Artificial intelligence is increasingly being positioned as a solution to one of healthcare’s most persistent challenges: there is simply more information, more complexity, and more work than many teams have the capacity to manage effectively.
For healthcare value analysis professionals, the appeal is understandable. AI can assist with evidence synthesis, organize large volumes of information, identify patterns in utilization or spend, compare documents, and reduce the administrative burden associated with preparing analyses for multidisciplinary review. Work that previously required hours may, in some circumstances, be completed in a fraction of the time. Those efficiencies are meaningful. They are also only part of the value equation.
As healthcare organizations accelerate their adoption of AI, we should be careful not to evaluate these technologies differently from the products, services, and clinical technologies that value analysis evaluates every day. The purchase price of a technology has never been synonymous with its total cost, and efficiency has never been synonymous with value. AI should be held to the same standard.
The Purchase Price Is Only the Beginning
The most visible cost associated with an AI platform is generally straightforward: a subscription, enterprise license, per-user fee, or usage-based charge. That number is easy to place into a budget and equally easy to use when comparing vendors. It is also an incomplete representation of cost.
Usage-based pricing deserves particular attention. Some generative AI services charge for tokens, the units into which text is divided for processing, with separate rates for input and output. Longer documents, repeated requests, and additional rounds of analysis can therefore increase expense. A fixed subscription may bundle usage rather than expose a separate token charge; the contract determines what the organization actually pays. Evaluating a tool on a brief demonstration alone tells us little about its cost across routine committee preparation or a sustained evidence review workload.
Successful implementation may require integration with existing systems, information security review, data governance, legal and regulatory assessment, employee education, workflow redesign, and ongoing validation. Organizations may also need to determine who is responsible for monitoring outputs, managing changes to the technology, and ensuring that appropriate human oversight remains in place. These activities consume both financial and organizational resources.
Employee training also extends beyond an initial demonstration. Staff need time to learn which tasks are appropriate for the tool, how to frame requests, and how to verify the results. That learning period may require additional support and temporarily reduce productivity. As tools and workflows change, education may need to be repeated. These are implementation costs even when they appear as staff time rather than a separate invoice.
Value analysis professionals understand this concept well. A product with a lower acquisition cost may create additional downstream expense through utilization, labor requirements, complications, waste, or operational inefficiency. Conversely, a product with a higher initial cost may generate sufficient clinical or operational benefit to justify the investment. The same principle should apply to artificial intelligence.
The question should not simply be, “What does the AI platform cost?” The more useful question is, “What does it cost the organization to achieve the outcome we expect from the technology?”
Efficiency Is Not Automatically Value
AI discussions frequently focus on time savings, and appropriately so. Time is one of healthcare’s most constrained resources.
Suppose a value analysis professional spends four hours gathering evidence, reviewing documents, organizing information, and preparing an analysis for committee review. An AI-supported workflow reduces that process to ninety minutes. That reduction demonstrates an improvement in efficiency, but establishing the business case requires an assessment of the resources needed to achieve it.
We should account for the software, integration, training, and validation required to achieve that reduction, including any ongoing usage charges. We should also examine whether the workflow has become simpler or whether another application has been added to an already fragmented technology environment.
We also need to establish how the recovered time will be used. If a technology saves several hours of work but those hours are immediately absorbed by other administrative tasks, the organization has improved efficiency, but the measurable impact may remain limited.
If that same capacity allows a value analysis team to evaluate additional initiatives, engage clinicians earlier, examine utilization more deeply, accelerate implementation, reduce unnecessary variation, or address projects that previously remained in a backlog, the organizational impact becomes considerably more significant.
For many healthcare organizations, the strongest AI business case may therefore be less about eliminating labor and more about expanding professional capacity.
Organizational Attitudes Shape the Return
An organization’s existing attitude toward AI can also influence the return on its investment. Enthusiasm may lead a team to accept a favorable demonstration without adequately testing the workflow or accounting for the effort needed to produce it. An assumption that AI has little legitimate role can discourage a useful evaluation before the evidence is considered. Both positions can distort the business case.
These attitudes have practical consequences. A purchased platform may remain underused if staff concerns are not addressed, while pressure to demonstrate rapid adoption may produce activity without measurable improvement. Concerns about accuracy, workload, or professional autonomy deserve examination. Leaders should establish evaluation criteria before the pilot begins, involve the people who will use and review the output, and allow the results to support adoption, modification, or rejection. The time required to build that shared understanding is part of implementation.
AI Must Function Within the Operating Model
The return also depends on how AI functions within the organization’s operating model, connecting the work of value analysis, supply chain, finance, quality, information technology, and clinical operations. A platform may perform an impressive analysis, but an analysis that remains disconnected from the people, data, and workflows responsible for acting on it has limited organizational value.
Consider a system that identifies significant variation in the utilization of a particular product category. The identification of variation is useful, but it is not the outcome. Value is created when the information can be placed into the appropriate clinical and operational context, reviewed with stakeholders, connected to financial and quality data, and translated into an informed decision.
This is where interoperability and workflow design become part of the value proposition. Does the technology complement the organization’s existing systems? Can relevant information move appropriately between platforms? Can stakeholders access and understand the output? Does AI reduce fragmentation, or does it create another location where information must be retrieved, reviewed, and reconciled?
Healthcare has no shortage of technology. Adding another platform is not inherently innovation. The objective should be to improve the way the organization makes decisions.
Human Oversight Belongs in the Cost Model
There is another cost associated with AI that deserves explicit consideration: human oversight. Generative and analytical AI systems can process information rapidly and produce outputs that appear highly credible. They can also omit context, misinterpret information, rely on incomplete data, or produce an answer that is simply incorrect.
That creates a particular concern in value analysis, where decisions can influence patient outcomes, clinician workflow, organizational expense, standardization, and risk.
The appropriate response is not to dismiss AI because it requires validation. Nor should we assume that human review represents an inefficiency that will eventually disappear. Human oversight is part of responsible AI implementation.
Professional oversight requires the ability to evaluate the evidence, recognize missing clinical or operational context, and remain accountable for the resulting decision. The staff time needed to validate outputs, correct errors, and document conclusions belongs in the cost model, including work that must be repeated when an initial output is unusable.
A Value Analysis Approach to AI
Value analysis professionals can contribute to AI adoption by applying the discipline’s established methods to the technology itself.
Before adopting an AI solution, organizations should be able to answer several fundamental questions:
- What problem are we attempting to solve?
- What does the current process cost in time, resources, and opportunity?
- What will implementation, integration, education, governance, validation, and maintenance require, and how will usage affect ongoing expense?
- What measurable capacity will the technology create?
- How will that additional capacity be used?
- What effect should the technology have on cost, quality, outcomes, workflow, or decision-making?
- And, perhaps most importantly, how will we know whether those improvements actually occurred?
These questions reflect the discipline of value analysis, applied to AI.
As AI becomes more deeply embedded across healthcare operations, the ability to distinguish technological capability from organizational value will become increasingly important. A system can be fast, sophisticated, and impressive without producing meaningful improvement.
Our evaluation should establish whether the technology improves the quality of the work, increases the capacity of our teams, strengthens collaboration across the organization, and ultimately supports better decisions.
But if we are going to claim that AI creates value, we should be prepared to demonstrate it.
Kyle Atkins, Ed.S., PMP, NRP, eFACHDM
Kyle Atkins is a healthcare leader, educator, and strategist focused on advancing the responsible use of artificial intelligence in healthcare value analysis. His work centers on applying AI within complex, human-centered systems — strengthening evidence synthesis, decision governance, and operational clarity without compromising ethics, trust, or clinical judgment. A contributor to the Association of Healthcare Value Analysis Professionals (AHVAP) Expert Position Statement on Artificial Intelligence, Kyle actively supports the development of ethical, explainable, and governed AI practices across value analysis programs. Readers interested in expanding their professional development and access to AI-focused resources, including the AHVAP AI Insight Hub, are encouraged to learn more about AHVAP membership at www.ahvap.org. Additional commentary and analysis on AI, governance, and value-based decision-making can be found at kyle.veritastech.io/blog.
Articles you may like:
Cost and Quality Optimization in the Age of AI: Moving Beyond the Spreadsheet
