.png)
The Future is Expert Led, AI-Enabled
- Nicole Nezat
- May 13
- 6 min read
Introduction: The Promise of AI
Artificial intelligence is, hands down, the hottest topic in healthcare and commercial organizations today. The potential of AI tools to accelerate or even revolutionize entire sectors of business have every department talking about how and where to leverage these new tools. This excitement is understandable. AI tools improve rapidly, generate ideas instantly, summarize large amounts of information, and create the impression of near-limitless capability.
Many organizations are already integrating AI into day-to-day operations and those that aren’t, are not far behind. Commercial pharma teams are exploring AI-supported planning, content creation, customer engagement, and operational workflows.
At the same time, most of us have experienced the limitations of generative AI and language-learning models firsthand. AI hallucinations, misleading statements, materials that sound nice but are seriously lacking in content, and subtle but significant changes to existing materials all make these tools less independent than the initial promise.
AI has the potential to reduce workload and automate certain tasks, but it cannot replace expertise. In many cases, it increases the value of expertise because teams still need people who can evaluate outputs, identify weak reasoning, apply context, and refine ideas into something strategically useful.
Because of this, human expertise is now required in two areas: subject matter expertise and AI fluency. Unsurprisingly, the organizations creating the most value from AI are combining both.
Real success with AI comes from putting powerful tools in the hands of people who know how to think critically, apply expertise, and direct those tools effectively.
The Situation: The AI Illusion
Challenges with AI
AI feels authoritative, fast, and limitless.
AI is getting a lot of attention today, and for good reason. Many executives see this boon in technology as a “gold rush”, eager to transform their business or risk being left behind. The urgency that goes along with this excitement, however, creates significant risk. On the surface, AI tools are perfect, full of limitless potential and AI agents that never say ‘no’. But underneath this perfect exterior is a high-potential tool with a lot of red flags.
Authority & Accuracy
Large Language Models (LLMs) are trained to sound authoritative even when there is a lack of information to form a conclusion. Language choice is geared to make content sound deep and comprehensive but, upon closer examination, falls far short of the depth and rigor expected of an expert.
In fact, in a 2025 clinical decision-making study, clinician accuracy dropped from 66% to 56% when inaccurate AI recommendations were introduced, demonstrating how easily flawed AI outputs can influence human judgment.
Hallucinations
Hallucinations remain one of the clearest examples of this challenge. Even as AI systems improve rapidly, confident inaccuracies still persist across major platforms. Recent testing found that newer systems such as ChatGPT-5 hallucinate less frequently than prior versions, while inaccuracies still remain across major AI platforms.
Lack of Substance
AI also struggles with depth, nuance, and sustained refinement. It can generate first drafts quickly, but maintaining strategic coherence across multiple rounds of iteration often requires substantial human intervention. This becomes especially visible in healthcare and pharmaceutical organizations where messaging, compliance, stakeholder dynamics, and business priorities are deeply interconnected.
Linguistic Red Flags
AI copy often feels… odd. Linguistic patterns that may occur occasionally in human writing are overly-present in AI content. Part of what makes AI sound so compelling is this inspirational, TED Talk-esque feel to the material it creates, like every thought is profound. A recent study of AI editing found that AI editors made more edits than human editors, and increased emotional language use significantly.
Inconsistent Outputs
Prompt Engineering, the customization of your AI requests and the parameters that guide them, is a critical skill for anyone navigating our new AI-enabled reality. However, even the perfect prompt can be ignored by the AI tool. This is frustrating when creating a new project with a contained output, say, a project proposal you want to bring to your financial team, but absolutely unacceptable in the world of embedded AI assistants.
Many teams experience the same pattern. The first draft arrives quickly, while the refinement process takes much longer than expected. Human experts still need to validate information, sharpen logic, restructure ideas, remove low-value content, and ensure the output aligns with real-world business objectives.
The gap between fast output and high-quality output remains significant.
The Need for Human Expertise and Governance
Human expertise remains essential because AI still requires oversight, interpretation, and validation. Experts are needed to verify statements, identify hallucinations, evaluate logic, and shape outputs into something meaningful and strategically useful. Without that layer of review, organizations risk making decisions based on inaccurate or low-value information.
This becomes especially important in healthcare and life sciences environments where trust, compliance, and precision directly influence business outcomes. The Healthcare AI 2025 Global Practice Guide highlights ongoing concerns around governance, liability, oversight, and accountability in healthcare AI implementation.
Organizations also need people who can sift through AI-generated noise and identify misleading conclusions, weak strategic thinking, and unnecessary complexity introduced during content generation. Low-value AI content damages trust quickly.
Even factually correct outputs can undermine business goals when they become overly generic, repetitive, convoluted, or filled with recognizable AI artifacts. Customers and stakeholders notice these weaknesses faster than many organizations expect. AI can accelerate output generation, but human expertise is critical to determine whether that output creates value.
The Solution: Expert Leadership with AI Enablement
The Winning Model: Expert Led, AI-Enabled
The strongest organizations are approaching AI with a clear operating model: expert-led, AI-enabled.
AI accelerates information synthesis, drafting, and workflow support. Human expertise provides judgment, strategic refinement, contextual understanding, and decision-making discipline. This distinction matters because AI does not independently produce business judgment. It generates outputs based on patterns learned from existing information. Human experts determine whether those outputs are accurate, relevant, compliant, and strategically useful.
Commercial pharma organizations are already beginning to structure AI adoption this way. The MIT Technology Review Insights and Globant report on agentic AI highlights growing interest in AI-enabled commercial workflows that still rely heavily on human oversight and orchestration.
Organizations that treat AI as a replacement for expertise often struggle with inconsistency, weak strategic alignment, and declining output quality. Organizations that combine AI with strong expert leadership create scalable advantage.
Human // AI Integration
AI performance depends heavily on how it is directed. To make the most of the AI tools available, one must be an expert on the subject matter at hand, understand how AI functions and the advantages & limitations of each tool, and maintain a critical thinking lens on all outputs generated – no data point or statement can go unchecked.
Prompting quality influences output quality. Iteration quality influences refinement quality. Evaluation quality determines whether outputs create business value or operational risk. Many organizations underestimate this capability gap. Teams are often given access to AI tools without structured guidance on how to extract high-value outputs consistently.
The difference becomes obvious quickly. Some individuals produce strong outputs because they understand how to structure prompts effectively, guide reasoning, refine iterations strategically, and evaluate outputs critically. Others struggle to move beyond generic or low-value outputs despite using the same tools.
AI capability is not simply technical training. Organizations need to intentionally develop critical thinking, evaluation capability, prompting strategy, judgment, and iterative refinement skills.
Without those capabilities, organizations struggle to consistently generate meaningful value from AI investments.
In Action: Maximizing Efficiency and Impact
What Leaders Should Do Now
Organizations that want meaningful AI impact should focus on capability building as much as technology deployment.
The first priority is investing in dual capability: domain expertise and AI fluency.
Strong subject matter expertise remains essential because teams still need to evaluate outputs, apply judgment, and connect information to real business situations. AI fluency becomes equally important because teams must understand how to direct, refine, and evaluate AI effectively.
Leaders also need to clearly define where AI augments workflows and where expertise continues to lead decision-making. Organizations should build workflows where AI supports critical thinking rather than replacing it. Teams should validate outputs, challenge assumptions, refine logic, and evaluate strategic fit consistently.
The organizations seeing the strongest results are doing four things consistently:
Investing in both domain expertise and AI fluency
Defining where AI supports workflows versus where expertise leads decisions
Building workflows that reinforce critical thinking
Setting expectations that AI amplifies expertise rather than replaces it
AI should be positioned as a multiplier that strengthens strong thinking and accelerates high-quality workflows. Organizations that position AI as a substitute for expertise often create inconsistent execution, weak strategic alignment, and declining output quality over time.
Closing: A More Demanding Standard
AI is raising the standard for what good work looks like.
Basic information access is becoming universal. Generic content generation is becoming easier for every organization. Surface-level outputs will become increasingly difficult to differentiate.
Advantage will come from organizations that can think better, refine better, and apply expertise more effectively.
The strongest organizations will combine deep subject matter expertise, strong AI fluency, effective governance, critical thinking capability, strategic judgment, and operational discipline.
Those capabilities will shape how effectively teams interpret information, refine ideas, engage customers, and make decisions in increasingly complex environments.
The future will favor organizations that can guide AI intelligently rather than simply deploy it broadly.
That creates an important leadership question moving forward:
Are your teams equipped to lead AI, or are they simply using it?
Comments