Cost Segregation in the Age of AI

An Analysis of Three Studies Against the IRS Audit Technique Guidelines

Over the last six weeks at Engineered Tax Services, three prospective clients have presented our firm with cost segregation studies they generated themselves using generative artificial intelligence — two with ChatGPT, one with Google Gemini — and asked our licensed engineering team to “validate” the work. In each case, we declined the engagement. In each case, the prospect was annoyed. And in each case, the practitioner-level questions raised by the request reach considerably further than the immediate sales conversation. They go to what constitutes a defensible cost segregation study under the IRS Audit Technique Guidelines (ATG); the taxpayer's exposure under IRC §6662; the preparer's exposure under IRC §6694; the AICPA Statements on Standards for Tax Services; the duty of due diligence under Circular 230; and the rules of professional engineering conduct that prevent a reputable engineering firm from attaching its name to work it did not perform.

This article documents our analysis of three real studies — the two AI-generated outputs we received from prospects, and an engineering-based study our team prepared on a comparable property — against the thirteen elements the IRS describes as marking a “quality cost segregation study.” It then examines the practitioner implications for taxpayers, CPAs, and the engineering firms that are increasingly being asked to put their licensed name onto AI-generated tax work.

The verdict the framework produces is unsurprising in direction but more nuanced than the typical “AI can or cannot do this work” framing suggests. The substantive issue is not AI per se. It is professional accountability for a tax position.

I. The Practitioner Issue

The use of generative AI to prepare tax-relevant work product is moving rapidly from novelty to routine fact pattern. Cost segregation is fertile ground for the experiment. The ATG does not mandate any particular methodology. The visual structure of an engineering report — categorized buckets, percentages, totals — is something AI tools reproduce competently at the surface level. The combination of “AI can do this” and “the IRS does not require a specific approach” produces an inference that the work has been accomplished. It often has not.

Our firm is seeing three distinct fact patterns emerge in 2026.

First, the taxpayer who has generated an AI cost segregation “study” and intends to rely on it directly when filing a return.

Second, the taxpayer who approaches an engineering firm asking us to “validate” the AI work — meaning, in practice, to attach our licensed engineering firm's name to the output so the taxpayer can defend the position on examination.

Third — and most relevant to practitioners reading this in AccountingToday — the CPA whose client brings them an AI-generated study and asks the CPA to file the return relying on the AI-derived classifications.

Each fact pattern raises distinct exposures. This article addresses all three.

II. The IRS Audit Technique Guidelines Framework

The IRS Audit Technique Guide for cost segregation studies, originally published by the Service in 2004 and updated through subsequent revisions, is the practitioner's framework for evaluating any cost segregation work product. The ATG is explicit that the IRS itself does not prescribe methodology:

“Neither the Internal Revenue Service nor any group or association of practitioners has established any requirements or standards for the preparation of cost segregation studies. The lack of specific requirements has resulted in a wide variety of approaches and methodologies.”

But the ATG identifies six methodologies, ranked roughly by reliability, and observes:

“While neither the Service nor any group or association of practitioners prescribes a specific methodology, there are certain approaches that produce more accurate and reliable allocations.”

The two highest-ranked methodologies under the ATG are the Detailed Engineering Approach from Actual Cost Records and the Detailed Engineering Cost Estimate Approach. Both presume engineering analysis performed by qualified individuals. The ATG further identifies thirteen “principal elements of a quality cost segregation study” — the framework an IRS examiner is trained to apply when a cost segregation position is challenged. These elements are not optional; they are the substantive criteria used at examination. They are reproduced in the table below alongside our assessment of how each of the three studies in this analysis performs against them.

Three structural points about the thirteen elements deserve practitioner attention before turning to the comparison.

Element 1 (qualified preparer) is binary. There is either a named individual whose expertise can be attested to, or there is not. AI does not satisfy this element regardless of how sophisticated the output appears.
Elements 8, 10, and 12 (unit costs/take-offs, reconciliation, §1245 identification) are analytical work products. The substantive question is not whether source documents exist — it is whether a qualified individual performed the analytical work on those documents.

Element 7 (legal analysis) cannot be satisfied by a recitation of generic tax language. It requires citation to controlling authority. The leading authority for cost segregation analysis remains Hospital Corp. of America & Subsidiaries v. Commissioner, 109 T.C. 21 (1997), which established the modern framework for distinguishing §1245 personal property from §1250 real property in the depreciation context. An AI tool can mention the case; the ATG element asks whether the case's analytical framework is being applied.

III. Analysis of Three Studies

Our analysis compares three real studies. The two AI-generated outputs were supplied to us by prospective clients; the engineering-based study is one our firm prepared in 2023, redacted to remove client-identifying details, used here with the client's permission as an illustrative comparator.

A. The Gemini summary

The first study was generated by Google Gemini at the request of an individual property owner with a $400,000 residential investment property. The complete output consists of six rows in a single-tab Excel worksheet. The rows identify $48,000 of “appliances, flooring, fixtures” as 5-year property; $12,000 of “misc equipment” as 7-year property; $40,000 of “land improvements” as 15-year property; and $300,000 as 27.5-year residential rental property, with a $100,000 subtotal of bonus-eligible basis.

There is no methodology statement. No preparer identification. No documentation reference. No analytical narrative. No citation to authority. The categories are heuristic estimates. To be fair to Gemini: the AI tool did not represent its output as a cost segregation study; the user asked for a summary and received a summary. The practitioner issue arises when the taxpayer treats the output as documentation supporting a return position.

B. The ChatGPT report

The second study presents the more interesting analytical case. The client is the developer of a $4.7 million new-construction athletic facility. The client uploaded a complete construction cost breakdown from the general contractor — including itemized invoices for each trade — and asked ChatGPT to produce a cost segregation report. The output is an Excel workbook with seven tabs.

The main “Trade Allocation Matrix” tab presents 38 construction trades. Each is assigned a reclassification percentage to 5-year and 15-year asset categories. The totals reconcile arithmetically: $390,915 reclassified to 5-year + $846,705 to 15-year = $1,237,621, approximately 26.1% of basis. A second tab contains a ∼225-row sub-component breakdown of the concrete trade. Five additional tabs — labeled site utilities, land grading, mechanical, plumbing, and electric — contain the contractor's actual invoices for each trade, embedded as nine PNG image files totaling approximately two megabytes.

The structural deficiencies relative to the ATG framework are several and worth itemizing for practitioners reviewing similar work product.

No methodology is stated. The report does not name which of the six ATG methodologies was applied. The implicit approach — estimating reclassification percentages by trade without engineering take-offs or component identification — most resembles the ATG's “Sampling or Modeling Approach” or “Rule of Thumb Approach,” the two least reliable methodologies in the ranking.
No engineer of record is identified. No P.E. license number. No firm credentials. No certification page. No statement of independence or absence of contingent fee.
No site inspection was performed. The output is generated from the cost data and invoice images the client uploaded. The ATG names the on-site engineering approach as the most reliable methodology.
No engineering take-offs were performed. The contractor invoices are present in the trade tabs but no unit costs were extracted from them and applied to identified components. ATG element 8 requires the analytical work, not merely the document custody.
No reconciliation was performed. The trade allocation matrix does not reference the invoices in the trade tabs. There is no analytic trace between the invoiced amounts and the reclassification percentages. ATG element 10 calls for reconciliation as an analytical work product, not arithmetic summation.
No indirect cost analysis is performed. The report lists $85,070 in permit fees and $98,816 in architectural fees but assigns 0% reclassification to both. The ATG element 11 requirement is that indirect costs be allocated to appropriate assets under §263A capitalization principles. Defaulting them to 39-year property is not allocation.
No §1245 vs. §1250 statutory analysis is performed. MACRS bucket assignment does not satisfy element 12, which requires the personal-property-versus-real-property statutory test applied to each identified component.
No interviews were conducted with the property owner, the architect, the general contractor, or any facility manager. ATG element 4.
No legal analysis or citation to controlling authority appears anywhere in the file. ATG element 7.
Substantial trade-level amounts are assigned 0% reclassification. The tilt-wall erector trade ($479,779), plumbing ($237,421), roofing ($200,789), painting ($118,092), insulation ($102,977), and combined construction management ($439,672) are all reclassified at 0% to short-life assets. A competent engineering analysis would identify components inside several of these trades that qualify for reclassification — specialty plumbing fixtures, decorative finishes inside painting work, specific architectural elements within the management scope. The aggregate amount left at 0% reclassification is approximately $1.59 million. This is not merely a technical deficiency under the ATG; it is also a real opportunity left on the table for the taxpayer.

C. The ETS engineering-based study

The comparator is a cost segregation study our firm prepared in 2023 for a high-rise apartment building with a $43,865,936 basis. The property was placed in service November 30, 2022. The report runs 139 pages.

For purposes of the ATG framework, the substantive contents of the report are:

Project engineer Robert Schultz is named on the report as the qualified individual who performed the analysis. He specializes in cost segregation studies and prepares them in accordance with the IRS ATG and the U.S. Master Depreciation Guide.
CEO Julio P. Gonzalez signs the certification page. ETS certifies that the firm has no contingent interest in the conclusions, no contingent fee arrangement, and no predetermined results.
ETS holds a Florida Professional Engineering license (#CA27824) and a Texas Professional Engineering license (#F-12634). Both license numbers are verifiable on the public state engineering board registries.
The methodology is named explicitly: the Detailed Engineering Approach from Actual Cost Records, combined with the Detailed Engineering Cost Estimate Approach — two of the six ATG methodologies, applied together, both at the top of the reliability ranking.
A physical site inspection was performed by the named project engineer. Photographic documentation, observations, and on-site notes are referenced in the certification.
Documentation reviewed includes architectural plans, drawings, AIA cost documents, contractor invoices, and the existing fixed asset depreciation schedule. All reconciled to actual cost records.
Interviews were conducted with facility management.
Legal analysis section cites Hospital Corp. of America & Subsidiaries v. Comm'r, 109 T.C. 21 (1997); IRC §263A capitalization of interest and production costs; the Tax Reform Act of 1986 (P.L. 99-514); and additional rulings collected in a 20-page appendix of statutory and case-law references.
The component-level breakdown runs to 28 pages — every reclassified component individually identified with its own cost basis and depreciation life.
A dedicated indirect cost allocation section (page 54) treats architectural fees, engineering fees, and permits under §263A principles.
1245 personal property is identified separately from §1250 real property — the actual statutory test.
Unit of Property analysis, bonus depreciation analysis, disposition and retirement analysis, and mis-capitalized fixed asset analysis are each addressed in dedicated sections.
Biographies and credentials section documents the engineering team's experience and specialization.

The numerical result on this specific property was an increase in accumulated depreciation through 2022 from $165,494 (without the study) to $9,635,547 (with the study), a difference of $9,470,053 in accelerated depreciation available to the taxpayer.

D. Scoring against the thirteen ATG elements

The table below presents the assessment of each study against each ATG quality element. We have assigned partial credit only where the element is materially satisfied even if not fully developed.

#	ATG quality element	Gemini	ChatGPT	ETS	Practitioner observation
1	Preparation by individual with expertise and experience	Fail	Fail	Pass	AI tools do not satisfy the qualified-preparer requirement. There is no named individual whose expertise can be attested to or contested.
2	Detailed description of methodology	Fail	Fail	Pass	Neither AI study names which of the six ATG methodologies was applied. The ETS report names two: Detailed Engineering from Actual Cost Records + Detailed Engineering Cost Estimate.
3	Use of appropriate documentation	Fail	Pass*	Pass	ChatGPT received contractor invoices from the client and embedded them in trade tabs. Document possession is satisfied. The work product built on those documents is not — see elements 8 and 10.
4	Interviews with appropriate parties	Fail	Fail	Pass	AI cannot conduct facility-manager or contractor interviews.
5	Common nomenclature	Fail	Partial	Pass	Gemini uses lay terms (“appliances, flooring, fixtures”). ChatGPT uses some standard MACRS categories.
6	Standard numbering system	Fail	Partial	Pass	Trade-level numbering only in ChatGPT; no component-level identification scheme.
7	Explanation of legal analysis	Fail	Fail	Pass	ETS cites Hospital Corp. of America v. Comm'r, 109 T.C. 21 (1997), IRC §263A, and the Tax Reform Act of 1986. Neither AI study cites controlling authority.
8	Unit costs and engineering take-offs	Fail	Fail	Pass	Take-offs require quantitative extraction of unit costs from source records, applied to identified components, performed by a qualified individual. AI tools did not perform this analytic step.
9	Asset organization	Fail	Partial	Pass	ChatGPT consolidates 38 trades; ETS report contains a multi-page component-level inventory.
10	Reconciliation to actual cost records	Fail	Fail	Pass	Mathematical summation is not reconciliation. Reconciliation requires tracing the reclassified basis to substantiated cost records with discrepancies identified and explained.
11	Treatment of indirect costs	Fail	Fail	Pass	ChatGPT defaults architectural and permit fees to 39-year classification rather than performing allocation analysis under §263A.
12	§1245 property identification	Fail	Fail	Pass	§1245 vs. §1250 is a statutory analysis applied to each identified component. MACRS bucket assignment does not satisfy this element.
13	Related aspects (§263A, accounting method, sampling)	Fail	Fail	Pass	ETS report addresses each. AI studies address none.
	TOTAL (of 13)	0	1	13	*ChatGPT satisfies element 3 because client-supplied invoices are present in tabs. Elements 8 and 10 nonetheless fail because no analytical work was performed on those documents.

The aggregate score is ETS 13 of 13; ChatGPT 1 of 13; Gemini 0 of 13.

The ChatGPT study satisfies element 3 because the client supplied the contractor invoices and they are present in the file. It satisfies no other element in full. The Gemini summary does not attempt to meet the framework.

IV. The Critical Distinction — Document Possession Versus Engineering Analysis

One nuance in the ChatGPT result merits separate treatment because it explains why “the AI had the source documents” does not produce the practitioner protection it might appear to. The same nuance is also what makes the validation request — “will you just sign off on the AI study?” — impossible to fulfill, as discussed below.

ATG element 3 is satisfied at the level of document possession. The taxpayer's contractor invoices were present in the workbook. A practitioner-level reviewer looking only at element 3 would mark it satisfied, as we did.

ATG elements 8, 10, and 12 require analytical work. They are not satisfied by the existence of the source documents. They are satisfied by the engineering analysis performed on the source documents by a qualified individual. Specifically:

Element 8 (unit costs and engineering take-offs) requires that quantities and unit costs be extracted from source records and applied to identified components. The ChatGPT output does not contain take-offs. It contains trade-level percentage assignments that do not reference the underlying invoiced amounts.
Element 10 (reconciliation) requires the analytical trace between the reclassified basis and the actual cost records, with discrepancies identified and explained. Arithmetic summation — demonstrating that the trade matrix subtotals to the total basis — is not reconciliation in the ATG sense. Reconciliation requires that each reclassified component traces back to a substantiated cost record.
Element 12 (§1245 identification) requires statutory analysis applied to each identified component. The personal-property-versus-real-property test under §1245 is performed under Hospital Corp. of America & its progeny and turns on the function and purpose of the asset, the manner of its installation, and whether it serves the building or the business operating inside. MACRS bucket assignment performed at the trade level does not perform this test.

The implication for practitioners reviewing AI-generated cost segregation work product is that the presence of source documents in the file is not, by itself, evidence that the elements requiring analytical work have been satisfied. A reviewer must look for the analytical work product — not merely the underlying records. Cost segregation studies prepared by qualified engineering firms include this analytical work in narrative form alongside the schedules; AI-generated outputs typically do not.

V. Taxpayer Exposure Under IRC §6662

The accuracy-related penalty under IRC §6662 applies to underpayments attributable to (among other things) a substantial understatement of income tax. §6662(d). The base penalty is 20% of the underpayment, increasing to 40% under §6662(h) for gross valuation misstatements or undisclosed transactions lacking economic substance. The penalty does not apply to portions of the underpayment for which the taxpayer establishes reasonable cause and good faith. §6664(c)(1).

Treas. Reg. §1.6664-4 sets the standard for the reasonable cause defense. The leading case authority is Neonatology Associates, P.A. v. Commissioner, 115 T.C. 43 (2000), affd. 299 F.3d 221 (3d Cir. 2002), which established the three-prong test for the “reliance on professional advice” form of the reasonable cause defense:

The adviser was a competent professional who had sufficient expertise to justify reliance.
The taxpayer provided necessary and accurate information to the adviser.
The taxpayer actually relied in good faith on the adviser's judgment.

An AI tool does not satisfy prong one. An AI output has no “professional” within the meaning of the Neonatology framework, no licensure or credentialing whose expertise can be tested, and no Circular 230 obligations that would support a reasonable cause defense. The taxpayer who files a return relying on an AI-generated cost segregation study cannot point to a competent professional whose expertise justified the reliance — there is no professional in the equation. The Neonatology defense fails at the first prong.

The taxpayer can attempt to argue that the AI tool is more like a calculator or a research database than a professional advisor — that is, a tool the taxpayer used to inform a return position the taxpayer themselves took in good faith. The argument runs into the difficulty that the AI output here is not a calculation or a research query but a substantive determination of the taxpayer's tax position. Cost segregation classifications are substantive tax determinations under the ATG and the Hospital Corp. framework, not arithmetic. An AI tool that generates the classification is functionally claiming the role of the qualified individual that the ATG element 1 requires — a role no AI tool can occupy under any current authority.

The result is that a taxpayer relying on an AI cost segregation study faces the §6662 base penalty if the position is reduced or disallowed, with limited prospect for a reasonable cause defense. On a $4.7 million property where the reclassification analysis would otherwise support seven figures of accelerated depreciation, the exposure is meaningful both in absolute and percentage terms.

VI. Preparer Exposure Under IRC §6694

Section 6694 imposes a penalty on a tax return preparer who prepares a return reflecting an understatement of liability attributable to an undisclosed position that lacks substantial authority, or a disclosed position that lacks a reasonable basis. §6694(a). The penalty is the greater of $1,000 or 50% of the fee derived. For positions involving willful conduct or reckless disregard, the penalty under §6694(b) is the greater of $5,000 or 75% of the fee derived. Substantial authority is interpreted in light of the standards in Treas. Reg. §1.6662-4(d).

AICPA Statement on Standards for Tax Services No. 1 sets the practitioner standard at a position that has a realistic possibility of being sustained on its merits — a standard the AICPA has historically interpreted as broadly consistent with the substantial authority threshold for undisclosed positions. The CPA who prepares a return reflecting cost segregation classifications drawn from an AI-generated study must, under SSTS No. 1, conclude that the position meets this standard.

Circular 230 §10.22 imposes a parallel due diligence obligation. The practitioner must exercise due diligence as to the accuracy of representations the practitioner makes to the IRS — and a return reflecting cost segregation classifications is a representation about the depreciable lives of identified property.

The practitioner's exposure when a client brings an AI-generated cost segregation study has several layers.

If the AI study satisfies one of the thirteen ATG quality elements and fails the other twelve, as in our ChatGPT example, it is difficult to articulate the substantial authority for the return position. The cost segregation classifications would be drawn from a work product that does not satisfy the IRS quality framework for cost segregation studies.
Disclosure (Form 8275 or 8275-R) shifts the standard from substantial authority to reasonable basis but does not eliminate practitioner exposure. The CPA who discloses a position based on an AI-generated cost segregation study may still face §6694(a) exposure if the position lacks even a reasonable basis.
The CPA's professional liability carrier may have specific positions on AI-derived work product in tax positions. Several carriers in 2025 and 2026 have updated their underwriting guidance to address generative AI use in return preparation. CPAs should consult their carrier on the firm's exposure before signing returns relying on AI cost segregation studies.

The practical implication for the CPA receiving an AI-generated study from a client is that the appropriate response is rarely “we will file it.” The appropriate response is more often: “We will not file the return reflecting these classifications without an engineering-based study that satisfies the ATG framework. We can refer you to qualified engineering firms.”

VII. The Validation Problem

Engineering firms are now routinely asked to “validate” AI-generated cost segregation studies — meaning, in practice, to put the engineering firm's licensed name on the AI's output so the taxpayer can defend the position on examination. The request is structurally impossible to fulfill.

There are two paths to a hypothetical “validation,” and both fail.

Path one is for the engineering firm to perform the substantive engineering analysis the AI did not perform — to extract unit costs from the source documents, identify components, perform the reconciliation, apply the §1245 statutory analysis, document the methodology, conduct interviews, and issue a certification page. At which point the “validation” is the study. The AI output adds nothing to the work product. The taxpayer has paid for the AI tool and then paid the engineering firm for the actual analysis. There is no meaningful sense in which the engineering firm has “validated” the AI output; it has supplanted it.

Path two is for the engineering firm to attach its licensed name to the AI's output without performing the substantive analysis — to issue a certification page over the AI's classifications. This path is foreclosed by professional engineering conduct rules in every state with engineering licensure, including Florida and Texas where our firm holds its licenses. The Florida Board of Professional Engineers' rules on signing and sealing work require that an engineer take responsibility only for engineering work the engineer either performed or directly supervised. See Fla. Admin. Code R. 61G15-23. The Texas Engineering Practice Act contains parallel provisions at Tex. Occ. Code §1001.402. Putting a licensed engineer's seal onto AI-generated cost segregation work product the engineer did not perform constitutes a violation of these rules and exposes the firm and the individual engineer to disciplinary action by the state board, in addition to whatever liability the firm assumes by signing work it did not perform. No reputable engineering firm signs unaudited AI output. The regulatory boundary explains why.

This is the practical answer to the prospective client who asks whether we can simply review and validate. We cannot. The structure of the engineering credentialing system, the regulatory rules governing engineering signatures, and the ATG quality framework all point to the same conclusion: AI cost segregation work is not work product that can be validated by a licensed engineering firm short of redoing it. “Redoing it” is, in fact, what we offer. Many of these conversations end with the engagement letter for an engineering-based study that meets the ATG framework from the beginning.

VIII. Practitioner Framework — Recommendations

For CPAs receiving cost segregation studies (or purported studies) from clients in 2026, our firm's recommendations are as follows.

Document what was received. If the client has provided AI-generated work product, the engagement file should reflect that. Note the tool used, the date generated, the input data provided, and any caveats the AI tool itself provided.
Score the received work product against the thirteen ATG elements. The framework is operational, not theoretical. If the work product fails substantial elements — in particular elements 1, 7, 8, 10, and 12 — the position likely cannot be supported on substantial authority grounds.
Decline to take the position on the return where the work product does not satisfy the framework. Recommend the client commission an engineering-based study from a qualified firm. The increased cost is typically modest relative to the deduction at stake and the penalty exposure averted.
Where the client is unwilling to commission a qualifying study, consider disclosure on Form 8275 and document the basis for the position. Confirm exposure with the firm's professional liability carrier.
Update engagement letter language. Several firms in 2025 added provisions specifically addressing AI-generated work product the client may provide for use in return preparation. The provisions document who bears the risk if AI-derived classifications are disallowed.

For new engagements, include cost segregation methodology questions in the client intake. Whether the client has previously used AI tools to prepare tax work product is now a relevant practitioner inquiry.

IX. The Right Frame for AI in Tax Practice

This article should not be read as anti-AI. Our firm uses generative AI substantially in our work — for research support, for drafting, for analytical review of documents, for client communications. AI is an excellent tool, and its role in tax practice will continue to expand.

The line that the cost segregation analysis above illuminates is professional accountability for a tax position. AI tools cannot be the named preparer of a tax position. They cannot be subject to Circular 230 due diligence requirements. They cannot stand behind a position under examination. They cannot defend a study against an examiner's challenge. They cannot be subject to professional discipline. They cannot carry professional liability coverage. They are an instrument; they are not a professional.

The framing that protects both taxpayers and practitioners is this: AI is a tool that licensed professionals use — not a substitute for licensed professionals. A cost segregation study prepared by a qualified engineering firm may legitimately involve AI tools in document review, in drafting, in analytical support. What it cannot involve is the AI tool occupying the role of the engineer of record. The engineering work — the take-offs, the reconciliation, the §1245 analysis, the site inspection findings — must be performed by qualified individuals whose names appear on the report and who can answer to examination if it is challenged.

This framing protects the taxpayer who relies on a study prepared by a licensed engineering firm that uses AI tools competently. It protects the CPA who files a return relying on such a study. It protects the engineering firm whose credentialed personnel performed the analytical work. It does not protect the taxpayer or the practitioner who treats an AI output as if it were itself the engineering work.

Conclusion

Three studies. One framework. The IRS examiner's question on audit will not be whether the taxpayer used AI to prepare the cost segregation position. It will be the same five document requests that have opened cost segregation examinations for the last decade: the engineering report with classifications and reasoning; timestamped photographic evidence; the engineer of record's credentials; reconciliation to actual cost records; documentation of the methodology applied. Whether AI was used in preparing those work products is largely irrelevant. Whether the work products exist, and whether a qualified preparer's name stands behind them, is dispositive.

For practitioners advising clients on cost segregation positions in 2026, the conversation must begin and end there.

About the author. Heidi Henderson is Chief Marketing Officer of Engineered Tax Services, a licensed engineering firm specializing in cost segregation studies, R&D tax credits, and engineered tax solutions. ETS holds Professional Engineering licenses in Florida (#CA27824) and Texas (#F-12634) and has prepared over 75,000 cost segregation studies since 2001. She can be reached at hhenderson@engineeredtaxservices.com.