Call note compliance monitoring, explained.

What a call note is, and why a compliance team wants to read it

From a doctor's office to a coached rep: a call note in five steps

1. The visitTuesday, 2:20 p.m. A rep sees a dermatologist for twelve minutes and, on the way out, offers to sponsor a dinner.
2. The noteThe rep dictates what happened the next morning. The note lands in the CRM with hundreds of others.
3. Someone reads itReader, sample or word list: this is where the offer is found, or left unread.
4. It goes to the right teamInvestigators for the offer, Medical Information for the question, a compliance reviewer for the claim.
5. A person decides, and the record staysAn investigator opens a case, the rep is coached, and the note, the decision and the reason remain on file.

Step 3 is the subject of this guide. The people and dates in this example are fictional.

An unbooked dinner can count, and three more surprises in call notes

Four terms a call note reader needs: call note, remuneration, off-label, Open Payments

Fourteen terms are defined in the one-page summary.

The five kinds of phrase a compliance reader hunts for in a post-call note

In shortField compliance teams read notes for five things: value offered in return for prescribing, gifts and meals outside policy, off-label questions, promotional language, and competitor or access remarks. Each goes somewhere different, and the fifth is useful business information that needs no compliance action.

A call note runs to a few sentences or a couple of minutes of speech, and one note can hold several of these. The example at the top of this page holds four: the dinner, a question about a 15-year-old, a comparison with a competing product, and a health plan's refusal. Teams use different labels. These five cover most of what commercial compliance policies list.

That shape, with the most serious finding as the smallest bar, explains why sampling is a poor fit. Pull 5% of notes at random and you would expect to see about three of those 61 offers. Which three would be luck.

Anti-Kickback, False Claims, Open Payments, FDA: the rules a call note touches, as of November 2026

In shortNo single law governs call notes. Four bodies of US law reach what a note records: the Anti-Kickback Statute, the False Claims Act, Open Payments and FDA's promotion rules. Over them sit voluntary codes and guidance from the federal health department's inspector general (OIG), PhRMA, which represents research drug companies, and the Justice Department (DOJ). We found no federal rule that requires a call note, or sets a deadline for reading one.

Which rule reaches which sentence in a call note

States add their own layer. Vermont bars most gifts to health care providers, food included, with a civil penalty of up to $10,000 for each violation.23 Minnesota bars gifts to practitioners, leaving out items worth $50 or less in total in a calendar year.24 Other states have their own rules. Companies outside the US work under different national laws and industry codes. Apart from the EU AI rules in section 11, this guide covers the US. Have counsel list what applies to your field force.

How a rep's offer becomes a kickback question: one purpose, four facts, no payment needed

In shortThe statute covers offering as well as paying, and OIG's guidance says an arrangement is in play if any one purpose is to generate business. So the words of the offer matter more than its price. Investigators need four facts: who offered, to whom, what of value, and in return for what.

The Anti-Kickback Statute has three features a newcomer should know. It covers whoever "offers or pays" remuneration.1 It does not require that the person knew about this statute or specifically meant to break it.1 And OIG reads it widely: arrangements such as meals, gifts and entertainment "potentially implicate" it "if any one purpose of the arrangement is to generate business".12 OIG adds that the outcome depends on the facts, and that following the PhRMA Code "should substantially reduce" a company's risk.12

Words in a note cannot prove a crime. They can show that someone needs to look. A note saying a dinner depends on three new patients states a condition. A note saying a dinner took place after an educational talk does not. Those sentences need different handling, even when they share most of their words.

Three fictional sentences, three different answers

Why a modest meal is taken seriously

A $20 lunch sounds trivial, and compliance teams still take it seriously. Research on Medicare prescribing suggests why. A 2016 study of 279,669 physicians found that 95% of the payments tied to four promoted drugs were meals, with a mean value under $20. Physicians who received a single such meal had higher relative rates of prescribing the promoted drug.25 The study is cross-sectional: it shows an association and cannot show that the meal caused the prescribing.

Causation and whistleblower suits: what the courts are still arguing about

Two live disputes affect how much a kickback case can cost, and neither makes an offer acceptable. Check the status of both with counsel before relying on either.

Meals, gift cards and hockey tickets: the Open Payments arithmetic behind a line in a note

In shortA meal is reported per person, split among everyone who ate, staff included, and shown only for covered clinicians who ate. Under $13.82 it may drop out, unless that clinician passes $138.13 for the year. So a note that names who was at lunch feeds a federal report, and a missing name makes that report wrong.

Open Payments requires manufacturers to report transfers of value to covered recipients: physicians, physician assistants, nurse practitioners, clinical nurse specialists, certified registered nurse anesthetists and anesthesiologist assistants, certified nurse-midwives and teaching hospitals.26 For a shared meal, the rule says to divide the cost by everyone who ate, including "non-covered recipients, such as office staff", and to report only for covered recipients who actually ate.22

Working out one catered lunch

Take the catered lunch from the third example in section 4 and add a figure, since the note gives none. Suppose the food cost $184, for eight people: the physician, Dr. Fischer; two nurses; others from the office; and two guests.

The $184 is our assumption, for the arithmetic only. The sample note gives no amount, which is why the rep would be asked.

How an expense report can quietly shrink a meal

In its August 2026 resolution, the drug maker Veloxis admitted that employees falsified expense reports by adding names to attendee lists, which lowered the apparent cost per attendee, and by leaving out clinicians who did attend.27 Its reports to the Centers for Medicare & Medicaid Services (CMS), the agency that runs Open Payments, drew on those records and understated the payments, which led to the $1.55 million civil penalty.28,19 A note that says who ate and an expense report that lists attendees can check each other, so read them together.

Open Payments reports a meal; the PhRMA Code and company policy limit it

How much money runs through the Open Payments database

CMS's own penalties are rare. In fiscal 2025 it sent 9 warning letters, completed 4 audits and issued no Open Payments penalties, with a decision on one audit pending.30 The rule sets $1,000 to $10,000 per unreported item, and $10,000 to $100,000 if the failure is knowing, adjusted for inflation.31 At the 2025 levels those ranges are $1,443 to $14,432 and $14,432 to $144,329.32 OMB canceled the 2026 adjustment for agencies.5 Veloxis's recovery came through the Justice Department.19

Off-label requests in call notes: who may answer, and what turns a question into a solicited one

In shortDoctors may ask about unapproved uses, and the company may answer through its medical team. FDA's draft guidance calls a question unsolicited when the rep did nothing to prompt it. A rep who mentions the use and invites a request makes the follow-up solicited, and it can then be evidence of intended use. So a note's order of events matters.

Off-label means outside the approved label: another disease, another age, another dose. Physicians often have good reason to ask. Companies must be careful about how they respond, because under 21 CFR 201.128 a firm's intended use can be shown by "oral or written statements", while its mere knowledge that doctors prescribe a drug off-label does not establish intent.7 So a rep's words can count as evidence of the firm's intent. A doctor's off-label prescribing, by itself, does not.

FDA's 2011 draft sets out how a firm should respond to a private question from one doctor. It recommends referring the question to the firm's medical or scientific department, keeping the answer non-promotional, and giving sales and marketing staff "no input" on its content. It also lists what to keep on record: the nature of the request, who asked, what was sent and any follow-up.9

This text is still a draft, so treat it as FDA's current thinking. Under a procedure built on it, the rep's job is simple to state: decline, offer the medical team, and write down what happened.

The same question, handled two ways

FDA's draft also gives the reverse case, which reviewers look for: if a sales representative mentions an unlabeled use and invites a professional to ask for more, any resulting request counts as solicited.9 A note that reads "I mentioned the data in teens and she asked for it" is a different event from one that reads "she asked about teens", even though both end in a request.

The January 2025 guidance covers something separate: scientific information a firm chooses to share on its own initiative. In an in-person visit, FDA says, it should stay apart from promotional communications, and those who share it should have specialized training, or know how to direct questions to medical or scientific staff.10 FDA's page says the guidance is not for current implementation until OMB decides on its information collection.10 Your policy should say who, if anyone, in the field may do it.

What whistleblower complaints say off-label programs looked like

The best public picture comes from a 2011 study of 41 unsealed whistleblower complaints from 18 cases, which led to $7.9 billion in recoveries. These are allegations, and they are old, but the patterns still match what compliance teams look for.33

The second chart is a reason to read for both kinds of phrase in the same pass. An offer and an off-label question in one note, as in this guide's example, is a pattern with history behind it.

"Works faster than anything else": promotional claims a rep can make out loud, and FDA's 2025 letters

In shortWhat a rep says to a doctor counts as promotion. FDA's rules treat unsupported comparisons and overstated claims as misleading, and its promotion office lists sales representative presentations among what it regulates. FDA sent 74 letters to drug and biologic makers in 2025, nearly all from September 9 on, most of them about TV ads.

FDA's Bad Ad program tells clinicians how to report misleading promotion, anonymously if they wish, and it lists "Sales representative presentations" alongside television ads and printed pieces.34 The advertising rule says a promotion is false or misleading if it claims a drug is better or safer than another without substantial evidence, or safer than has been demonstrated.8 That gives a compliance reviewer three questions about any sentence: does it make an absolute claim, does it compare without head-to-head evidence, and does it say more than the label?

What FDA's 2025 letters tell a field compliance team

One law firm counted 74 letters to pharmaceutical and biologic manufacturers in 2025, 10 Warning Letters and 64 Untitled Letters, FDA's two kinds of enforcement letter. That was the highest annual total in nearly 25 years. Only five came before September 9, 2025. Most concerned consumer television ads, and 16 challenged promotion aimed at health care professionals. Benefit claims cited included overstated efficacy and unsubstantiated superiority.18 We found no equivalent public count for 2026.

These letters mostly concern ads and printed material. A spoken claim leaves little trace except a rep's own note, which is one reason compliance teams want to read those notes, and coach from the actual words.

Keyword lists against plain-word offers: what call note review misses and flags in error

In shortA phrase list finds what someone thought to list. Real offers use ordinary words, and real false alarms come from words that are harmless where they stand. Reading for meaning changes both errors, and it brings new ones you must measure. No public benchmark exists, so test any method on your own notes.

The common first tool is a list of phrases. One CRM vendor's documentation shows the idea: compliance users choose the free-text fields to watch and the phrases to flag, with "save money" as its example, and then review the notes that match.35 This is cheap and simple to explain, and it does catch the words on the list. It has two blind spots, and tuning the list trades one for the other.

A caution on those numbers: nobody can count what no one found. In a sample month, "61" means what the better reader found, so the keyword list's 9 is measured against that reader. A reader for meaning makes its own mistakes: it can read intent into a joke, or miss an offer told in shorthand.

We found no published recall (the share of real issues found) or false-alarm figure for any method on real call notes. Vendors publish descriptions: one 2026 guide from a CRM vendor says large language models now read free text with a consistency and speed impractical for human reviewers, and screens notes as they enter the system, but it gives no test results.3 A 2026 article promoting a review tool has no figures either.36 Ask each vendor for its numbers on your notes.

What field conduct has cost drug makers: Veloxis, Gilead, Biohaven, Medtronic and a record False Claims year

In shortRecent settlements turn on meals, trips, guests and speaking fees, and three of the four below began with a whistleblower suit. The Justice Department reported more than $6.8 billion in False Claims Act recoveries for fiscal 2025 and 1,297 whistleblower suits, both records. Employees' own messages and records can become the evidence.

Fiscal 2025 recoveries topped $6.8 billion, the most in the False Claims Act's history, and over $5.7 billion of that involved health care.40 These totals cover many kinds of matter, far beyond sales conduct.

For the long view, Public Citizen counted 412 pharmaceutical settlements worth $38.6 billion from 1991 to 2017. It also found that penalties for unlawful promotion fell 94%, from $8.7 billion in 2012 to 2013 to $527 million in 2016 to 2017.41 That count ends in 2017. The four matters above, all later, turn on meals, speakers and payments.

What a corporate integrity agreement asks of a field force monitoring program

In shortA drug maker that settles a federal fraud case may also sign a Corporate Integrity Agreement. These spell out monitoring in detail, and the two read for this guide name call notes. OIG's 2003 guidance and DOJ's 2024 questions ask for similar things in general terms. So a company without such an agreement can still read what one requires as a guide to good practice.

A Corporate Integrity Agreement, or CIA, is a multi-year contract with the federal health department's inspector general (OIG). Its monitoring sections are the most concrete public description of this work. The agreement Veloxis signed in August 2026 calls for a Field Force Monitoring Program "to identify potential inappropriate promotional activities or other improper conduct", in two parts.42 Endo's 2014 agreement used similar language, with call notes on its list.43

Results go to the Compliance Officer. A probable violation becomes a Reportable Event, which the company must report to OIG in writing within 30 days of determining that it exists.42 That clock runs from the company's determination and applies under the agreement. It says nothing about how soon a company must find out, which is the gap a monitoring program is there to close.

For a sense of scale, take this guide's fictional field force of 612 reps: 15% is 92 full-day ride-alongs a period. The records review leaves more choice of method. In 2003 OIG said a commitment to compliance "can be measured by" a company's "commitment to training and monitoring its sales force".12 DOJ asks whether compliance staff can reach the data they need, and whether technology given to compliance matches technology given to sales.16

Here is what a 5% sample leaves unread in this guide's fictional month.

Voice notes and AI review: what transcription gets wrong, and which AI rules actually apply

In shortSoftware can read every note, and that changes coverage more than judgment. Speech tools mishear, unevenly across speakers. Most pharma AI rules do not cover this work, because it is a commercial compliance task, while DOJ's questions and, in Europe, the AI Act's worker-monitoring rules do apply. A person decides every case.

Automation changes one thing outright: reading every note stops being an expensive idea. It does not change who decides. Intent, context and fairness to a named employee need a person, a point even the promoters of review software make: strong programs "intentionally combine AI-driven detection with expert review and decision-making".36

What speech recognition gets wrong

Dictated notes depend on a transcript, and a transcript can be wrong in ways that matter when the text may lead to a case against a person. A study run in 2023 and presented in 2024 found that roughly 1% of one speech tool's transcriptions, on clips that included people with aphasia, a condition that limits speech, held whole invented phrases. Its lead author says the tool has since improved.44 A 2020 study of five commercial systems found an average word error rate of 0.35 for Black speakers and 0.19 for white speakers.45

Both are older than today's tools and neither tested call notes, so treat them as reasons to test. Keep the audio beside the transcript, have a person listen before any case opens, and compare error rates across your own reps.

Which AI rules reach a call note reader

Treat any claim that such a tool is certified with suspicion. GxP (the good-practice rules for regulated manufacturing, laboratory and clinical work) and 21 CFR Part 11 (FDA's rule on electronic records and signatures) cannot be certified for a product, and a vendor can only offer controls that support your own compliance. Sample records under 21 CFR 203 have their own rules.11 Ask instead what is logged, who can change it and how you would show it to an examiner.

Samples, word lists and a monthly review: how field compliance teams read call notes now

In shortA small compliance team pulls a sample, reads it weeks later, chases the rep or the manager for missing facts, and refers what it finds. The longest single wait comes before anyone reads. Four ways of working exist, and many companies combine them.

From rep to investigator: who handles a call note

Sample in house, list in the CRM, review software or outside reviewers: four ways to monitor call notes

Why many drug makers limited free-text call notes, and why that is being reopened

Not every company has free-text notes to read. A CRM vendor's March 2026 guide for legal teams says that from the mid-2000s, legal and compliance leaders limited free-text fields because notes "could be used as evidence" in enforcement and litigation, and because fewer free-text records meant fewer to review. It says the result was notes kept in personal apps, calendars and email, which remain discoverable and are harder to monitor.3

The vendor sells a tool for this, so weigh its account accordingly. We found no published count of how many companies allow free text. The choice belongs to your legal team, and it now turns on whether you can read everything that gets written.

Sam Patel's dinner offer: from one highlighted sentence to a case in front of the investigators

In shortThis is the example from the top of the page, inside the product. Agents read every note, overnight or as it arrives. Lena Ortiz, the field compliance lead, opens Sam Patel's voice note on the morning of October 7 and sees the dinner highlighted, the four facts filled in and 4 hours 15 minutes left. She decides where each finding goes.

1

Every note read before anyone starts work

The nightly run finished at 2:41 a.m., and notes filed after it, such as Sam's at 8:05 a.m., are read as they arrive. The list counts 1,284 of 1,284 notes read, typed and spoken. Each row shows what the agents found, its stage and the clock from the call. Findings sort to the top, with the nearest deadline first.

Lena does not choose a sample. Ten notes need her decision today.

2

The note itself is the screen

CN-40817 is Sam's 1:48 voice note about Dr. Elena Ruiz in Denver. Each phrase that matters is marked in the rep's own words, in its color: red for a possible inducement, violet for an off-label request, amber for promotional language, blue for competitor and access remarks.

One click on a highlight opens its finding on the right. The words are never changed. Sentence numbers tie each finding to its source.

3

The clock runs from the call

The call was Tuesday at 2:20 p.m. The note was filed Wednesday at 8:05 a.m., 17 hours and 45 minutes later. Under this sample company's procedure the deadline is a business day from the call, so it falls at 2:20 p.m. Wednesday, whatever time the note came in. At 10:05 a.m., 4 hours and 15 minutes remain.

Offered by, offered to, what of value, in return for: four of four, each cited to the call report or to a sentence. Had one been missing, the agents would have drafted a question to Sam for Lena to send.

4

The old keyword list, on the same note

One button runs the comparison. The list missed the dinner, because no listed word is in the sentence. It also missed the 15-year-old question and the comparative claim, and flagged "pediatric" in the practice's name, Aldervale Pediatric & Adult Dermatology.

This helps at a pilot, when someone asks what the old method would have done with your own notes.

5

The 15-year-old question goes to Medical Information

Dr. Ruiz asked, unprompted, whether she could start Veltrimab in a 15-year-old. The agents mark the request as unsolicited, citing sentence 5, and mark Sam's reply as correct, citing sentence 6. They set the physician's question out as the request will read, with the sample label passages.

Omar Haddad's Medical Information team answers in writing, by email, as Dr. Ruiz agreed. Sam needs no coaching for this one.

6

A comparative claim, with the approved line beside it

Sam said Veltrimab "clears skin faster than Zorelta". The agents find no head-to-head study in the label or approved material, cite the sample company's promotional standard, and offer the approved sentence from the company's sales material, its detail aid. They also note that this is Sam's second language flag in 30 days.

Lena chooses to coach, escalate or dismiss. Dismissing needs a reason. Coaching goes to Sam's district manager, Priya Ranvelle, with the sentence, the rule and the approved line.

Weekly staff lunches for five patients: a 38-second voice note read as it is transcribed

In shortAt 10:04 a.m., a new voice note arrives from Jordan Lee in Atlanta. As the words appear, an offer turns red and the clock starts from the 9:15 a.m. call. Five agents finish in 4.1 seconds on the sample note. Lena sends the case to the investigators two minutes after it arrives, with 23 hours and 9 minutes to spare.

1

A note waits at the top of the list

Jordan Lee called on Dr. Nia Robinson at Copperleaf Skin & Laser at 9:15 a.m. and dictated a note 49 minutes later. It sits at the top of the list, marked New. Nothing has been found yet, and no clock is running. If the note holds value offered to a prescriber, the clock will start from the call rather than from now.

2

The clock starts when the words appear

Fifteen seconds in, the transcript reaches "bring lunch for her whole staff every Friday once she has five patients on Veltrimab". The phrase turns red and the four facts fill in. The clock starts at 23 hours 10 minutes, counted from the call, and the route to investigators appears.

Nothing goes anywhere yet. The panel says so: nothing is sent without you.

3

The whole note, read

The other findings follow: "double the dose for patients who flare" as an off-label question, "still start on Zorelta because of its copay card" as a competitor mention, and "works faster than anything else out there" as promotional language. The note records no reply to the dose question, so that finding shows the rep's response as not recorded.

4

Lena sends it; the case carries the call date

At 10:06 a.m., Lena presses Send to Compliance investigations. Case CI-26-0440 opens with the call date carried over, her name and the time on record. That night's reconciliation will confirm the case exists and the date matches. The off-label question and the promotional language wait for her decision.

A lunch on a personal card, referrals by destination, nudges for the rep and what doctors say about access

In shortThe other screens follow one note's findings to their destinations, show the rep what will go to Compliance before submitting, and turn the rest of the notes into de-identified insights. A dashboard reports coverage and speed, and settings hold your own rules.

A gift or meal referral with one fact missing

The keyword list, side by side, across every note

Referrals by destination, reconciled every night

What the rep sees while dictating

Access barriers and competitor remarks, with patients' details left out

Coverage and speed on one dashboard, and rules in your own terms

Design targets for hours from the call to Compliance

These are design targets and estimates carried in the solution's sample data, to be measured in your own pilot: 100% of notes read, against an estimated 5% before; a median of 6.6 hours from the call to a referral, against an estimated six days; and 38 false alarms a month to clear, against 1,212 on the sample keyword list.

Six agents read and route; the field compliance lead decides each one

In shortEach agent has one job and a written limit. Together they read, cite and propose. A named person decides each route, nothing about a possible inducement closes without a person, and every dismissal goes to the investigations lead for a weekly second look.

Lena decides, Grace takes a second look, Omar answers, Priya coaches

The agents work from the words in the note. A person's decision protects the rep as well as the company: a flag becomes a finding only after a person has decided, with the audio and the sentence in front of them.

CRM notes in; cited escalations, spend records and Medical Information requests out

In shortThe solution connects to the systems field and compliance teams already run, reads the notes where they live and writes each decision back. It is built for large field forces, with roles, maker-checker and an audit trail. Your rules are settings, and they travel with the record of who changed them.

Case, spend and Medical Information systems: where the routes end

Escalations go to the compliance case system, meals to the spend system and questions to the Medical Information system. Acknowledgments come back, and each night every referral is matched to its case or record. A person can also ask about any note, clock or referral in plain words, and the answer points back to the sentence.

Built for commercial compliance at field force scale: roles, records and security

What an unread call note costs you: settlements at one end, monitor hours and false alarms at the other

In shortYour largest exposure is rare and very expensive, and the steady cost is reviewer time spent on the wrong notes. Public figures frame the first. Nobody publishes the second, so count it yourself in a week.

Seven numbers to find in a week: your hours, your delay and your misses

The offer no one read, and other costs a monitoring calculator misses

What reading every call note returns when you count monitor hours against its yearly cost

In shortPut in your own numbers. The defaults are assumptions, marked as such. The calculator values reviewer time only: reading, clearing false alarms and chasing facts. It leaves out the main reason to read every note, the case you catch early, because no honest formula prices that.

False alarms, fact-chasing and coverage: where call note monitoring pays

Your hours back across a year of 320,000 call notes

How ready your field compliance team is: notes, policies, clocks, owners and tests

In shortPick the answer closest to your team today. Eight questions score 0 to 2, and the last is for reflection. The meter adds up as you go, and each question says why it matters and what a strong answer sounds like.

What a score of 0–6, 7–11 or 12–16 asks of your field compliance team

What to ask any call note monitoring vendor about recall, records, voice and reps' rights, including us

In shortA good answer comes with a number measured on your notes, or a screen you can test. Ask to see each of these working. Then ask us the same questions.

Testing call note review on past notes in ten weeks, with planted offers and a go or no-go

In shortStart with past notes, so nothing is at stake while you measure. Read a sample by hand as your answer key, plant some offers, run the software, and compare. Move to live notes only when your team and counsel are happy with the numbers.

Catch rate, wrong flags and hours from the call: what your pilot should report

Pharma call note compliance: the questions people search for, answered

In shortFourteen short answers. Each stands alone, so you can send one to a colleague.