---
title: Before AI Reads Your Asset Records, Run the 20-Addition Test
description: "Before AI tools read your equipment data, run the 20-addition test: check your last 20 fixed asset additions against the floor to see if the record holds."
image: https://www.solotruth.com/hubfs/Blog%20CA%2025-01.jpeg
---

[![SoloTruth](https://www.solotruth.com/hubfs/logo.svg)](https://www.solotruth.com/)

- [Home](https://www.solotruth.com/)
- [Inspection Management](https://www.solotruth.com/inspections)
- [Asset Management](https://www.solotruth.com/fixed-assets)
- Resources
  
  [Blog](https://www.solotruth.com/blog) [Press](https://www.solotruth.com/press) [Glossary](https://www.solotruth.com/glossary)
- [About Us](https://www.solotruth.com/about-us)

[Contact Us](https://www.solotruth.com/contact-us)

[← Blog](https://www.solotruth.com/blog) / Continuous Asset Verification

Continuous Asset Verification Fixed Asset Management Fixed Asset Register ARM

# Before AI Reads Your Asset Records, Run the 20-Addition Test

Tim Harris

Nov 15, 2026, 10:59:17 AM

![](https://www.solotruth.com/hubfs/Blog%20CA%2025-01.jpeg)

[← Back to Blog](https://www.solotruth.com/blog)

The 20-addition test is a quick check of whether your equipment records are fit for AI to work from. Pull the last 20 assets added to your fixed asset register and check whether someone who has never seen them could walk out, find each one, and describe its condition from the record alone.

AI tools built for plants, warehouses, and maintenance teams are starting to arrive. They draft work orders, summarize repair histories, and answer questions about equipment in plain English. Every one of them works from the records a company already keeps, and most of those records have never been checked against the floor. In a 2025 Manufacturing Leadership Council survey, only 39 percent of manufacturers said they have a process to verify the accuracy and quality of their manufacturing data. That survey covers manufacturing data broadly, not equipment records specifically, so treat it as context rather than a measure of any one register. This post is for chief operating officers, chief executives, and the finance and maintenance leaders who report to them. It explains why the asset record is the first thing AI will expose, how to test yours in an afternoon, and what it takes to keep it right once you know where it stands.

## What Is the 20-Addition Test?

The 20-addition test checks your 20 most recent fixed asset additions against the floor, to see whether the record alone is enough to find each asset, identify it without doubt, and know what shape it is in.

A fixed asset register is the list of equipment a company owns and depreciates, kept by accounting and usually held in the ERP. Additions are the newest lines on it: the forklifts, compressors, racking, and vehicles bought and placed in service in recent months. They're the lines most likely to be right, because nobody has had time to move, scrap, or lose them yet. If the newest records can't pass, the older ones won't do better.

Run it like this. Pull the 20 most recent additions and hand each record to someone who has never seen the asset. A line passes only if that person can do all four of these things from the record alone:

- **Find it.** The location on the record is specific enough to walk to, not just a site name or a cost center.
- **Identify it.** The tag or serial number on the record matches the one on the machine.
- **Describe it.** There is dated evidence of its condition, not just a purchase price and an in-service date.
- **Trust it.** The record shows when the asset was last verified in person, and by whom.

The last check matters more than it looks. Suppose the crew moved forklift 10432 to Site B in the spring, and the register still shows it in service at Site A. Ask an AI tool where the forklift is and it answers Site A, because nobody who saw the move told accounting. A record that shows when the asset was last verified in person, and by whom, tells you how far to trust that answer.

Score the result as a simple count out of 20, and note which of the four checks each failing line missed. That tells you whether the fix is location, identity, condition evidence, or how recently anyone verified the asset in person.

## Why Asset Records Drift Before AI Ever Reads Them

Asset records drift because the people who see equipment change every day are rarely the people who update the register, and nothing in most companies connects the two.

1. **Records start from paperwork.** An addition is usually created from an invoice or a purchase order. That proves a purchase happened. It doesn't prove the asset was installed where the record says, or that it's still there.
2. **Maintenance and the register don't share changes.** A shared asset number links the two records, but a status change or a move logged in maintenance doesn't update the register on its own.
3. **Changes on the floor don't travel.** A move, a breakdown, or a retirement gets logged in a work order, or not at all. The register keeps the old version until someone thinks to tell accounting.
4. **Counts are infrequent and start from the list.** A count that begins with the register can confirm the equipment it lists. It can't find equipment the register never recorded, and it rarely says anything about condition.
5. **Nobody owns the space between the two records.** Operations owns the equipment and accounting owns the register. The question of whether they still describe the same thing belongs to neither.

### Record as Written vs. Record Backed by Field Evidence

|  | **Record as written** | **Record backed by field evidence** |
| --- | --- | --- |
| **Where it comes from** | Invoices, purchase orders, and manual entry | Observations at the asset, reviewed by accounting |
| **Status and location** | Whatever was true when the record was last updated | Confirmed at the asset and approved by accounting |
| **Condition** | Unknown until someone looks | Dated photographic and condition evidence |
| **Location** | A site or a cost center | Where the asset was last verified, with a timestamp |
| **How often it is checked** | At the next count | During work crews already do |
| **What an AI tool inherits** | Every gap in the paperwork | Evidence it can point back to |

 

The difference between the two columns isn't software. It's who last looked at the asset, and whether what they saw reached the register in a form anyone can use. An AI tool connected to the left-hand column will be fluent and wrong. Connected to the right-hand column, it can show where every answer came from.

## The Real Cost of an Unverified Asset Record

An unverified asset record costs money twice: once in the decisions made from it, and again when someone has to stop and check it by hand.

- **Equipment that is gone but still on the books.** Kroll's fixed asset register health check found that more than 15 percent of recorded assets at one manufacturing client were no longer in service. That's one client, not a rate for anyone else, and it shows what a real check can turn up.
- **Equipment that can't be verified when someone looks.** A federal audit of routine equipment purchases found that 43 percent of items tested, worth $1.8 million, could not be verified as existing. The auditors called their sample nonrepresentative, so it isn't a rate for any other organization either.
- **Paperwork that isn't proof.** Auditing standards state that management's written word that equipment exists is not sufficient evidence on its own. An auditor still needs something behind the line.
- **AI answers nobody can rely on.** A summary drafted from a wrong record is wrong faster, and it arrives sounding confident.
- **Time spent re-checking.** Every AI answer that has to be verified by hand takes back the time the tool was bought to save.

*"Every ERP records what should be on the floor. Verifying it is a different job, one no ERP was built to run. SoloTruth runs it."*

Tim Harris, CEO, SoloTruth

The cost that rarely shows up in a budget is trust. Once a maintenance lead or a controller catches an AI tool reporting a machine as in service after the crew retired it, they stop believing its answers, and the investment stalls before it pays back.

## Who Is Most Affected?

The companies most exposed are the ones with a lot of equipment spread across sites, and a plan to put AI on top of the records that describe it.

- **Chief operating officers of multi-site operations.** Warehousing, distribution, manufacturing, and transport fleets move equipment between sites every week, and the record follows slowly or not at all.
- **Maintenance and reliability leaders considering AI tools.** They'll be the first to see an assistant answer from a work order that the register contradicts, because work orders are where changes are logged first.
- **Chief financial officers and controllers.** The register drives depreciation, insurance values, and property tax, and the auditor will ask how it was checked.
- **Companies preparing for an ERP migration or a sale.** Every record gets moved or examined, accurate or not, and a buyer's diligence team reads the same register the AI tool does.
- **Utilities and other plant-heavy businesses.** Long asset lives and large registers mean small status and location errors add up over decades.

## What to Look For in a Way to Keep the Record Right

Not every approach to asset accuracy will hold up once AI starts reading the record. Look for six capabilities.

1. **Evidence from the asset, not from paperwork.** The record should be confirmed by someone standing next to the equipment, not inferred from an invoice.
2. **Captured by the people already there.** Crews and technicians record it during work they already do, on devices they already carry, so verification is frequent without becoming a separate project.
3. **Findings from the floor reach the register.** When a crew finds equipment moved, idle, or gone, the finding goes to accounting for approval and updates the register, rather than waiting for the next count.
4. **Condition and location with a timestamp.** Photographic and condition evidence, verified with where and when it was gathered, rather than a simple yes, it's here.
5. **Human review before the record changes.** A named person in accounting approves each finding before the register moves, and the reason is kept with the change.
6. **An evidence trail for every asset.** Anyone, whether an auditor or an AI tool, can trace a line on the register back to what was seen, by whom, and when.

## What Good Looks Like

An asset record that's ready for AI is one where every line can be traced to something a person saw, and changes on the floor reach the register in days rather than at the next count.

1. **Run the 20-addition test and keep the score.** It takes an afternoon, it gives you a baseline, and it tells you which of the four checks fails most often.
2. **Give floor findings a path to the register.** Decide who reviews a finding from the floor, such as a move, a breakdown, or a retirement, and how fast it reaches the register. Every later step depends on that handoff.
3. **Put capture into rounds that already happen.** Maintenance rounds, inspections, and shift walks already put people next to equipment. Adding a scan and a photo to that work costs minutes, not a project.
4. **Route every finding through review.** Evidence first, approval second, register last. That order is the control, and it's what an auditor will want to see.
5. **Start with one slice.** Pick one site or one asset class and check it in both directions: from the register out to the floor, and from the floor back to the register.
6. **Connect AI to a record you can defend.** I believe the order matters: rules first, a verified record second, AI third. Reverse it and you automate uncertainty at a higher price.

## Common Misconceptions About AI and Asset Records

Three assumptions come up often when companies plan to put AI on top of their equipment data.

### Misconception: AI will clean up the data.

**Reality:** An AI tool can spot that the work order and the register disagree. It can't confirm which one is right, whether the forklift is still in the yard, or what condition it's in. Only someone at the asset can.

### Misconception: The count reconciled, so the register is right.

**Reality:** A count that starts from the register confirms the equipment it lists. It can't find equipment that was never recorded, and it says little about condition, which is what most AI use cases need.

### Misconception: Fixing this needs new tags and readers.

**Reality:** Most equipment already carries a QR code, a barcode, or a serial plate, and the crews nearby already carry phones. The gap is the process that connects them to the record, not the hardware.

## Frequently Asked Questions

### What is the 20-addition test?

It's a quick check of your 20 most recent fixed asset additions. For each one, someone who has never seen the asset tries to find it, identify it, and describe its condition using only the record.

### Why test the newest assets rather than the oldest?

The newest records should be the most accurate, because there has been the least time for equipment to move, break, or leave. If recent additions fail the test, older records are very unlikely to do better.

### Can an AI tool fix an inaccurate asset register?

An AI tool can spot likely duplicates and gaps in the records it reads. It can't confirm what exists on the floor or what condition it's in. That evidence has to come from people at the asset.

### How often should equipment records be checked against the floor?

More often than once a year. The most practical approach checks equipment during work that already happens, such as maintenance rounds and inspections, so verification is frequent without a separate counting project.

### Who should own the accuracy of the asset register?

Accounting owns the register and operations owns the equipment, so accuracy needs both. Operations supplies evidence from the floor, and accounting reviews and approves each change before the register moves.

### What is asset relationship management (ARM)?

Asset relationship management keeps the fixed asset register matched to the equipment on the floor. SoloTruth's Asset Relationship Management platform (ARM) gathers field evidence through the crews already on site and routes each finding to accounting for review.

## Check the Floor Before AI Reads the Record

AI tools will read whatever asset record you give them, and they'll repeat its mistakes with confidence. The 20-addition test tells you in an afternoon whether yours is ready.

This is the gap ARM was built to close. ARM schedules the checks from our portal, your crews gather and verify photographic, condition and other asset evidence, with location and a timestamp, on the phones they already carry, and accounting reviews every finding before the register changes. There's no extra hardware, and it works alongside the ERP and maintenance systems you already run. SoloTruth has used the SoloTruth/VirtuSpect platform to manage and complete inspections, verifying condition and location, at over 1,000 businesses across the US. The first step is a proof of concept on one slice of equipment.

Book a 30-minute strategy call at calendly.com/tim-harris-solotruth/30min to see where your asset record stands before AI starts reading it.

*Last Updated: November 2026*

Share this article: [LinkedIn](https://www.linkedin.com/sharing/share-offsite/?url=https://www.solotruth.com/blog/ai-asset-records-20-addition-test) [X](https://twitter.com/intent/tweet?url=https://www.solotruth.com/blog/ai-asset-records-20-addition-test)

#### Ready to close the gap?

Schedule a 30-minute strategy session with the SoloTruth team.

[Schedule a Call](https://calendly.com/tim-harris-solotruth/30min)

#### Topics

Continuous Asset Verification Fixed Asset Management Fixed Asset Register ARM

#### SoloTruth ARM One-Pager

The evidence and control layer above ERP. Two pages. No fluff.

[Download PDF](https://www.solotruth.com/hubfs/SoloTruth-ARM-One-Pager_2026_03_28.pdf)

### Ready to close the gap between your records and reality?

Schedule a 30-minute strategy session with the SoloTruth team.

[Schedule a Strategy Session](https://calendly.com/tim-harris-solotruth/30min)

![SoloTruth](https://www.solotruth.com/hubfs/logo.svg)

- [Home](https://www.solotruth.com/)
- [Inspection Management](https://www.solotruth.com/inspections)
- [Asset Management](https://www.solotruth.com/fixed-assets)
- [Blog](https://www.solotruth.com/blog)
- [Press](https://www.solotruth.com/press)
- [About Us](https://www.solotruth.com/about-us)
- [Contact Us](mailto:info@solotruth.com)

© 2026 SoloTruth. All rights reserved. [info@solotruth.com](mailto:info@solotruth.com) [Privacy Policy](https://www.solotruth.com/privacy-policy) [Terms of Use](https://www.solotruth.com/terms-of-use)

<https://www.facebook.com/solotruthhq> <https://x.com/TimothyHar9882> <https://www.linkedin.com/company/solotruth/> <https://www.youtube.com/@SoloTruthHQ>

```json
{
  "@context" : "https://schema.org",
  "@type" : "BlogPosting",
  "author" : {
    "@type" : "Person",
    "name" : "Tim Harris",
    "url" : "https://www.solotruth.com/blog/author/tim-harris"
  },
  "dateModified" : "2026-11-15T22:05:38.754Z",
  "datePublished" : "2026-11-15T17:59:17.000Z",
  "headline" : "Before AI Reads Your Asset Records, Run the 20-Addition Test",
  "image" : [ "https://www.solotruth.com/hubfs/Blog%20CA%2025-01.jpeg" ],
  "mainEntityOfPage" : {
    "@id" : "https://www.solotruth.com/blog/ai-asset-records-20-addition-test",
    "@type" : "WebPage"
  },
  "publisher" : {
    "@type" : "Organization",
    "logo" : {
      "@type" : "ImageObject",
      "url" : "https://www.solotruth.com/hubfs/LIGHT_Logo_RGB.png"
    },
    "name" : "SoloTruth"
  }
}
```

```json
{
  "@context" : "https://schema.org",
  "@type" : "Article",
  "dateModified" : "2026-11-15",
  "datePublished" : "2026-11-15",
  "mainEntityOfPage" : {
    "@id" : "https://www.solotruth.com/blog/ai-asset-records-20-addition-test",
    "@type" : "WebPage"
  }
}
```