AI Goodwill Impairment Testing Under ASC 350: A 2026 Practitioner Walkthrough

If your team runs annual goodwill impairment reviews, you already know the pain: weeks of data gathering, DCF model population, comparables research, and documentation packaging, all compressed into the worst possible window of the year-end close. AI tools are being applied to this workflow in 2026, and some of the applications are genuinely useful. But the accounting literature and the vendor marketing are telling completely different stories, and finance teams are caught in the middle.

This walkthrough maps the ASC 350-20 impairment process step by step, identifies exactly where automation adds defensible value, and explains what your auditors and the SEC will scrutinize when AI-generated outputs show up in your workpapers. It is written for controllers, technical accounting teams, and CFOs at acquisition-active companies who are evaluating whether and how to embed AI tools into their impairment process without creating audit risk.

Key takeaway: AI can meaningfully accelerate the data-intensive, repeatable sub-tasks in goodwill impairment testing. It cannot replace the professional judgment that determines whether an impairment exists, and it introduces new internal control obligations under SOX that most teams have not yet addressed.

How the ASC 350-20 Goodwill Impairment Framework Works in 2026

Under ASC 350-20, goodwill must be tested for impairment at the reporting unit level at least annually. The current framework, simplified by ASU 2017-04, uses a single-step test: if a reporting unit's carrying amount exceeds its fair value, an impairment loss is recognized equal to that excess, capped at the carrying amount of goodwill. The legacy Step 2 hypothetical purchase price allocation is gone.

Before running the quantitative test, companies may perform an optional qualitative assessment (sometimes called Step 0, codified at ASC 350-20-35-3A). If the qualitative assessment concludes it is not more likely than not (probability greater than 50%) that fair value is below carrying amount, no further testing is required. In practice, that relief is narrower than it looks: as PwC's Deals practice notes, many auditors require companies to perform a full quantitative analysis every couple of years regardless of the qualitative conclusion.

One complexity that catches teams off guard: the current single-step model can produce materially different impairment amounts than the old two-step model because it captures fair value declines in non-goodwill assets within the goodwill charge. A reporting unit holding long-lived assets that are recoverable under ASC 360 but whose fair values have fallen below carrying amounts, or a significant portfolio of financial assets carried at amortized cost in a rising-rate environment, can produce a goodwill impairment charge that would not have existed under the prior framework. This interaction is one reason the quantitative test remains judgment-intensive and cannot be fully automated.

For companies with multiple reporting units, the sequencing matters too. Per PwC's guidance, other assets (inventory, financial assets, indefinite-lived intangibles except goodwill, long-lived assets under ASC 360) must be tested for impairment before goodwill. An AI tool that automates goodwill impairment in isolation without integrating this sequencing can produce incorrect results.

Private company note: Companies eligible for the ASC 350-20 private company alternative (ASU 2014-02) may amortize goodwill straight-line over up to 10 years and test only when a triggering event occurs. If that is your situation, the automation ROI calculation is fundamentally different, and most of what follows applies with much less force.

The Full ASC 350-20 Impairment Workflow: Step by Step

Here is the complete workflow, with an honest assessment of where AI helps and where it does not.

Step 1: Reporting Unit Identification

This is the most consequential and least automatable step in the entire process. A reporting unit is an operating segment or one level below an operating segment (a component), provided it constitutes a business and discrete financial information is available and reviewed by segment management. Getting this wrong creates errors that cascade through every subsequent step, and it is one of the areas the SEC's Division of Corporation Finance most frequently questions in comment letters.

AI has limited applicability here. The determination requires legal entity analysis, organizational chart review, management reporting structure assessment, and professional judgment about whether a component meets the definition. AI can help organize the documentation and flag changes in organizational structure from prior periods, but the conclusion is management's.

Step 2: Triggering Event Monitoring (Interim Testing)

This is arguably the highest-ROI automation use case in the entire workflow. Deloitte's DART guidance identifies specific triggers requiring interim testing between annual dates, including:

• Cash or operating losses at the reporting unit
• Consecutive operating results significantly below analyst or internal forecasts
• Significant revisions to internal or external forecasts
• New restructuring plans
• Market capitalization below book value
• Recognition of a goodwill impairment loss in a subsidiary's financial statements
• Sustained decreases in share price (for public companies)

AI monitoring tools can be configured to flag these triggers in near-real-time by pulling from ERP systems, market data feeds, and internal forecasting tools. A company that previously relied on a quarterly manual review of triggering events can replace that process with continuous monitoring, reducing the risk of a missed interim test and the SEC comment letter that follows.

As PwC puts it: "Companies should take a fresh look at existing processes and controls for assessing asset impairment, as proper identification of triggering events is integral to appropriately measuring goodwill impairment."

Step 3: Qualitative Assessment (Step 0)

AI can assist with data gathering; it cannot reach the conclusion. The qualitative assessment under ASC 350-20-35-3A requires evaluation of:

• Macroeconomic conditions (economic deterioration, capital market access, foreign exchange)
• Industry and market considerations (competitive environment, regulatory developments)
• Cost factors (raw materials, labor)
• Overall financial performance (cash flows, revenue vs. prior periods and forecasts)
• Entity-specific events (management changes, strategy shifts, litigation, contemplation of bankruptcy)
• Events affecting a specific reporting unit (composition changes, divestiture expectations)
• For public companies: sustained decreases in share price in absolute terms and relative to peers

AI tools can pull and organize the data underlying each of these factors, compare current-period metrics to prior-period benchmarks, and flag where conditions have deteriorated. Climate-related risk factors, increasingly relevant as a qualitative input, are an area where AI tools trained on ESG data can assist with the data-gathering layer. The weighting of factors and the overall conclusion, however, remain management's judgment call and are subject to auditor challenge.

Step 4: Quantitative Testing (DCF and Market Comparables)

This is where the workflow is most resource-intensive and where the automation boundary is sharpest.

PwC's guidance is direct on the cross-functional coordination this requires: "Early and ongoing cross-functional coordination between accounting, FP&A, valuation and tax professionals is critical to confirming an effective and efficient impairment analysis." AI workflow tools that facilitate data handoffs between these teams, track open items, and maintain version control on model inputs are a legitimate efficiency play. The assumptions themselves are not.

Step 5: Documentation and Disclosure Packaging

AI can draft and organize; management must own and validate every word. ASC 350-20-50-2 requires specific financial statement disclosures when an impairment is recognized, including a description of the facts and circumstances leading to the impairment. AI can compile the supporting documentation package, draft the disclosure narrative, and cross-reference workpaper sections. But the disclosure is a management representation, and any AI-generated text must be reviewed and validated by qualified professionals before it appears in the financial statements or notes.

This is not a theoretical risk. The SEC's Division of Corporation Finance has issued comment letters questioning the adequacy of goodwill impairment disclosures, focusing on: (1) the identification and composition of reporting units; (2) the reasonableness of DCF assumptions including discount rates, terminal growth rates, and revenue projections; and (3) the basis for concluding that a qualitative assessment was sufficient. AI-generated documentation that cannot be traced to specific, auditable data sources is particularly vulnerable to these inquiries. You can search the SEC EDGAR comment letter database for recent examples on goodwill disclosures.

What Auditors Will Scrutinize When AI Is Involved

The PCAOB has consistently identified auditing of goodwill and other intangible assets as a recurring area of audit deficiency. Specific findings have included deficiencies in auditors' testing of management's significant assumptions in DCF models and in the evaluation of the completeness and accuracy of data used in fair value measurements.

When AI tools generate the data inputs or comparables used in a client's impairment model, auditors cannot simply accept those outputs. Under PCAOB AS 2502 (Auditing Fair Value Measurements and Disclosures), auditors must evaluate the completeness and accuracy of the data underlying fair value estimates. That obligation extends to AI-generated outputs. Expect your auditors to ask:

• What AI tool produced this output, and how does it work?
• What data sources did the tool draw from, and how were they validated?
• Who reviewed the AI output before it was incorporated into the model?
• Is there a documented review procedure, and is it operating effectively?
• How does the tool handle edge cases or data anomalies?

The practical implication: AI-generated comparables, sensitivity tables, and documentation packages need the same audit trail as manually prepared workpapers, plus an additional layer documenting the tool's inputs, logic, and the human review step. Presenting AI output without that trail creates audit friction, not efficiency.

SOX and ICFR: The Control Obligation Most Teams Are Missing

This is the gap almost no one in the vendor market is talking about, and it is where the most significant compliance risk sits.

Under SOX Section 302 and 404, management is responsible for the design and operating effectiveness of internal controls over financial reporting (ICFR). When an AI tool is embedded in the goodwill impairment testing process, it becomes part of the ICFR environment. The tool is a control, and it must be treated as one.

That means:

1. Document the control. Describe the AI tool, its function in the impairment workflow, the inputs it receives, and the outputs it produces. This documentation belongs in your ICFR narrative.
2. Assess the design. Does the tool have appropriate input validation? Are there controls over changes to the tool's configuration or underlying model? Is access restricted to authorized users?
3. Test operating effectiveness. Someone must periodically verify that the tool is producing accurate outputs. This is not a one-time implementation check; it is an ongoing testing obligation.
4. Document the human review step. The human-in-the-loop review of AI outputs is itself a control. Document who performs it, what they are reviewing for, and how they evidence their conclusion.

Failure to treat an AI tool as part of ICFR is a potential ICFR deficiency under AS 2201. If your external auditors identify it before you do, it becomes an audit finding. For a more detailed framework on governing AI tools in financial reporting processes, see Finrep's AI governance framework for finance and the AI audit trail requirements for SEC filers.

What a Human-in-the-Loop Governance Model Looks Like

The right governance model is not complicated, but it needs to be explicit and documented.

Before the review cycle:

• Inventory every AI tool used in the impairment workflow and document its function.
• Confirm the tool is included in your ICFR documentation and has been assessed for design effectiveness.
• Assign a named reviewer for each AI-generated output type (comparables, sensitivity tables, documentation drafts).

During the review:

• AI pulls and organizes data; a qualified accountant or valuation professional reviews the output against source data before it enters the model.
• AI runs sensitivity analyses; the finance team reviews the assumption ranges for reasonableness and documents their conclusion.
• AI drafts documentation packages; technical accounting reviews and edits before any text is finalized.
• All management assumptions (discount rate, terminal growth rate, revenue forecasts) are set by humans and documented with supporting rationale independent of any AI output.

After the review:

• Retain the AI tool's input files, output files, and the evidence of human review in the workpaper package.
• Document any instances where the AI output was overridden and why.
• Include the AI tool usage in your ICFR operating effectiveness testing.

For teams also using AI in the broader close process, the AI financial close automation guide covers the ICFR and SOX implications in more depth.

The FASB Goodwill Project: What It Means for Your Automation Investment

Before committing significant resources to building an AI-assisted impairment workflow, finance leaders should be aware of FASB's active project on the subsequent accounting for goodwill, part of its broader Identifiable Intangible Assets and Subsequent Accounting for Goodwill project.

FASB is actively considering whether to require amortization of goodwill for public companies, as is already permitted for private companies under ASU 2014-02. If the Board moves toward mandatory amortization, the annual impairment testing requirement for public companies would change materially, and the ROI of AI automation tools built specifically for the impairment workflow would shift accordingly.

The project is ongoing as of September 2026. Teams building impairment automation should design for modularity: triggering event monitoring and documentation tooling will remain valuable regardless of how the amortization question resolves, while DCF-specific automation is more exposed to a framework change.

FAQ

Can AI determine whether a goodwill impairment exists under ASC 350-20? No. The conclusion requires professional judgment about fair value, which is management's responsibility and a management representation in the financial statements. AI can assist with the data inputs to the analysis, but the conclusion cannot be delegated to a tool.

What is the biggest audit risk when using AI in goodwill impairment testing? Presenting AI-generated outputs without a documented human review step and without treating the AI tool as part of ICFR. Auditors applying PCAOB AS 2502 must evaluate the completeness and accuracy of data underlying fair value estimates, and that obligation extends to AI-generated inputs.

Does using AI for impairment documentation create SEC comment letter risk? Only if the documentation cannot be traced to specific, auditable data sources. The SEC focuses on the substance of disclosures, not the method of preparation. AI-generated text that is reviewed, validated, and supported by a clear audit trail is no more exposed than manually drafted text.

Should private companies invest in AI impairment testing tools? Generally not at the same level as public companies. Private companies electing the ASC 350-20 alternative under ASU 2014-02 amortize goodwill over up to 10 years and test only on triggering events, which significantly reduces the impairment testing burden and the automation ROI.

What happens to impairment automation if FASB requires goodwill amortization for public companies? Triggering event monitoring and documentation tooling remain valuable. DCF-specific automation becomes less central. Design your workflow for modularity rather than building a monolithic impairment automation stack.

How should we document AI tool usage in our impairment workpapers? Include: the tool name and version, the data sources it drew from, the specific outputs it produced, the name and title of the person who reviewed those outputs, the date of review, and any adjustments made to the AI output before it was incorporated into the analysis. This documentation satisfies both PCAOB AS 2502 and AS 2201 requirements.

The goodwill impairment review will not become a push-button process anytime soon. But teams that map the workflow honestly, automate the right sub-tasks, and govern AI tools as the ICFR controls they are will run materially faster, more defensible reviews than those still doing everything manually.