To AI or Not to AI – That is the Wrong Question

Jon Wilkenfeld
May 8, 2024

Hamlet-Inspired Rationale for the Thoughtful Use of AI

The Increasing Prominence of AI for Compliance Professionals

If you took a drink every time you heard the word “AI” at the recent Pharmaceutical Compliance Conference (PCC), you’d have been drunk before 10 AM. It seemed as if every presentation mentioned “AI”, including several exclusively focused on the topic.

“To AI or Not to AI?”

Image by OpenAI’s DALL-E, generated 5/2/2024

Many of the discussions centered on whether companies should use AI.

In one corner were those emphasizing the impact AI can have in streamlining manual tasks and advancing data analytics. Oversimplified, no need to hire, AI can do that.

In the other corner were those claiming that AI can never (or at least not in the near future) replace the human component of compliance programs.

The naysayers added that AI creates a myriad of new risks for the organization including access to confidential information, privacy concerns, and potential for error and put simply, needs to be approached with extreme caution.

With apologies to the Bard, I believe “To AI or Not to AI”? is the wrong question. Like the debate on rules vs. values-based compliance programs, the answer is both.

Generative AI is here to stay. And for any who refuse to approach it head on, I cite Hamlet’s Queen Gertrude — “the lady doth protest too much, methinks.”[1] Let us not be like Hamlet and allow our conscience to make cowards of us all. Instead, let’s focus on three critical questions:

  1. What is AI best at?
  2. How do we use generative AI effectively?
  3. When should we not use AI?

I conversed with ChatGPT to flesh out these questions and found it immensely helpful. All text in blue was drafted by ChatGPT (often edited for brevity). The subsequent Hamlet quotes were largely identified by ChatGPT — including one from Henry IV. More on that later…

[1] Please note, gendered language used throughout this piece is specifically from Hamlet, written in ~1600).

What is AI Best at?

Automation of Repetitive Tasks

“Though this be madness, yet there is method in’t.” – Polonius

First and foremost, AI — and technology more broadly — should be used to automate manual processes. At Potomac, we often talk about the importance of sharpening your saw as conveyed in the 7 Habits of Highly Effective People. Undo the madness of performing the manual activity, identify the consistent method to it, and automate it. Technology is great at this.

There are many use cases for compliance professionals. Some are easily identifiable and have been available for years, such as aggregate monitoring and automated sanction screening. Others are newer, such as using tools like Microsoft Copilot to reduce formatting time. I cringe watching transparency professionals look up HCPs one at a time in the NPPES registry (if you do this, please, please reach out to me to stop that madness).

And as natural language processing continues to improve, there will be a growing host of automatable functions. For example, highly educated (and compensated) human beings in the Medical Affairs function frequently review doctors’ résumés to “tier” them as part of a “fair market value” assessment. In the future, this first pass could and should entirely be handled by machines.

Enhancing Creativity

“There are more things in heaven and earth, Horatio, than are dreamt of in your philosophy.” – Hamlet

The potential of generative AI in ethics and compliance transcends our current understanding, offering opportunities for innovation and advancement that may surpass our wildest imaginations.

Generative AI is exceptional at expanding initial idea generation. In Hamlet, the “play’s the thing.” When dealing with AI, you could say “the prompt is the thing.” The key is knowing where and how to prompt the discussion. Because Generative AI is best when building on existing creative ideas. It becomes an instant group brainstorming session. If you have an inkling of an idea, ask the AI for more. See what it says, and then ask for more details. It’s like having colleagues working with you on a whiteboard.

I don’t know about “surpassing our wildest imagination.” Maybe I haven’t mastered prompts. But in building this piece, I initially asked for Shakespeare quotes to help support this piece. When it provided an example from Hamlet, I pivoted and asked for more quotes from Hamlet and how to tie those quotes into my central theme.

If Hamlet is not your thing, ask generative AI for lyrics from Taylor Swift songs that relate to your latest Code of Conduct training. Or famous movie quotes…or…. Like any good brainstorming session, you’ll get good ideas and bad ideas; the key is to zone in on what you want to keep.

Motivation

“Something is rotten in the state of Denmark.” – Marcellus

I’m feeling a little “unmotivated” to write this one.

I submitted the following prompt to ChatGPT, “Can you tie the concept of motivation to starting something new with the quote “Something is rotten in the state of Denmark.” Aiming for something centered around how hard it is to get going when you are feeling “rotten” or unmotivated.“ (Note: I have edited the text below for length by deleting redundant paragraphs).

“Something is rotten in the state of Denmark” serves as a poignant metaphor for the internal state of disarray or discontent that can hinder our motivation to embark on new endeavors. Just as the characters in Hamlet confront the pervasive sense of corruption within their society, individuals grappling with feelings of disillusionment or stagnation may find it challenging to summon the motivation needed to initiate change or pursue new opportunities.

However, just as Hamlet ultimately confronts the corruption within his kingdom and takes action to restore order, individuals can overcome their internal barriers and find the motivation to initiate change.

Staring at a blank sheet of paper is hard. Generative AI can be a useful starting point to get a first draft moving. It need not cover everything but can help overcome that initial inertia. Especially when you are feeling unmotivated to tackle the corruption within your kingdom.

How do We Use Generative AI Effectively?

Validate Independently

“Doubt thou the stars are fire, Doubt that the sun doth move, Doubt truth to be a liar, But never doubt I love.” – Hamlet, as read by Polonius

“Here’s a list of 30 additional Hamlet quotes… 26. “The better part of valour is discretion.”

To which I immediately asked ChatGPT where I could find that quote in Hamlet. It responded:

“The better part of valour is discretion” is a famous quote from Shakespeare’s play “Henry IV, Part 1,” spoken by Falstaff in Act 5, Scene 4.

Image by OpenAI’s DALL-E, generated 5/7/2024

For those looking to use Generative AI to help provide compliance answers to complicated questions via chatbots and similar tools, I have serious concerns. While natural language processing is improving, there is still a tendency by GPTs to “hallucinate” and guess at responses even when facts are straightforward and easily verifiable. It’s obviously harder when different words are used in the questions than those specifically stated in the policy or training are fed into the system. There is a real risk of providing inaccurate guidance.

Others have covered this topic much better than I can. Maybe these hallucinations are the ghosts of Hamlet’s father. Over time, I imagine AI hallucinations will improve, potentially immensely. But right now, answers unchecked by human eyes can be troubling. “Doubt truth to be a liar.”

Edit for Clarity and Conciseness

“Brevity is the soul of wit.” – Polonius

Polonius’ wisdom reminds us of the importance of clarity and succinctness in communication.

GenAI ≠ “Clarity and succinctness.”

A friend and colleague of mine (who inspired this article) mentioned at PCC how much bad writing we’ve fed into the AI engine. It has not learned to be concise. Edit the output.

Organizational Preparation

“The readiness is all.” – Hamlet

ChatGPT did a good job with this one. I’ll let it explain:

Hamlet’s profound realization before the climactic duel with Laertes speaks to the importance of preparedness. When it comes to automating repetitive tasks with AI, readiness in terms of data preparation, infrastructure setup, and organizational alignment is crucial for successful implementation.

For those looking to use AI to develop risk-based insights, I’ll remind everyone of the importance of data quality: garbage in, garbage out.

And there must be organizational alignment on how and when to use AI securely to prevent loss of confidential information, protect privacy, and be net beneficial to compliance efforts.

When Should We Not Use AI?

People and Culture

“This above all: to thine own self be true.” – Polonius

Polonius encourages Laertes to prioritize authenticity and self-awareness…to maintain one’s integrity and moral compass…

Human interaction plays a crucial role in fostering trust and building relationships, especially in compliance investigations and audits….

Ethical dilemmas and complex compliance decisions require human expertise to consider factors such as intent, mitigating circumstances, and ethical implications…

Humans are better equipped to communicate findings, address concerns, and collaborate with stakeholders…

Human oversight is essential to ensure that AI systems are deployed ethically, responsibly, and in alignment with organizational values and regulatory requirements.

Our role as compliance professionals must expand with AI. We need to shift our own focus from the monotony of following the black and white “rules” towards the gray area of “values.” We need to step up our game and show thoughtfulness in evaluating complex scenarios. Only humans will be able to solve new and evolving compliance concerns that cannot be fed into the algorithms.

The job for building and enforcing the company’s culture will remain the job of the Compliance Officer. At the heart of that is our individual relationships with people. AI will not replace that.