doctrine may further limit tort recovery where the claim arises from the commercial sale of a defective product, and the alleged losses are purely economic. In Neibarger v. Universal Cooperatives, Inc., the Michigan Supreme Court held that the economic-loss doctrine can confine a commercial purchaser to remedies available under the Uniform Commercial Code when defective goods cause economic losses.⁷ Whether a particular AI subscription or software arrangement is treated as a sale of goods, a license, a service, or a hybrid transaction will depend on its substance and the governing law.
The Mortenson decision demonstrates the practical consequence. Even when allegedly defective estimating software contributes to a multimillion-dollar underbid, enforceable contractual limitations may prevent recovery of the contractor’s full loss.⁶ Construction companies should therefore negotiate important technology contracts with the same care applied to major subcontracts.
Vendor responsibility may be stronger when the vendor provides implementation, custom training, project-specific data conversion, or consulting services. A vendor that merely licenses a general platform may have less project-specific responsibility than one that promises to configure and validate a system for a particular estimating or modeling workflow.
Owners may contribute to an AI-related failure by requiring use of a particular platform, supplying incomplete data, compressing the design schedule, restricting professional review, or treating a preliminary model as suitable for construction.
An owner that demands an aggressive early estimate but refuses to fund adequate investigation should not assume that the estimate carries the reliability of a completed design. Similarly, an owner that directs contractors to rely on a centralized digital twin may assume obligations concerning access, accuracy, cybersecurity, maintenance, and version control.
The owner should clearly identify the purpose of the AI-enabled deliverable. A conceptual model used to select among design alternatives should not automatically become the basis for procurement or fabrication. Each transition to a higher-consequence use should require confirmation that the information has reached an appropriate level of development and has been reviewed by the responsible professionals.
Owners should also avoid creating conflicting instructions. A contract cannot safely state that the model is merely informational while the project team is simultaneously directed to use it for quantity takeoff, coordination, prefabrication, and field layout. Courts and arbitrators will examine actual project conduct as well as formal disclaimers.
A defective model may injure or economically harm parties that have no direct contract with the model author. A subcontractor may rely on an engineer’s model even though the engineer contracted only with the architect. A later purchaser may discover a defect allegedly caused by digital design information. A worker or neighboring property owner may suffer physical injury or property damage.
Michigan law distinguishes between contractual duties and independently existing duties. In Loweke v. Ann Arbor Ceiling & Partition Co., the Michigan Supreme Court explained that entering a contract does not extinguish separately existing common-law or statutory duties owed to noncontracting third parties.⁸ A party may therefore face tort liability to a noncontracting person where an independent duty exists, particularly when the alleged conduct causes physical injury or damage to other property.
Purely economic claims by noncontracting parties are more difficult. Courts are often reluctant to impose unlimited liability for financial losses suffered by everyone who may foreseeably rely on project information. The analysis may involve negligent-misrepresentation principles, intended third-party-beneficiary provisions, the economic-loss doctrine, and express limitations on model reliance.
This is another reason BIM agreements should specify who is entitled to rely on each model and for what purpose. The broader the permitted group and use, the greater the potential liability exposure of the model author.
Even after identifying an error, the claimant must ordinarily prove that the error caused the claimed loss. AI-related construction disputes may involve multiple contributing causes.
An omitted quantity may have been generated by the software, but the estimator may also have failed to perform the required review. A model may contain a dimensional error, but the subcontractor may have fabricated from an obsolete version. A clash may have existed in the design model, but the contractor may have failed to run the agreed coordination process. Material costs may exceed the estimate because of market escalation rather than an algorithmic defect.
The parties will therefore need evidence showing how the estimate or model was created, modified, reviewed, approved, distributed, and used. Relevant evidence may include source drawings, data inputs, prompts, assumptions, system settings, training or reference datasets, software versions, model versions, clash reports, validation records, approval histories, audit logs, emails, meeting minutes, and testimony from the personnel responsible for review.
A party that cannot reproduce the output or identify the version used may face substantial difficulty proving or defending the claim. AI governance is therefore also evidence preservation. The project should retain enough information to reconstruct material decisions without necessarily preserving every inconsequential interaction.
Traditional construction insurance policies were not drafted with every AI-related risk in mind. Coverage may depend on whether the claim alleges professional services, defective work, bodily injury, property damage, economic loss, cyber failure, or technology error.
A commercial general liability policy may respond to covered bodily injury or property damage but exclude damage to the insured’s own work or purely economic losses. Professional liability coverage may respond to negligent design or professional services but contain exclusions relating to warranties, guarantees, cost estimates, or technology failures. Cyber coverage may address data breaches and system interruption without covering the cost of replacing defective physical construction. A technology errors-and-omissions policy may protect the software vendor but include contractual-liability and consequential-damage limitations.
Owners, designers, contractors, and technology vendors should review their policies before a project begins. The contracts should not require a party to accept AI-related liability that is materially broader than its available insurance.
The strongest protection is a contract that addresses the actual technology being used rather than relying on generic language written for paper drawings.
The agreement should define the AI-enabled systems and identify the functions for which they may be used. It should distinguish administrative uses, such as summarizing meeting notes, from higher-risk uses involving cost, design, code compliance, safety, fabrication, or field layout.
The contract should allocate responsibility for data quality. The output of an AI system may be unreliable because the underlying information is incomplete, outdated, improperly formatted, or taken from the wrong project. Each participant should know which data it must supply, whether it warrants the data’s accuracy, and who must identify conflicts.
The agreement should also require appropriate human review. The level of review should increase with the potential consequences. A conceptual rendering may require less validation than a structural calculation or fabrication model. The contract should identify who performs the review, what qualifications that person must possess, and how approval is documented.
Model provisions should establish a clear hierarchy among drawings, specifications, schedules, addenda, models, and later revisions. They should identify whether any model version is a contract document and explain how conflicts are resolved. The parties should not wait until litigation to determine whether the model or the drawing controlled.
Version control is equally important. The project should have a designated location for current files, a naming convention, publication procedures, access controls, and an archive of superseded versions. Participants should receive prompt notice of material changes and should be prohibited from using obsolete models.
The contract should define authorized uses and prohibited uses. A model authorized for coordination should not be presumed suitable for quantity takeoff or fabrication. An estimate intended for early budgeting should not be treated as a guaranteed maximum price. Clear limitations are most effective when they reflect actual project practices.
The parties should address the technology vendor’s role and contractual limitations. If the project depends on a critical platform, the contracting party should investigate whether the vendor warrants availability, accuracy, interoperability, data portability, and technical support. It should also understand the vendor’s liability cap and whether project data may be used to train external systems.
Indemnity and limitation-of-liability provisions should be tailored to the responsibilities each participant can control. A model author should not indemnify the entire team for every misuse of the model. A recipient should not bear sole responsibility where the model was expressly represented as suitable for the disputed use. Broad exclusions of consequential damages should be evaluated carefully because many estimating and modeling losses involve delay, lost productivity, rework, escalation, and lost profits.
Finally, the contract should require preservation of material records. Where AI contributes to a major estimate, design decision, or model revision, the parties should retain sufficient documentation to establish the data used, output produced, review performed, and approval given.
There is no single rule making the owner, contractor, architect, or software company automatically liable whenever AI produces an incorrect result. Responsibility will usually follow control, contractual assumption, professional duty, reasonable reliance, and causation.
A contractor that privately uses AI to prepare its lump-sum bid will generally remain responsible for the bid unless the error resulted from defective owner-provided information or another recognized basis for relief. An architect or engineer that incorporates AI-generated content into professional deliverables may remain responsible for exercising the applicable professional standard of care. A subcontractor that creates a fabrication model may be responsible for its own model elements and delegated design. An owner that requires reliance on an inaccurate contractual model may face responsibility under contract principles comparable to those governing defective plans and specifications. A software vendor may bear responsibility for a defective product or service, but its exposure may be heavily restricted by the license agreement.
In many disputes, responsibility will be shared. The software may generate an error, the professional may fail to identify it, the contractor may ignore a discrepancy, and the owner may demand reliance on an incomplete deliverable. Comparative-fault principles, contractual indemnity, and contribution claims may then determine how the loss is distributed.
Artificial intelligence can improve construction estimating, design, coordination, and project management, but it does not eliminate the traditional need to define responsibility. The most dangerous arrangement is one in which every participant uses the digital output, but no participant clearly owns the duty to validate it.
Owners should identify which AI-assisted deliverables may be relied upon. Architects and engineers should preserve professional judgment and review. Contractors should independently validate estimates and understand the limits of their software licenses. Subcontractors should confirm model versions and authorized uses before fabrication. Technology vendors should be evaluated not only for performance but also for contractual accountability and insurance.
When those responsibilities are documented, AI becomes another manageable project tool. When they are left undefined, a single incorrect quantity, assumption, or model element can produce a dispute involving nearly every participant in the project.
The central legal question will rarely be whether the AI made a mistake. It will be whether the responsible people and companies reasonably selected, governed, reviewed, communicated, and relied upon the technology.
This article is provided for general informational purposes and does not constitute legal advice. The application of construction, contract, professional-liability, and insurance law depends on the governing jurisdiction and the particular project agreements and facts.
Contact Tishkoff
Tishkoff PLC specializes in business law and litigation. For inquiries, contact us at www.tish.law/contact/. & check out Tishkoff PLC’s Website (www.Tish.Law/), eBooks (www.Tish.Law/e-books), Blogs (www.Tish.Law/blog) and References (www.Tish.Law/resources).
Sources:
1- National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework (AI RMF 1.0), NIST AI 100-1, January 2023. https://nvlpubs.nist.gov/nistpubs/ai/nist.ai.100-1.pdf
2- American Institute of Architects, Guidance for the Responsible Use of AI by Architecture and Design Firms, updated October 22, 2025. https://www.aia.org/aia-architect/article/architects-and-ai-practical-guidance-changing-profession
3- AIA Contract Documents, Introducing AIA Contract Documents’ 2022 BIM Documents, discussing E201-2022, E202-2022, E401-2022, E402-2022, G203-2022, G204-2022, and G205-2022. https://learn.aiacontracts.com/articles/6523765-introducing-aia-contract-documents-2022-bim-documents/
4- ConsensusDocs, ConsensusDocs 301—Building Information Modeling Addendum. https://www.consensusdocs.org/contract/301-building-information-modeling-bim-addendum/
5- United States v. Spearin, 248 U.S. 132; 39 S Ct 59; 63 L Ed 166 (1918). https://supreme.justia.com/cases/federal/us/248/132/
6- M.A. Mortenson Co. v. Timberline Software Corp., 140 Wash 2d 568; 998 P2d 305 (2000). https://law.justia.com/cases/washington/supreme-court/2000/67796-4-1.html
7- Neibarger v. Universal Cooperatives, Inc., 439 Mich 512; 486 NW2d 612 (1992). https://law.justia.com/cases/michigan/supreme-court/1992/88206-3.html
8- Loweke v. Ann Arbor Ceiling & Partition Co., LLC, 489 Mich 157; 809 NW2d 553 (2011). https://law.justia.com/cases/michigan/supreme-court/2011/20110606-s141168-55-loweke-final.html
9- Michigan Model Civil Jury Instruction 30.01, Professional Negligence/Malpractice, amended October 2023. https://www.courts.michigan.gov/4abd23/siteassets/rules-instructions-administrative-orders/jury-instructions/civil/adopted/m-civ-ji-30.01-professional-negligence-malpractice-as-adopted-oct-2023.pdf
10- Michigan Occupational Code, MCL 339.2011, concerning plans, specifications, and estimates for covered public works. https://michigancontractorauthority.com/michigan-engineer-licensing-law/