The 60-Second Version

The technology industry learned to manage trust in fragments: engineering owned one piece, cybersecurity another, privacy, legal, clinical practice, and operations theirs. AI ends that era, because an AI decision moves across all of them: model, infrastructure, workflow, human being, institution. The failure that matters rarely happens inside a discipline. It happens between them, at the seams, where assumptions change, evidence thins, and responsibility can quietly move without anyone noticing. The seam cannot belong to nobody. Someone has to manage enough of the whole to know whether responsibility survived the handoff and whether the institution can stand behind the outcome. That is general management, and it has become business critical: trust now shows up in engagement, repeat purchase, market value, and breach costs. We are living this transition ourselves at Medigram. AI adoption is not the outcome. Better healthcare performance is the outcome.

Do not mistake this for easy work.

To reproduce what Medigram has built to this point, we estimate that it would require four to five exceptional, full-time general managers, representing roughly 17 to 25 cumulative GM-years of leadership, supported by approximately 20 to 25 distinct professional capabilities. Those numbers are not a blueprint for building Medigram. They are a measure of the management burden required when consequential technology has to work across many disciplines at once.

There was a time when technology organizations could manage trust in fragments.

Engineering owned one fragment. Cybersecurity owned another. Privacy, safety, clinical practice, legal, governance, standards, finance, evidence, and operations owned others.

Each discipline developed increasingly sophisticated ways to manage its part.

The technology industry learned to manage fragments of trust long before it learned to manage trust as an institutional outcome.

Now accountability is here.

AI moves through organizations differently. A decision can begin with a model, move through infrastructure and software, enter a clinical or operational workflow, affect a human being, and ultimately become the responsibility of an institution.

The boundaries between those activities do not matter very much to the person affected by the outcome. They matter enormously to the people responsible for producing it.

The fragments have to become a whole

Our experience building Medigram has changed how I think about this problem.

Trustworthy technology is not created because every specialist performed well within the boundaries of a function. You can have excellent engineering and still fail. You can have excellent cybersecurity and still fail. You can have strong clinical expertise, governance, evidence, legal review, and operations and still fail.

Because the failure does not necessarily occur inside any one of those disciplines.

It can happen between them.

Someone has to manage the seams

A requirement changes as it moves from a standard into architecture. An architectural decision changes the security surface. A technical decision changes what can be demonstrated later. A model enters a workflow and encounters clinical reality. An operational constraint changes what can actually be implemented. A locally reasonable choice creates consequences somewhere else.

These are not merely handoffs. They are seams.

The seam cannot belong to nobody.

Someone has to understand enough of the whole to ask whether responsibility survived the handoff, whether evidence still supports the claim, whether one discipline's solution created another discipline's risk, and whether the institution can ultimately stand behind the outcome.

That does not diminish specialists. It makes their expertise more consequential.

That is general management.

Trust has become business critical

For years, trust could be discussed as though it were an intangible benefit: important to reputation, desirable for customers, and good for institutions to possess.

That is no longer an adequate way to understand it.

Trust increasingly affects whether people engage, whether customers buy again, whether employees remain committed, whether institutions can form durable relationships, and ultimately financial performance and enterprise value.

The economic signal is becoming measurable.

PwC reported in 2025 that adding a trust-and-safety feature made 72% of surveyed consumers more likely to engage and 68% more likely to consider purchasing a new product or service. Deloitte reports that 88% of customers who trust a brand buy from it again, and that highly trusted companies can outperform low-trust companies by up to 4x in market value.

In healthcare, the downside can also become a balance-sheet event. IBM reported an average healthcare data-breach cost of $7.42 million in 2025, the highest industry average for the fourteenth consecutive year.

Those figures do not mean trust alone caused every outcome. They mean trust is no longer reasonably treated as a soft concept detached from operating performance.

There is another warning for leadership. PwC found that 90% of executives believed customers highly trusted their companies, while only 30% of consumers said they did (2024 Trust in US Business Survey).

An institution therefore cannot simply assume trust exists because each internal function believes it has done its job.

Trust has to be produced, maintained, demonstrated, and protected as an institutional outcome.

AI makes this business critical because it increases both the number of seams and the speed at which decisions cross them. Engineering decisions become security decisions. Security decisions become operational decisions. Technical performance becomes clinical or business consequence. Governance decisions become evidence obligations.

The fragments are becoming economically connected.

A useful way to think about the value of trust is:

Trust value = growth enabled + friction removed + losses avoided + enterprise resilience preserved.

Growth can come from adoption, strategic relationships, new products, and greater institutional willingness to deploy AI. Friction can be removed when credible evidence makes diligence, procurement, approval, and assurance easier. Losses can be avoided when organizations reduce the likelihood or impact of failed implementations, security events, operational disruption, remediation, litigation, and liability exposure. Resilience is preserved when customers, clinicians, employees, partners, and boards retain confidence in the institution.

No single specialist owns all four.

The enterprise does.

Trust is an institutional outcome

AI trust will not be achieved by adding another isolated trust function to an already fragmented organization. Nor will it come from asking one profession to absorb responsibilities that properly belong to many.

Specialists produce essential fragments of trust. General management makes those fragments function as a whole.

The model may belong to a vendor. The infrastructure may belong to technology. The workflow may belong to operations. The decision may belong to a professional.

The accountability ultimately belongs somewhere.

The organizations that understand this will not merely become better at AI. They will become better institutions.

And that is the standard consequential technology now requires.

We believe enough in AI to live the transition ourselves

I believe AI can and should become an extraordinarily important means of improving healthcare.

In 2025, I was honored to receive the AI Champion of the Year award at AIMed25. I value that recognition because championing AI, to me, has never meant deploying technology simply because it is available. It means doing the work required to make AI relevant, helpful, and effective for the people and institutions depending on it.

That obligation starts with us.

The point of Medigram is not AI for its own sake. Medigram exists to relieve workforce burden, and to do it as infrastructure rather than as one more tool competing for a clinician's attention.

Healthcare runs on dedicated people who quietly absorb the gaps between systems, and that absorption is exhausting the workforce institutions depend on. Infrastructure is the difference. A tool asks people to do one more thing. Infrastructure takes work off them, carries context across the seams, and holds up in the moments when people are most stretched.

We are deliberately learning how to operate Medigram as the kind of organization we believe healthcare institutions will increasingly need to become: one in which people and intelligent systems work together deliberately, human accountability remains clear, evidence survives important decisions, and technology increases the effectiveness of the people responsible for outcomes.

This is not only our experience. Boston Consulting Group reported in 2025 that only 5% of companies in its global study were generating AI value at scale, while 60% reported little material value despite substantial investment. The companies BCG describes as “future-built” are going beyond automation and productivity toward innovation, reinvention, better decisions, and faster action.

Deloitte has reached a complementary conclusion. Its research found that technology-first approaches are more likely to disappoint, while organizations that intentionally redesign roles, workflows, and decision-making around human-machine collaboration are more likely to exceed AI return expectations. Deloitte's 2026 research reports that organizations prioritizing work design are twice as likely to exceed their AI ROI expectations.

We are applying those lessons quietly inside Medigram.

Some recurring intelligence, monitoring, synthesis, and decision-support activities are increasingly supported by autonomous and semi-autonomous workflows. Their purpose is not to remove people from consequential decisions. It is to help the people responsible for those decisions see important changes sooner, retain institutional context, focus their attention, and make better-informed choices.

We expect this approach to evolve as the technology evolves. We will learn where machines are most useful, where human judgment is indispensable, where responsibilities should change, and where new human capabilities are required.

We do not expect our customers to copy how we operate. Healthcare institutions are far more complex, and every organization has different responsibilities.

But we do believe we should be willing to live the transition we are asking others to navigate.

Our aspiration is simple: Medigram should be an organization that stakeholders can trust and that demonstrates, through its own behavior, how AI can make an institution more relevant, helpful, effective, accountable, and capable.

AI adoption is not the outcome. Better healthcare performance is the outcome.

AI can and should become an extraordinarily important means of achieving it.

Selected evidence

Quantitative figures above are external research indicators of economic materiality and are not presented as proof that trust alone causes the reported outcomes.

Companion letters: the Medigram founder letter on evidence and authorship at /ceo-letter/, and the chair letter of the Trustworthy Technology & Innovation Consortium, “Someone Must Own the Seams,” at trustworthytechnologyinnovation.com/blog-own-the-seams.

Sherri DouvilleCEO, Medigram

Companion leadership letter on trust, seams, accountability, and general management.