In-Kind Gift Valuation
A donor drops off forty laptops, a law firm bills you zero for eighty hours, and a restaurant caters your gala. Someone now has to put a number on all of it, for your financial statements, your Form 990, and your donor acknowledgment letters. That job has always been slow, inconsistent, and quietly risky. AI can do most of the research behind it in minutes, which is genuinely useful and also the fastest way to produce a number you cannot defend.
In-kind valuation is one of those tasks that sits awkwardly between departments. Development knows what the donor said it was worth. Finance knows what the auditors will accept. Program knows what the organization would have paid if it had bought the thing. These are three different numbers, and in many organizations nobody has ever written down which one goes where. The result is a set of records that are internally inconsistent, hard to explain, and reconstructed under pressure every year when the audit starts.
The volume makes it worse. A food bank receiving daily donated product, a legal aid organization tracking pro bono hours, a school receiving donated equipment, and a disaster relief agency receiving mixed pallets of goods all face the same problem at different scales, which is that valuation is a research task repeated hundreds or thousands of times a year by people who have other jobs. When research is expensive and repetitive, organizations take shortcuts. The most common shortcut is accepting whatever number the donor supplies, which is precisely the practice that produced the sector's worst valuation scandals.
This is a natural fit for AI, and the fit is real. Finding comparable prices, checking what a service normally bills at, categorizing a shipment, and drafting the documentation that supports a valuation are all research and writing tasks that language models handle well. What they do not handle is judgment about which market applies, whether a restriction changes the answer, and whether a number is defensible to someone who is skeptical of it. That distinction is the whole subject of this article.
Below we cover the two separate valuation obligations most nonprofits confuse, what fair market value actually means when applied to donated goods, the history that explains why auditors are suspicious of gifts-in-kind, exactly where AI earns its place in the workflow, and the controls that keep an AI-assisted process audit-ready. If your challenge is the logistics of receiving and tracking these gifts rather than pricing them, our guide to in-kind donation management covers that side.
You Have Two Valuation Jobs, and Only One of Them Is Yours
The most consequential misunderstanding in this area is that the nonprofit determines the value of a donor's tax deduction. It does not. The donor is responsible for substantiating their own deduction, and for noncash gifts above certain thresholds that means obtaining a qualified appraisal. Your organization's role on IRS Form 8283 is limited to the donee acknowledgment section, where you confirm that you received the described property on a given date. Signing that section is not an endorsement of the donor's claimed value, and staff should understand that clearly before they sign anything.
The thresholds are worth knowing even though they are the donor's problem. Section A of Form 8283 covers claimed deductions under five thousand dollars per item or group of similar items, along with publicly traded securities. Section B covers claimed deductions of five thousand dollars or more, and those generally require a qualified appraisal performed by someone with demonstrable expertise in that property type, prepared no earlier than sixty days before the date of the gift. Separately, under Internal Revenue Code section 170(f)(8), a donor cannot deduct any single contribution of two hundred fifty dollars or more without a contemporaneous written acknowledgment from you that describes the property and states whether you provided anything in return.
One follow-on obligation does land squarely on your organization. If you dispose of charitable deduction property that was reported on Form 8283 Section B within three years of receiving it, you generally have to file Form 8282 and provide a copy to the donor. Organizations that receive donated vehicles, equipment, or art and later sell them frequently miss this, and it is exactly the sort of thing that surfaces in an audit rather than at the time.
The valuation that genuinely is yours is the accounting one. Under generally accepted accounting principles, contributed nonfinancial assets are recognized as revenue at fair value on the date received, and that number flows into your statement of activities, your functional expense allocation, and your Form 990. This is the number your auditor will test, and it is entirely independent of whatever the donor writes on their own return. Keeping these two separate in your records, with separate fields and separate documentation, eliminates most of the confusion in one move.
The donor's number
Their deduction, their responsibility
Substantiated by the donor, appraised at their expense above the threshold, reported on their return. Your obligations are the written acknowledgment describing the property and, where applicable, the donee signature on Form 8283.
Never state a dollar value for the property in an acknowledgment letter. Describe what you received and let the donor and their advisor handle the number.
Your number
Financial statements and Form 990
Fair value at the date of receipt, determined under your own accounting policy, disclosed with the valuation technique used. This is what the auditor tests and what feeds your revenue, expense, and efficiency ratios.
It can legitimately differ from the donor's claimed deduction. Documenting why is much easier than explaining later that you never noticed the two were different.
Fair Market Value Is a Question About Markets, Not About Price Tags
Fair market value is conventionally defined as the price at which property would change hands between a willing buyer and a willing seller, neither under compulsion, both having reasonable knowledge of the relevant facts. Every word in that definition is doing work, and the word that causes the most trouble for nonprofits is the implied one: which market. A donated item does not have a single price. It has a price in a retail store, a different one in a wholesale channel, another on a resale platform, and another again in a different country.
Accounting guidance resolves this by pointing to the principal market for the asset, meaning the market with the greatest volume and level of activity for that item, considered from the perspective of the entity holding it. The practical translation for a nonprofit is that you should value the thing in the market where it would actually be sold or where you would actually buy it, not in the market with the highest number. A pallet of near-expiry canned goods is not worth its shelf price at a grocery store, because that is not the market in which that particular pallet would trade.
Contributed services carry an additional wrinkle that catches many organizations by surprise. Not all donated time is recognizable. Under accounting standards, contributed services are recognized only if they create or enhance a nonfinancial asset, or if they require specialized skills, are provided by someone possessing those skills, and would typically need to be purchased if not donated. An attorney donating legal work is generally recognizable. A board member volunteering at the registration desk generally is not, however valuable that help was. Organizations that record all volunteer hours as revenue tend to discover this during their first audit.
For the services that do qualify, the rate matters as much as the hours. The defensible rate is what the organization would have paid to purchase that service, which is often meaningfully below the professional's standard billing rate. A firm that bills commercial clients at seven hundred dollars an hour may have a stated nonprofit rate, and if your organization would realistically have hired at a lower market rate, that lower rate is the better answer. This is one of the places where documented reasoning matters far more than the precise figure.
Four questions that settle most valuations
Answer these in writing and the number usually follows
- In what market would this item actually trade, given its condition and quantity?
- Would we have purchased this if it had not been donated, and at what price?
- Are there restrictions on how we can use or dispose of it that affect what it is worth to anyone?
- What evidence supports this figure, and would a skeptical reader find it persuasive?
Why Auditors Treat This Category With Suspicion
It helps to understand why gifts-in-kind attract more scrutiny than almost any other line on a nonprofit's financial statements. The reason is a well-documented history of organizations inflating these values, sometimes by extraordinary multiples, in ways that made their program spending ratios look far better than reality. Because donated goods appear as both revenue and program expense, an inflated valuation can transform an organization's apparent efficiency without a single additional dollar reaching anyone.
The pharmaceutical cases are the best known. CharityWatch's reporting on charity valuation documented deworming medication carried on nonprofit financial statements at values many multiples above world market prices, and a Forbes account of one large charity described a restatement in which hundreds of millions of dollars in donated pills were revalued to a small fraction of the original figure once revised guidance was applied. The California Attorney General has brought enforcement actions against organizations for overvaluing in-kind pharmaceuticals at multiples of their applicable international prices.
The mechanism in these cases was consistent and instructive. The organization valued goods at the price they would command in the United States retail market while distributing them in countries where they could be purchased for a tiny fraction of that. Both numbers were real prices in real markets. The choice of which market to use was where the distortion entered, and it was a choice that could be argued in good faith right up until the magnitude made it indefensible.
This history is directly relevant to AI-assisted valuation, because a model asked what a donated item is worth will typically return a retail figure. Retail prices are what the internet is full of. Without an explicit instruction about which market applies and why, an automated valuation process will systematically drift toward the highest defensible number, which is exactly the failure pattern that produced the scandals. The tool is not doing anything wrong. It is answering the question it was asked, and the question was underspecified.
Valuation errors flatter you, which is why they persist
Nearly every incentive in this area points one direction. A higher in-kind valuation increases total revenue, increases program expense, improves the program spending ratio that watchdogs and funders look at, and makes the annual report more impressive. Nobody in the organization is naturally motivated to argue the number down.
That asymmetry is why the process needs an explicit skeptic and a written policy rather than case-by-case judgment. If your valuation method has never produced a number lower than what the donor suggested, that is a finding about your method rather than about your donors.
Where AI Genuinely Earns Its Place
With the risks named, the useful applications become much easier to identify. In every case the pattern is the same: AI does the research and the writing, a person makes the determination. That division sounds like a platitude until you notice how much of the actual labor sits on the research side.
Comparable price research is the clearest win. A staff member handed a donated commercial refrigerator has to figure out make, model, age, condition, and what similar units sell for in the secondary market. That is twenty minutes of searching per item, and it is the reason most organizations do not do it. A model that returns a range with sources attached, framed explicitly as candidate comparables for a human to assess, turns twenty minutes into two. The important design detail is that it must return a range and its sources rather than a single number, because a single number invites acceptance without review.
Service rate benchmarking works similarly. Determining what your organization would have paid for donated accounting, legal, marketing, or technical work requires knowing local market rates for that specialty, which is research a model can compile quickly. Pair it with your own history of what you have actually paid for comparable services, and you have a defensible rate with a documented basis rather than a professional's standard rate accepted uncritically.
The third application is the least glamorous and possibly the most valuable: writing the documentation. Every valuation needs a short record of what was received, what market was used, what evidence supported the figure, and who approved it. Nobody enjoys writing these, so they get skipped, and the absence of contemporaneous documentation is what turns a reasonable valuation into an audit finding. A model that drafts that memo from the structured inputs takes the friction out of the step most likely to be abandoned. The same principle underpins good practice in building audit trails for AI-assisted decisions.
Comparable research
Ranges with sources, never single figures
Identifying what similar items sell for in the relevant secondary market, with links a reviewer can check. Specify the market in the prompt, because an unspecified request will return retail pricing by default.
Categorization at intake
Feeding the disclosure requirement
Sorting incoming gifts into the categories your notes will have to disaggregate anyway. Doing this at receipt rather than at year end is what makes the disclosure a report rather than a reconstruction project.
Valuation memos
The step everyone skips
A short written basis for each significant valuation, drafted from structured inputs and approved by a person. Contemporaneous documentation is worth more at audit than a more precise number arrived at without a record.
Consistency checking
Catching drift across the year
Reviewing the full year of valuations for similar items priced differently, or rates that moved without explanation. Inconsistency is the finding auditors reach for first, and it is easy to detect in bulk.
Where an AI Valuation Will Not Hold Up
The hard limit is the qualified appraisal requirement. Where a donor needs an appraisal, that appraisal must come from a qualified appraiser meeting specific regulatory criteria, and no model output substitutes for it regardless of how well reasoned it looks. This is not a matter of accuracy. It is a matter of who is permitted to sign, and the answer is a person with credentials who accepts responsibility for the opinion. Organizations that receive significant property gifts should have a relationship with an appraiser in place before they need one.
Unique and thinly traded items are the second category. Artwork, collectibles, real property, donated business interests, specialized equipment, and anything with provenance or condition as a major driver of value all resist the comparable-based approach that AI does well. There is no dataset of similar transactions to reason from, and a model asked to value them will produce a confident number derived from loose analogies. The tell is that the reasoning cites categories rather than transactions.
Restrictions and conditions are the third, and they are where quiet errors accumulate. Donated goods that cannot legally be resold, software licenses that expire or cannot be transferred, product that must be distributed rather than monetized, near-expiry inventory, and equipment that requires costly installation all carry values materially different from an unrestricted equivalent. These facts live in the gift agreement and the loading dock, not in any price database, so they have to be supplied as inputs. A valuation process that does not have a field for restrictions will systematically overstate.
Finally, there is the international pricing question that produced the sector's most damaging cases. Any goods valued in one market and used in another require an explicit, documented decision about which market governs, made by a person who understands why it matters. This should be written into your gift acceptance policy rather than decided per transaction, because deciding it repeatedly under time pressure is how organizations end up with a pattern they cannot explain.
A Workflow That Survives Fieldwork
The practical design that works for most organizations is tiered by value, so that effort scales with materiality. Routine low-value goods can be handled with standard rates set annually and applied consistently, which is both defensible and far less work than pricing each item. Mid-value items get AI-assisted comparable research with human review and a short memo. High-value items get a person, and above the appraisal thresholds they get a professional.
Setting those standard rates once a year is the single highest-leverage step available to a food bank, thrift operation, or supply distribution program. Rather than valuing every donation, you establish a documented per-unit or per-pound rate based on a defensible methodology, apply it consistently, and revisit it annually with a written rationale. Auditors are generally comfortable with this approach because it is transparent and repeatable, which is worth more to them than item-level precision that varies with whoever happened to do the entry.
Whatever the tier, three records should exist for anything significant: what was received and in what condition, what market and method produced the number, and who approved it. Those three facts answer nearly every question an auditor will ask. Building the capture of them into the intake process rather than the close process is what separates organizations that get through fieldwork cleanly from those that spend three weeks reconstructing a year of decisions, a pattern we discuss further in our guide to AI-assisted audit preparation.
It is also worth tracing where these numbers land downstream. In-kind revenue and its matching expense flow into your functional expense allocation and therefore into the program ratio that appears on your Form 990 and in watchdog assessments. An error here propagates into your public financial profile, which is why the topic connects directly to defensible functional expense allocation and to the narrative you eventually write in your Form 990 disclosures.
A tiered valuation policy
Set the thresholds once, apply them without debate
- Routine goods: annually reviewed standard rates, applied consistently, no per-item research
- Mid-value items: AI-assisted comparables, human selection from the range, short memo
- High-value items: finance staff determination with documented market rationale
- Appraisal-threshold property: qualified appraiser, with the relationship established in advance
- Anything restricted, unique, or valued across borders: escalate regardless of dollar amount
The Disclosure Requirement Changed What You Have to Explain
Accounting Standards Update 2020-07 reshaped how contributed nonfinancial assets appear in nonprofit financial statements, and its practical effect is that your valuation reasoning is now visible rather than internal. Contributed nonfinancial assets must be presented as a separate line item in the statement of activities rather than folded into total contributions, which means anyone reading your statements can see the magnitude immediately.
The note disclosures go further. Organizations must disaggregate contributed nonfinancial assets by category, and for each category describe whether the assets were monetized or used in programs, any donor restrictions on their use, the policy for monetizing rather than utilizing them, and, critically, the valuation techniques and inputs used to determine fair value along with the principal market used. That last requirement is the one that changes behavior, because it obliges you to state in a public document which market you chose.
An organization that has been quietly valuing goods at United States retail while distributing them internationally now has to write that down where funders, watchdogs, and journalists can read it. This is precisely what the standard was designed to accomplish, and it is why the workflow discipline described above is worth building rather than improvising. If your process cannot articulate the principal market for each category, the disclosure will either be vague, which invites questions, or specific and unsupported, which is worse.
The upside for organizations that do this well is real. Clean, specific disclosure of valuation methodology signals a finance function that knows what it is doing, and it distinguishes you from peers whose gifts-in-kind line is a large number with no explanation attached. Sophisticated funders read these notes. A clear account of how you value donated goods is a small but genuine credibility asset.
Conclusion
The honest answer to whether AI can determine fair market value is no, and the more useful answer is that determination was never the expensive part. The expensive part was the research behind it and the documentation after it, both of which are now dramatically cheaper. An organization that uses AI for comparables, categorization, consistency checking, and memo drafting, while keeping the determination and the market choice with a person, gets most of the efficiency without acquiring any new risk.
The risk to watch for is subtle and worth stating plainly. An automated process left unsupervised will drift toward higher valuations, because retail pricing is what is most available and because nothing in the organization pushes back against a number that improves every metric it touches. That drift is not a hypothetical failure mode. It is the exact mechanism behind the sector's most damaging valuation scandals, and it operated for years inside organizations staffed by people acting in good faith.
The controls that prevent it are modest. Specify the market in the policy rather than the transaction. Require a range with sources rather than accepting a figure. Give someone the explicit job of arguing the number down. Write the memo at the time, not at the close. Escalate anything unique, restricted, or valued across borders. None of these require sophisticated tooling, and together they make the difference between an AI-assisted process that speeds up good practice and one that industrializes a bad habit.
If you want a single starting point, write down your standard rates for routine goods and the principal market for each category of gift you regularly receive. That document takes an afternoon, answers most of what your disclosure now requires, and turns hundreds of individual judgment calls into the consistent application of a decision you made once, carefully, when nobody was waiting on you.
Need a Valuation Process You Can Defend?
We help nonprofit finance teams build in-kind workflows that use AI where it saves real time and keep human judgment exactly where the auditors expect to find it.