DOT Audit Checklist for a Small Fleet: Organise Records With AI

A small US trucking company should organise its records by driver, vehicle and company, then use AI to rename and index scanned files, read expiry dates into a tracker, and list gaps against a checklist. A compliance specialist then confirms what the AI flagged. AI sorts and flags; it does not certify.

The speed gain is real because most preparation time goes into hunting for paper, not judging it. The risk is also real: driver files hold personal and medical data, and a model that misreads a date can tell you everything is fine. So every AI step below is paired with a human check, and the most sensitive files never go near an AI tool at all.

Follow me on Instagram@sagnikteaches

What a records review usually looks at

FMCSA, the national agency that regulates motor carrier safety, can look at a carrier's records through a safety audit, which its New Entrant pages say a new carrier gets within 12 months of starting operations, or a compliance review (also called an intervention), which can come at any time FMCSA's safety data points to a problem. Both generally take place at the carrier's principal place of business. Rules differ by vehicle weight, cargo and whether you operate for hire, so treat this as a map and confirm what applies to you on FMCSA's regulations pages or with a compliance specialist. The groups that come up most often:

Connect on LinkedInSagnik Bhattacharya

• Driver qualification files, one per driver
• Driver investigation history, kept separately
• Hours of service logs and their supporting documents
• Vehicle inspection, repair and maintenance records, per vehicle
• Drug and alcohol testing programme records
• The accident register

I have checked the retention periods below against the text of the national rules on eCFR in November 2026. Rules change, and one of these sections was amended in July 2026, so confirm the current wording before you set a destruction schedule.

Subscribe on YouTube@sagnikteaches

The checklist, group by group

1. Driver qualification file (one per driver)

Under the rule on driver qualification files, a carrier keeps one for each driver it employs, and it may be combined with the personnel file. Tick each item per driver:

• Application for employment, completed and signed.
• Motor vehicle record (MVR) from the licensing authority, obtained at hiring.
• Road test certificate, or a copy of the licence or certificate accepted as equivalent.
• Annual MVR inquiry: at least once every 12 months, with the record on file.
• Note of the annual review of the driving record, giving the name of the person who reviewed it and the date.
• Medical examiner's certificate, valid, with the expiry date visible. For CDL holders the rule works differently: where the driver's licensing record (the CDLIS motor vehicle record) shows medical certification status, the carrier meets this item by obtaining that record and filing it. Ask your specialist which route applies to each driver.
• Medical variance or exemption papers, if the driver holds any.
• Verification note for medical examiner listing on the National Registry of Certified Medical Examiners, which the current rule asks for in the files of drivers who do not need a CDL.

The file is kept while the driver is employed and for three years afterwards. The annual MVR, the annual review note and the medical certificate may be removed three years after their date. Check how your own files are organised before you cull anything.

2. Driver investigation history file (separate, restricted)

This one holds the driver's written authorisation for the alcohol and drug history check, and the responses from previous employers (or proof of a good-faith attempt to get them, with the employer's name, date contacted and what was said). The rule says it must be kept in a secure location with controlled access, limited to people involved in the hiring decision, and used only for hiring. It is kept for as long as the driver is employed and three years after.

That access rule is the reason I would keep this file out of general AI tools. Index it by name and date by hand, or let the AI see only a list of file names.

3. Hours of service

• Records of duty status for every driver who must keep them, from an electronic logging device (ELD) or on paper where an exception applies.
• Supporting documents that let someone check the logs: fuel receipts, bills of lading, dispatch records, toll and expense records.
• Retention: at least six months from receipt. The driver must also carry the previous seven consecutive days of logs while on duty.

What AI can do here: match supporting documents to log days, and list days with no supporting document at all. Whether a log is correct under the hours rules is a separate judgement for your safety manager.

4. Vehicle records (one set per vehicle)

• Identification: company number, make, serial number, year, and who owns it if not you.
• Schedule: what inspections and maintenance fall due and when.
• History: dated record of inspections, repairs and maintenance. Retained where the vehicle is housed or maintained, for one year and for six months after it leaves your control.
• Annual inspection report: kept for fourteen months from the date of the inspection report, where the vehicle is housed or maintained.
• Driver vehicle inspection reports, the certification of repairs and the certification of the driver's review: kept for three months from the date the report was prepared. A driver has to write a report only when a defect is found or reported, so a missing report on a clean day is not a gap. The rule has exceptions, including a carrier running only one commercial vehicle.

5. Drug and alcohol testing programme

• Written policy and proof each driver received it.
• Pre-employment query of the FMCSA drug and alcohol Clearinghouse before a driver performs safety-sensitive work, and at least one query a year for each driver after that.
• Retention varies: five years for positive results, refusals and programme administration records, two years for collection process records, one year for negative results. Training records for supervisors and drivers are kept while the person does the job and for two years after.

These records carry strict confidentiality rules. I would not upload test results to an AI tool at all; track only the dates (query done, annual summary filed) in your tracker and ask your third-party administrator or specialist what they allow.

6. Accident register

A list of each accident with the date, nearest city and state, driver name, number of injuries, number of fatalities and whether hazardous materials were released, plus copies of the accident reports required by governmental bodies or insurers. Kept for three years after the date of each accident.

Retention at a glance

Nine drivers, seven trucks and a shoebox of paper

Say a for-hire carrier runs seven trucks with nine drivers. The owner's records are a mix: scanned PDFs on a shared drive with names like "scan0012.pdf", paper folders, an ELD portal and a maintenance spreadsheet kept by the shop. A compliance specialist has asked for a review-ready index in two weeks. These figures are illustrative.

The plan has five passes, with the time each took in this scenario:

1. Collect (2 hours). Scan or export everything into three folders: Drivers, Vehicles, Company. One subfolder per driver and per vehicle.
2. Name and sort (about 1.5 hours with AI). Rename each file by what it is.
3. Extract dates into a tracker (about 2 hours with AI, plus a hand check). One row per document.
4. Run the gap check (30 minutes). Compare the tracker with the checklist.
5. Hand over to the specialist (1 hour). The tracker, the gap list and the folder index.

Total: roughly seven hours of owner time, where the same work done by hand from a shoebox of paper tends to take a couple of days. The specialist then spends their time judging, not hunting.

Step 1: name files by what they are

A file called "scan0012.pdf" is unfindable. Use a naming pattern the whole company shares: driver code, document type, and the date that matters (expiry for certificates, document date for the rest). Illustrative before and after:

BEFORE                  AFTER
scan0012.pdf            DRV04_MedicalCert_exp2027-03-14.pdf
scan0013.pdf            DRV04_MVR_2026-02-03.pdf
IMG_4471.jpg            DRV04_AnnualReviewNote_2026-02-05.pdf
Truck 3 stuff.pdf       TRK03_AnnualInspection_2026-04-22.pdf

Give the assistant a batch of pages (up to the plan's file limit) with a prompt along these lines, after removing licence numbers and other identifiers you do not need it to see:

For each uploaded document, tell me: (1) document type from this list: Application, MVR, RoadTest, MedicalCert, AnnualReviewNote, AnnualInspection, DVIR, Maintenance, AccidentReport, Other; (2) driver name or vehicle number; (3) the date that matters and which kind it is (issue date or expiry date); (4) how confident you are, High, Medium or Low, and why. If a date is handwritten or unclear, say so and give your best reading. Do not guess a document type: use Other.

The confidence column is the important part. A handwritten expiry date on a scanned medical certificate is exactly where a model reads a 3 as an 8. Everything marked Medium or Low goes to a person, and for expiry dates on medical certificates I would check every one by eye, whatever the label says. It takes seconds per file and a single misreading can mean a driver operating on an expired certificate.

Step 2: build the tracker

Ask the assistant to put its answers into a table you can keep in a spreadsheet. A filled-in extract:

Two things to notice. DRV07's annual review is already past its next-due date, so it is a gap even though the document exists. And DRV09's missing road test certificate may be perfectly fine if the driver was hired on a licence the rule treats as equivalent, which is exactly the kind of question the specialist answers, not the AI. Keep the "checked by" column. It forces a human name against every row.

Use a spreadsheet formula for the next-due logic rather than asking the AI each time. If the MVR date sits in D2, the next annual inquiry due is:

=EDATE(D2,12)

and a status column can flag anything due within 30 days:

=IF(E2<TODAY(),"OVERDUE",IF(E2<TODAY()+30,"DUE SOON","OK"))

The tutorial on tracking expiring driver documents with AI goes deeper on the ongoing reminders, which matter more than the one-off clean-up.

Step 3: the gap check

With the tracker built, ask the assistant to compare it with the checklist. The prompt:

Here is my document tracker (a table) and my checklist of required items per driver. For each driver, list each checklist item and mark it Present, Missing, Overdue (the annual item is older than 12 months), or Expiring within 60 days. Do not assume a missing item is not required. Only compare what I give you. End with a count per status and a list of the drivers with the most gaps.

Illustrative output for the nine-driver carrier: 81 checks (nine drivers, nine items each: the eight qualification-file items plus the investigation history file), 63 Present, 4 Missing, 6 Overdue, 8 Expiring within 60 days. Three of the six overdue items belong to one driver hired in 2024 whose annual reviews were never filed. A pattern like that tells the owner the fix is the process, not a one-off file chase.

The numbers look neat, so check them. Count the rows yourself against the tracker. Models sometimes report 81 checks when there are 79, and a wrong count is a sign that a row was skipped.

Step 4: do the same for vehicles and logs

For vehicles, put each truck in the tracker with its annual inspection date and its next due date, plus a column for whether the maintenance history is complete. The annual inspection report is kept fourteen months, so in the nine-driver example the owner would simply flag any truck whose last report is older than twelve months.

For hours of service, sampling beats full review. Pull the last six months of logs and supporting documents for two drivers and ask the assistant to find log days with no supporting document. Illustrative output: "Driver 02, 11 of 94 days lack a supporting document; 9 of the 11 fall on days the truck was at the yard." That kind of result suggests the paperwork was never created, not that something was hidden, but your specialist decides how to treat it.

What AI must never be allowed to decide

• Whether a document meets the rule, such as whether a medical certificate is valid for a given driver or whether a driver is exempt from logging.
• Whether a driver is qualified to drive.
• Anything to do with drug and alcohol test outcomes.
• Whether to disclose a gap, or how.

Never recreate, backdate or fill in a missing document. If the road test certificate is missing, record that it is missing. Fixing it going forward is fine, such as pulling an up-to-date MVR today and dating it today. Pretending it was there all along is the one move that turns a paperwork problem into a serious one.

The medical certificate AI read as valid

Here is how this failure tends to look, as an illustration. A carrier's tracker said a driver's medical certificate expired on 2028-03-14. The scanned original showed a handwritten expiry date that, on a closer look, read 2026-03-14: the model had read the "6" as an "8", so the tracker showed two more years of validity on a certificate that had already run out. The tracker's "Checked by" column was blank for that row, so the check never happened, and the error surfaced only when the specialist asked to see the original.

The fix was a rule rather than a better model: every date that drives a driver's eligibility is verified by a person against the original, and the tracker row stays red until the person's initials are in. AI gets you to the right pile fast. It does not get to sign off a pile.

Privacy and tool choices

Driver files contain licence numbers, birth dates, medical information and employment history. A few practical limits:

• Use a business-grade plan whose terms state it does not train on your content, not a free personal account.
• Redact or crop licence numbers and birth dates before upload where the task does not need them.
• Keep the investigation history file and all drug and alcohol test results out of general AI tools.
• Keep the tracker itself, with names and dates only, in a spreadsheet with access limited to the people who run compliance.
• Delete uploaded files from the AI tool's project or chat when the job is done.

If you already use a trucking accounting or fleet system, some of this may be built in. What to check in AI software for a small fleet lists the questions to ask before paying for a new tool.

When a request for records actually arrives

Preparation changes what the first day feels like. A request usually names the records wanted and a deadline, and the rules require records to be produced within the time the requesting official specifies. A short response plan, written before you need it, keeps the day calm:

1. One contact. Name one person who receives the request, logs the date and time it arrived and keeps a copy. Everyone else sends questions to that person.
2. Read the request itself. List exactly which records, which dates and which drivers or vehicles it names. Answer what is asked and no more.
3. Call your specialist the same day. They will want the tracker and the gap list, so the work you did above becomes the first thing you send them.
4. Pull from the index, not from memory. The folder structure and naming pattern let you assemble the exact set in minutes. Copy files rather than moving them, so the originals stay where they are.
5. Keep a delivery log. Note what you sent, when, and how, so nothing is disputed later.

An up-to-date records index is useful to have on hand before anyone asks. Ask the assistant to generate it from the tracker once a month and save the dated copy:

From this tracker, produce a records index for the specialist: one section per driver with every document on file (type, date, expiry), one section per vehicle with the last annual inspection, last maintenance entry and open defects, and a company section listing the accident register entries for the last three years. Add a final list headed "Gaps and open items" and nothing else. Do not add comments or advice.

Keeping "nothing else" in that prompt matters. Left alone, assistants add reassuring commentary such as "this looks compliant", which is the opposite of what you want in a document that may go to a reviewer.

A short example of the gap list in use

Take the nine-driver carrier again. Its gap list had four missing items. On the specialist call, two turned out to be non-issues: one driver was hired on an equivalent licence that the road-test rule accepts, and one note belonged in a different file. The other two were real: a missing annual MVR for one driver and an annual inspection report for a trailer that the shop could not find. The carrier requested a fresh MVR the same day, put it on file dated that day, and asked the shop for a copy of the missing report. The tracker carried both items until closed. Half the AI's flags needed a human explanation, which is the normal ratio and the reason the specialist stays in the loop.

A monthly routine so the next review is boring

A one-off clean-up decays unless something keeps it current. A short routine for the owner or office manager:

1. First working day of the month: open the tracker, filter for OVERDUE and DUE SOON, and assign each to a named person.
2. Add new documents to the right folder using the naming pattern, the same day they arrive.
3. Log each accident in the register within a week, with the reports attached. The accident report form and claim summary tutorial gives a structure for this.
4. Once a quarter, run the gap prompt again on the updated tracker and send the result to your specialist.

Fifteen minutes a month beats the two-week scramble before a review. And the safety policies a reviewer asks to see, such as your written plan and driver training notes, fit the same folder structure; drafting a trucking safety plan and toolbox talks with AI covers writing them.

Questions about preparing records for a DOT review

Can AI tell me whether I would pass a DOT audit?

No. AI can sort files, read dates and list what appears to be missing, which saves hours of preparation. Whether a record meets the rule is a judgement for a compliance specialist or your safety consultant, and the regulator decides in a real review.

Is it safe to upload driver files to an AI tool?

Driver files hold licence numbers, medical details and employment history. Use a business-grade plan whose terms say it does not train on your content, redact what the task does not need, and ask your compliance specialist before uploading drug and alcohol test records at all.

How far back should I keep each record?

It varies by record: six months for hours-of-service logs, three months for daily vehicle inspection reports, three years for the accident register, and the length of employment plus three years for driver qualification files. The retention table in this tutorial lists them with the rule each comes from.

What if the AI finds a missing document?

Do not recreate or backdate anything. Record the gap, fix it going forward, for example by requesting a fresh motor vehicle record today, and take advice from a compliance specialist on whether and how to disclose it.

Does this apply to my business if I only run two trucks?

Many of the same records apply whatever the fleet size, but which rules apply depends on your vehicles, cargo, operating area and whether you are a for-hire or private carrier. Check with FMCSA or a compliance specialist before assuming you are exempt from anything.

Further reads

• AI in the Trucking Industry: What a Small Fleet Owner Can Use Now — See other jobs in a small carrier where AI helps.
• Driver Scorecard Template: Fair Scoring From Telematics With AI — Score driver behaviour fairly from telematics data.
• How to Use AI in a Trucking Company: Load Paperwork Without Retyping — Stop retyping load paperwork that also supports your logs.
• Detention Pay: AI Tracks Wait Times and Invoices Brokers — Wait-time records double as supporting documents.
• Fleet Admin for Small Businesses: Automate Services, Checks and Fuel — Automate service reminders so maintenance records stay current.
• AI Dash Cams for Fleets: Cost, Privacy and Driver Consent — Hardware, subscription and contract costs for a small fleet, then the privacy notice, consent and no-emotion-inference rules to settle first.
• AI Tools for Fleet Management: Spot Fuel-Card Misuse in a Spreadsheet — A worked example: a fuel-card CSV and a location export, five flag rules, a sample exceptions list and how to verify each flag with the driver.
• AI in Fleet Management: Auditing Yard Time and Tolls From Telematics — Use AI on a telematics export to find drivers idling in the yard on the clock and tolls that never reached an invoice.
• Driver Walkaround Check: Triage Defect Reports With AI — Sort daily walkaround defect reports into severity bands with AI, while the off-the-road rules come from DVSA or FMCSA pages you hold yourself.
• Operator Licence Compliance: AI Tracks Checks, Audits and Renewals — A diary-and-sheet system for safety inspections, driver checks, maintenance records and change notifications, with AI prompts that keep it current.
• Owner-Operator Settlement Statement: Check Deductions With AI — Turn a lease and a settlement statement into two tables, compare them line by line, and send a dispute email with the evidence.
• Trucking Companies Using AI: A Claude Audit of Yard Time and Tolls — Export a month of telematics data, let Claude flag long yard stops and toll gaps, then check every flag against the raw rows.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: eCFR text of 49 CFR sections 382.401, 382.601, 382.701, 390.15, 391.25, 391.51, 391.53, 395.8, 396.3, 396.11 and 396.21 (eCFR API, current to 3 November 2026), read 5 November 2026; FMCSA New Entrant Safety Assurance Program page.