AI Meeting Notes and Transcripts in GxP Decisions: Are They Records?

Where AI Note-Takers Already Show Up in Quality Work

Most pharma and biotech companies did not decide to put AI into their quality meetings. The feature arrived inside tools they already used. A video platform added a summary button. A calendar integration let a third-party note-taker join any meeting its user attends. A quality lead found that the transcript helped her write minutes faster, and her team copied the habit. Within a year, AI-generated text is part of how GxP decisions get documented, and nobody wrote a procedure for it.

This is not a reason to ban the tools. A good transcript can make minutes more complete, help a team member who missed the call, and give an investigator a better view of how a decision was reached. The problem is that record status, retention, accuracy, and confidentiality have been left to default settings. Defaults were designed for general business use, not for regulated decisions.

Five Meetings That Make GxP Decisions

Not every meeting in a regulated company is a GxP meeting. A budget review, an IT roadmap session, or a hiring discussion is ordinary business. The meetings that matter here are the ones where people with quality responsibility make, approve, or change a decision that a GxP rule requires a company to document. Five come up again and again.

Clinical and safety teams have their own versions of these meetings: study team decisions on protocol deviations, safety review committee discussions, and signal evaluation meetings. The MHRA guidance discussed below is written for all of GXP, including clinical and pharmacovigilance work, so the same reasoning applies there.²

Why the Tool Arrived Before the Policy

AI meeting tools spread through individual choice, not system projects. No one submits a change request to turn on a summary button. The tools are often licensed as part of a collaboration suite that IT already manages, so they never pass through the computerized system inventory that quality maintains. And the output looks like a draft, which makes it easy to assume it does not count. That assumption is the one this article tests.

When a Transcript or Summary Becomes a GxP Record

Two regulators have written the definitions that matter most for this question. Neither definition mentions AI, meetings, or note-takers. FDA’s data integrity guidance dates from December 2018, and MHRA’s from March 2018. Both still describe how inspectors think about records, and both apply to any data a company creates, whatever tool creates it.

The FDA Test: Generated to Satisfy a CGMP Requirement

FDA’s guidance on data integrity and drug CGMP asks directly when electronic data becomes a CGMP record. Its answer is short: “When generated to satisfy a CGMP requirement, all data become a CGMP record.”¹ The same guidance defines metadata as “the contextual information required to understand data” and says data should be kept with “all associated metadata required to reconstruct the CGMP activity.”¹

For a meeting, the CGMP requirement is usually not “hold a meeting.” It is “record and justify the deviation,” “make a written record of the investigation,” or “have the quality unit review and approve the change.” If the AI summary of a deviation triage call is the document that records why the deviation was classified as minor, it was generated to satisfy that requirement, and it is a CGMP record. If the summary is a personal reminder and the classification rationale is written into the deviation record in the eQMS, the summary is probably not.

A footnote in the same guidance matters even more for meeting artifacts. FDA states that it “routinely requests and reviews records not intended to satisfy a CGMP requirement but which nonetheless contain CGMP information (e.g., shipping or other records that may be used to reconstruct an activity).”¹ A transcript of a CAPA board debate contains CGMP information whether or not anyone intended it to be a record. If it exists when an inspector asks how a decision was reached, it can be requested.

The MHRA Test: First Capture and Full Reconstruction

MHRA’s GXP data integrity guidance defines data to include original records, true copies, and “all subsequent transformations and reports” that allow full reconstruction and evaluation of the GXP activity.² It defines raw data as “the original record (data) which can be described as the first-capture of information, whether recorded on paper or electronically.”²

Read with meetings in mind, these definitions point in an uncomfortable direction. The recording or transcript is the first capture of what was said. The AI summary is a transformation of it. The minutes are a report built from both. If the minutes are the only thing kept, and they leave out the reasoning that an inspector later needs, a company may struggle to show full reconstruction of the decision.

MHRA also says that “Metadata form an integral part of the original record. Without the context provided by metadata the data has no meaning.”² For a transcript, the metadata is who spoke each line, when, in which meeting, with which attendees, and which tool and version produced the text. A transcript stripped of speaker labels and timestamps loses most of its value as evidence.

Business Practice Decides as Much as Intent

FDA’s guidance on the scope of 21 CFR Part 11 (FDA’s rule for electronic records and electronic signatures) adds a practical point. It notes that if a company relies on an electronic record to perform regulated activities, FDA may treat that electronic record as the record even when a paper version also exists, and that “the Agency may take your business practices into account in determining whether part 11 applies.”³ It then recommends that companies decide in advance, for each record required by predicate rules, which version they will rely on, and document that decision in a procedure.³

This is the heart of the meeting problem. A company may write a procedure saying the approved minutes are the record. But if quality staff routinely open the AI summary to check what was agreed, forward it to a site that needs to act, or paste it into an investigation report, the company’s practice says the summary is being relied on. What people do with the artifact matters as much as what the procedure says it is.

A Four-Question Test for Meeting Artifacts

Putting the FDA and MHRA definitions together gives a short test any quality team can apply to a meeting artifact.

Four Meeting Artifacts, Four Different Answers

Much of the confusion in this area comes from treating “the AI notes” as one thing. A typical AI meeting tool produces up to four distinct artifacts, and each one has a different relationship to the decision. Policy works best when it names each artifact and says what happens to it.

The Recording

A recording is the closest thing to a complete record of a meeting. It captures tone, hesitation, and the moment someone raised a concern that nobody followed up. That completeness is also its burden. Recordings are large, slow to review, and full of material nobody intended to keep: side conversations, comments about colleagues, discussion of unrelated products. For most GxP decision meetings, a recording is not needed as a record if the minutes are good. It may be useful as a short-term aid for the minute-taker. The question is how long it lives after the minutes are approved.

The Raw Transcript

The transcript is the artifact most likely to be treated as “what was said.” It reads as verbatim, which makes people trust it more than they should. Speech recognition makes errors with technical vocabulary, product codes, lot numbers, and names, and it assigns lines to the wrong speaker when people talk over each other. The research discussed later also shows that some transcription models occasionally add whole sentences that nobody spoke.⁹ A transcript can still be valuable. It just cannot be assumed to be accurate.

The AI Summary

The summary is the artifact people use most, because it is short. It is also the furthest from what happened. A summary model chooses what to include, how to group points, and how to phrase a conclusion. “The team discussed whether the excursion affected lots A and B” can become “The team agreed the excursion did not affect lots A and B.” That small shift turns a discussion into a decision. A summary should be treated as a first draft of minutes, never as minutes.

The Approved Minutes

Minutes are the artifact a GxP process is built to control. They are authored, reviewed, corrected, and approved by named people, and they can be stored in a validated system with an audit trail and a defined retention period. Using AI to draft minutes is reasonable. Using AI output as minutes, without review and approval, is where most of the risk is.

The Metadata That Gives Each One Meaning

Whichever artifact becomes the record, its metadata has to come with it. For a GxP meeting record, the useful metadata usually includes:

• Meeting identity: title, date, start and end time, time zone, and the GxP record it relates to (deviation number, change number, CAPA number).
• Attendees and their roles, including who held decision authority and who attended as a subject matter expert.
• For a transcript: speaker labels, timestamps, and a note of any segments where recording was paused.
• For AI-drafted text: the tool and version used, and a statement that the text was AI-drafted and then reviewed.
• For minutes: author, reviewer, approver, approval dates, and version history.

MHRA’s point that data without metadata has no meaning applies directly.² A summary pasted into an email loses the link to the meeting, the attendee list, and the tool that produced it. A summary attached to the deviation record in the eQMS, with a reference to the meeting and an approver, keeps them.

What Record Status Means: Retention, Review, Approval, and Audit Trail

Once an artifact is a GxP record, four obligations follow. None of them is new. They are the same obligations that apply to a paper deviation form or a signed change request. The work is in applying them to a file that a collaboration tool created.

Retention Follows the Decision, Not the Tool

A meeting record takes the retention period of the GxP record it supports. Minutes of a deviation triage that classified a deviation on a batch follow the batch retention rules: under 21 CFR 211.180(a), records associated with a batch are kept for at least one year after the batch expiration date.⁶ EU GMP Chapter 4 sets batch documentation retention at one year after expiry of the batch or at least five years after certification by the Qualified Person, whichever is longer.⁷ Records must also be “readily available for authorized inspection during the retention period,” which 211.180(c) allows to be met by electronic retrieval from another location.⁶

Collaboration tools do not know any of this. Microsoft’s documentation for Teams states that meeting recordings and transcripts have a default expiration time of 120 days, after which OneDrive and SharePoint move them to the recycle bin, and that recordings of non-channel meetings are stored in the organizer’s OneDrive.¹² An admin can change the default or turn expiration off, but changes affect only newly created recordings and transcripts.¹² Other platforms have their own defaults. The point is the same: if a transcript is a GxP record, its retention cannot be left to a storage-management setting in someone’s personal folder.

EU GMP Chapter 4 adds a requirement that suits this problem well: “It should be clearly defined which record is related to each manufacturing activity and where this record is located.”⁷ A procedure that names the approved minutes as the record, and names the eQMS record where they are filed, meets that expectation. A situation where the rationale is spread across three people’s AI summaries in three personal storage locations does not.

Accuracy Review Before Anyone Relies On It

FDA’s data integrity guidance lists among the CGMP data requirements that output from computer systems “be checked for accuracy,” citing 211.68.¹ MHRA expects a procedure for data review and approval, and says that the review record should include “a positive statement regarding whether issues were found or not, the date that review was performed and the signature of the reviewer.”²

For AI-drafted minutes, a practical accuracy review checks four things: the decision is stated as the decision-maker understood it, the rationale is complete, every action has an owner and date, and any disagreement is recorded. It is not a proofread. A reviewer who only fixes grammar has not reviewed the record. The UC Davis study of AI-drafted clinical notes found that 14.9% of notes in its editing data were left entirely unedited, and physician editing varied widely from person to person.¹¹ That was a clinical setting rather than a quality meeting, but the behavior is familiar: when a draft looks finished, people tend to accept it.

Approval by the People Accountable for the Decision

The GxP rules say who must approve the underlying decisions. Changes to written production and process control procedures “shall be drafted, reviewed, and approved by the appropriate organizational units and reviewed and approved by the quality control unit.”⁵ Investigations of discrepancies require a written record that “shall include the conclusions and followup.”⁴ ICH Q10 lists “Documentation and timely and effective communication of the results of the management review and actions” as an outcome of management review.⁸

Meeting minutes that document these decisions should be approved by the chair or the decision owner, with quality approval where the decision needs it. An AI tool cannot approve anything. MHRA’s guidance on scribes offers a useful parallel. Where a second person records an activity on behalf of the person performing it, MHRA says “the records should identify both the person performing the task and the person completing the record.”² An AI note-taker is a kind of scribe. Minutes should show that the text was AI-drafted, who edited it, and who approved it.

An Audit Trail for Corrections

FDA defines an audit trail as “a secure, computer-generated, time-stamped electronic record that allows for reconstruction of the course of events relating to the creation, modification, or deletion of an electronic record.”¹ EU GMP Chapter 4 says any alteration to a document entry should be signed and dated, and “the alteration should permit the reading of the original information.”⁷

Most collaboration tools let any participant with edit rights change an AI summary, and some keep only the current version. That is fine for a draft. It is not fine for a record. Once minutes are approved, they should be in a system that preserves versions and shows who changed what. The usual answer is to finalize minutes in the eQMS or a validated document management system, not in the meeting tool.

Where AI Meeting Records Go Wrong

The risks below are not reasons to stop using AI in meetings. They are the reasons a human review step and a clear policy are needed. Each one has shown up in published research, litigation, or professional guidance.

Transcription Errors and Invented Text

Speech-to-text has improved a great deal, but it is not error-free, and some errors are not the kind people expect. A 2024 study of OpenAI’s Whisper model by Koenecke and colleagues found that “roughly 1% of audio transcriptions contained entire hallucinated phrases or sentences” that did not exist in the audio, and that 38% of those hallucinations included explicit harms such as violent content, inaccurate associations, or implied false authority.⁹ The authors also found that hallucinations occurred more often for speakers with longer pauses and non-vocal periods.⁹

That study tested one model at one point in time, and newer models may perform differently. The lesson for a quality team is not the exact rate. It is that a transcript can contain a sentence nobody said, and the sentence can read naturally. In a deviation triage meeting, an invented clause such as “no impact to product” is not a harmless typo. Long pauses are also common in quality meetings, where people stop to look up a batch record or check a specification.

Ordinary recognition errors matter too. Lot numbers, equipment IDs, product codes, and the names of people and suppliers are exactly the words speech recognition gets wrong most often, and exactly the words a GxP record needs to get right.

Summaries That Drop the Point

Summarization adds a second layer of error on top of transcription. Researchers who built a dataset of 200 AI-generated meeting summaries annotated for nine error types, including structural, omission, and irrelevance errors, wrote that large language models “still struggle to maintain relevance and avoid hallucination” in meeting summarization.¹⁰ Their approach adds a separate step that finds and corrects those errors before a summary is used, which is the same job a human reviewer does for minutes.

The closest real-world evidence comes from healthcare, where AI scribes draft notes from recorded conversations. In the UC Davis pilot, physicians reviewed 356 AI-drafted clinic notes. Accidental omissions were the most frequent error (64 notes, 18%), followed by hallucinations (41 notes, 11.5%) and accidental inclusions (33 notes, 9.3%). Most errors were mild to moderate, but 19 notes (5.3%) contained errors rated as posing serious or imminent risk if not corrected.¹¹ The authors concluded that careful review remains necessary and that organizations should pilot the technology with a review process before deployment.¹¹

A clinic visit is not a change control board, and those figures should not be read as an error rate for quality meetings. But the shape of the finding transfers well. The most common failure was not invented text. It was something important left out. In a GxP decision meeting, the thing most likely to be left out is the condition attached to a decision: “approved, provided the cleaning validation report is signed first.”

Missing Dissent and Flattened Uncertainty

This is the risk that matters most for GxP decisions. Summaries tend to report consensus. When a QA reviewer says “I’m not comfortable closing this without the second lab result,” and the group moves on, a summary may record the outcome (“CAPA approved”) without the objection. Weeks later, when the second lab result comes back out of specification, the record shows a clean approval and no trace that the concern was raised.

That is a data integrity problem as well as a quality culture problem. ALCOA+ expects data to be complete, and MHRA’s definition of complete data is that “the data must be whole; a complete set.”² MHRA also warns, in the context of summary reports passed between organizations, that “summary reports are limited and critical supporting data and metadata may not be included.”² The same caution applies to an AI summary of a meeting. It is a report, and reports leave things out.

Dissent matters for another reason. FDA’s data integrity guidance says management should create a culture where employees “are encouraged to identify and promptly report data integrity issues.”¹ If people learn that objections raised in meetings disappear from the record, they stop raising them in meetings, or start raising them only in private. Minutes that record disagreement and how it was resolved support the culture regulators expect.

Speaker Misattribution

Speaker labels are part of the metadata that makes a transcript attributable, the “A” in ALCOA. They are also one of the weakest parts of current tools, especially when people join from a shared conference room, talk over one another, or dial in by phone. Professional guidance for lawyers on AI note-takers notes that AI transcripts can misattribute speakers, misunderstand content, and fail to capture context or nuance.¹⁴ In a GxP setting, a transcript that attributes the release decision to the wrong person is not a minor flaw. It attaches accountability to someone who did not make the call.

Confidential Data Going to a Vendor

Quality meetings discuss unpublished product data, supplier problems, inspection findings, and sometimes patient-level safety information. An AI note-taker sends that audio to a vendor for processing. MHRA’s guidance on cloud and software-as-a-service providers says companies should pay attention to “understanding the service provided, ownership, retrieval, retention and security of data,” and that the responsibilities of each party should be defined in a contract or technical agreement.²

The questions that matter are concrete. Where is the audio processed and stored? Does the vendor use meeting content to train its models? Who at the vendor can access transcripts? Can the company retrieve and delete everything at the end of the contract? These questions have already reached the courts. A class action filed in August 2025 in the U.S. District Court for the Northern District of California alleges that a widely used note-taker joined meetings and recorded participants without their consent, and used recordings to train its AI; the company has said audio is de-identified before it is used for training.¹³ Those are allegations, not findings, and the case outcome is not settled. The case is still a clear signal that a note-taker connected to one person’s calendar can bring a third party into meetings that other attendees assumed were private.

The risk is greatest with tools that individuals connect on their own. An enterprise collaboration suite under a negotiated agreement is a different situation from a free note-taker that an employee or a visiting contractor links to their calendar. Policy should address both.

Privilege and Discovery Exposure

Quality decisions sometimes involve legal counsel: a potential recall, a field alert decision, a response to an FDA Form 483. Legal commentators have warned that bringing an AI note-taker into these meetings creates risks. Lawyers writing for Columbia Law School’s corporate governance blog note that because AI vendors may have access to transcript data under their terms of use, an opposing party could argue that using the tool disclosed the conversation to a third party, and that “Courts have not yet squarely addressed this question.”¹⁵ An ABA Journal report from May 2026 describes lawyers advising clients to use human note-takers instead, partly because AI tools transcribe jokes, inaccurate statements, and offhand comments that could later surface in discovery.¹⁶

For a pharma or biotech company, the practical point is that a verbatim transcript of a quality debate is a much broader document than the minutes. It may include speculation about root cause that the investigation later disproved, informal comments about a supplier, or a frustrated remark about timelines. None of that makes the decision wrong. All of it can be read later without the context that the people in the room had. Decisions about keeping transcripts should involve legal as well as quality.

Consent and Recording Law

Recording laws vary by jurisdiction. California, for example, requires the consent of all parties to record a confidential communication, including those conducted by telephone or other device, under Penal Code section 632(a).¹⁷ Other U.S. states require only one party’s consent, and many countries outside the U.S. have their own rules, including data protection laws that treat voice recordings as personal data. Quality meetings often include people in several states and countries, plus contractors and supplier staff. A recording policy needs a consent step that works for the strictest jurisdiction involved, and legal should confirm what that step looks like.

Policy Options That Work in Practice

There is no regulatory text that tells a company exactly how to handle AI meeting tools. That leaves room to choose, and the right choice depends on the company’s size, systems, and risk tolerance. The options below are the ones we see working. Most companies need some version of all five.

Option 1: Approved Tools Only

Allow AI transcription and summarization in GxP meetings only through tools the company has assessed and approved. The assessment does not need to be a full computerized system validation for a tool whose output is a draft. It does need to answer the vendor questions above (data location, training use, access, deletion), confirm the tool can be configured to the company’s retention rules, and record the decision. Block or prohibit personal and third-party note-takers that individuals connect to their own calendars, and tell external attendees, including supplier and contractor staff, that their own note-takers are not permitted in the company’s GxP meetings.

If the approved tool’s output does become a record, or is used in a way that makes it one, it needs the level of assurance that use calls for. In most cases it is simpler to design the process so the tool’s output is never the record, which is Option 2.

Option 2: Human-Reviewed Minutes as the Official Record

This is the single most useful decision a company can make. The procedure states that for any meeting that makes a GxP decision, the official record is the approved minutes, filed in the related eQMS record or a controlled document system. AI may draft the minutes. A named person reviews and corrects them, and the chair or decision owner approves them. The minutes state that they were AI-drafted and human-reviewed. Nobody acts on a decision from a raw transcript or an unreviewed summary.

This follows FDA’s Part 11 guidance, which recommends deciding in advance which version of a record will be relied on and documenting that decision in a procedure.³ It also follows EU GMP Chapter 4’s expectation that it be clearly defined which record relates to each activity and where it is located.⁷ And it puts accountability where it belongs, with the people who made the decision.

The procedure has to be backed by practice. If people keep forwarding the AI summary as the decision, the procedure says one thing and the practice says another. Training, and a habit of linking to the approved minutes instead of the summary, close that gap.

Option 3: Transcript and Recording Retention Rules

Once minutes are the record, the company has to decide what happens to the recording, transcript, and AI summary. There are three workable approaches, each with a trade-off.

Whichever approach is chosen, three conditions apply. First, deletion must follow a written rule that applies to everyone, not individual choice. Deleting a transcript after a disputed decision, when similar transcripts are normally kept, looks very different from deleting it under a routine rule. Second, a legal hold must override the rule. Third, the rule must be enforced by configuration where possible, so that tool defaults, such as an automatic expiration period, match the policy rather than working against it.¹²

One point deserves care. FDA has said it is not acceptable to record data on scraps of paper that are discarded after the data are transcribed into a permanent notebook.¹ That statement is about laboratory and production data that must be recorded at the time of performance, not about meeting discussion. But an inspector may ask a similar question about a transcript that was the only contemporaneous capture of a decision and was deleted after someone wrote minutes from it. The best answer is minutes that are drafted during or right after the meeting, reviewed by attendees, and complete enough to stand on their own, together with a documented rule for the working aids.

Option 4: Clear Rules for When to Turn Recording Off

Some meetings, or parts of meetings, should not be recorded or transcribed at all. A short list in the procedure removes the need for judgment calls in the moment:

• Any meeting or segment where legal counsel gives advice, or where a recall, field alert, or regulatory response strategy is discussed, unless legal approves recording.
• Discussions of individual employees, such as performance, training failures, or disciplinary matters tied to a deviation.
• Patient-identifiable information, including safety cases discussed in detail, unless the tool is approved for that data.
• Meetings with external parties who have not consented, or whose contracts do not allow recording.
• Discussions of suspected data falsification or whistleblower reports, which FDA expects to be handled within the documented quality system rather than informally.¹

The chair should announce when recording stops and starts, and the minutes should note that a segment was not recorded and why. A gap in the transcript with no explanation invites the question of what happened during it.

Option 5: Record Dissent on Purpose

Because AI summaries tend to flatten disagreement, the minutes template should have a field for it. A simple section, “Concerns raised and how they were resolved,” prompts the reviewer to look for objections in the transcript and record them. It also makes it normal for minutes to show that a decision was debated. That is a sign of a working quality system, not a weakness.

A Workflow for a GxP Decision Meeting

Policies work when they turn into a routine people can follow without thinking about it. The workflow below is a completed example for a change control board, but the same steps fit deviation triage, CAPA review, and management review.

Step 3 has more effect than any technical control. A chair who says “Decision: approved, conditional on the updated risk assessment being signed by Friday, owner Maria” gives both the AI and the reviewer a clear sentence to capture. The habit improves minutes whether or not AI is in the room, because the decision is stated once, clearly, by the person who owns it.

Step 7 provides the evidence that the process works. MHRA describes periodic audit of the data generated as a way to verify that existing control measures are effective.² A small, regular sample of AI-drafted minutes compared with their source is that review for meeting records. It also tells the company whether the tool is improving or getting worse over time, which matters because vendors can update their models on their own schedule.

What an Inspector May Ask

As AI-drafted text appears in quality records, inspectors’ questions will follow from the guidance above. A company that has made the decisions in this article can answer them in a few minutes. A company that has not may find itself explaining why the rationale for a batch release is in someone’s chat history.

Transcripts Can Be Requested Even If They Are Not Records

FDA’s footnote on records that contain CGMP information is worth repeating to every team that uses AI note-takers.¹ A company cannot avoid producing a transcript by calling it a working aid. If it exists, it can be asked for. That is the strongest practical argument for a clear retention rule: what has been deleted under a routine, documented rule is not available to produce, and what has been kept is kept on purpose, in a controlled place, with its limits known.

Common Mistakes

• Treating the AI summary as minutes. Forwarding it as “what we agreed” without review makes it the record by practice.
• Leaving retention to tool defaults. Records either expire too early or transcripts are kept indefinitely in personal folders.
• No rule for external note-takers. A supplier’s or contractor’s note-taker sends the company’s quality discussion to a vendor the company never assessed.
• Review that only fixes typos. The reviewer corrects spelling but does not check decisions, conditions, or dissent.
• Deleting transcripts case by case. Ad hoc deletion, especially after a disputed decision, is far harder to defend than a routine rule.
• Recording everything. Legal advice, personnel matters, and patient data end up in transcripts that did not need to exist.

Conclusion

The question of when AI meeting transcripts are GxP records has a clear answer under existing rules. A meeting artifact becomes a record when it is created to meet a GxP requirement or when people rely on it for a GxP decision, and FDA and MHRA guidance written years before AI note-takers existed already describe what that status requires: retention for the right period, accuracy review, approval by accountable people, an audit trail, and the metadata that gives the record meaning. What the rules do not do is choose which artifact should carry that weight. That choice belongs to the company, and making it on purpose is the difference between a controlled process and a collection of AI summaries in personal folders. Our view is that approved, human-reviewed minutes should be the record, AI should be the drafting aid, and everything else the tool produces should follow a written rule that quality and legal have agreed.

Sakara Digital works with pharma and biotech organizations putting practical controls around AI in quality operations, from tool assessment and procedure design to the review habits that make AI-drafted records trustworthy. If you are working out how AI note-takers should fit into your deviation, CAPA, change control, or management review meetings and want an independent perspective on where to start, we are happy to have that conversation.

For Further Reading