1. Start with the enterprise problem, not the startup pitch

The strongest investment thesis begins inside the health system. Leaders should identify a small number of problems that are strategically important, operationally consequential, and measurable. Examples may include avoidable clinical deterioration, delayed access, excessive documentation burden, fragile care transitions, revenue leakage, supply disruption, medication risk, cybersecurity exposure, or inequitable outcomes. A solution becomes relevant only after the organization defines the problem, the affected population, the current workflow, the baseline performance, and the cost of leaving the problem unresolved.

This sequence prevents technology from defining the agenda. Startup presentations often begin with a product category, a proprietary capability, or a market forecast. Health systems must begin with a care or operating need. If leadership cannot describe the problem without mentioning the product, the investment case is probably not mature.

The executive team should ask five initial questions:

1. What patient, workforce, financial, or operating problem are we trying to solve?
2. Why is this problem important now?
3. Who owns the outcome and the affected workflow?
4. What measurable change would constitute meaningful value?
5. Why might an external startup be better positioned than an internal team or established partner?

The fifth question is often neglected. A startup may offer speed, specialized expertise, a novel data asset, or a more focused product team. It may also introduce immaturity, limited implementation capacity, uncertain financing, and dependence on a small number of founders. Executives should compare the startup option with improving an existing platform, redesigning the process, using an established vendor, building internally, or collaborating with an academic partner.

The proper comparison is not startup versus doing nothing. It is startup versus the best feasible alternative.

2. Define a healthcare-specific investment thesis

A health system investment thesis should specify where the organization will engage, what strategic value it expects, what risks it will accept, and what evidence will justify additional capital or adoption. It should be narrow enough to guide decisions and broad enough to support learning across several opportunities.

A useful thesis includes six elements.

Strategic domains

Identify priority domains such as access, care navigation, workforce capacity, clinical decision support, chronic disease management, behavioral health, revenue cycle, supply chain, data infrastructure, cybersecurity, or home-based care. These priorities should come from the enterprise strategy and capital plan rather than from external deal flow.

Target populations and settings

Specify whether the organization seeks value in acute care, ambulatory services, rural access, post-acute care, specialty care, employer health, or community-based services. A product that succeeds in a technology-forward academic center may not translate directly to a rural hospital, safety-net environment, or multi-state system.

Value definition

Describe the outcomes that matter. Clinical quality, patient safety, equitable access, workforce retention, throughput, total cost of care, margin improvement, consumer trust, and learning value may all be relevant. Leadership should rank these outcomes rather than treating every benefit as equally important.

Risk boundaries

State what the organization will not compromise. Patient safety, privacy, informed consent, regulatory compliance, cybersecurity, nondiscrimination, clinician accountability, and continuity of care should have clear minimum standards. A high-growth opportunity does not justify an unclear clinical boundary.

Capital and resource envelope

Define the amount of financial capital, implementation support, executive attention, clinical time, data access, and technical capacity available. A small equity check can create a large operating obligation if the startup expects access to clinicians, data, integration teams, or a flagship customer relationship.

Decision horizon

Set the expected period for validation, adoption, and strategic review. Some clinical technologies require longer evidence generation or regulatory pathways. Others should demonstrate workflow or financial value within months. The timeline must fit the risk and the mechanism of value.

The investment thesis should be approved by the appropriate executive and board bodies. It should be reviewed periodically, but leaders should avoid changing it to justify a favored deal. A thesis that expands whenever an attractive presentation arrives is not a thesis. It is a record of exceptions.

3. Choose the right engagement model

Healthcare organizations can work with startups through several models. Each creates different rights, obligations, and risks. Leaders should select the model that fits the objective instead of assuming that every relationship requires equity.

Customer relationship

The organization purchases a product or service through standard contracting and procurement. This model is appropriate when the need is established, the offering is sufficiently mature, and the health system does not require strategic rights. Procurement discipline remains essential, especially when a startup depends on the contract for survival or lacks implementation depth.

Structured pilot

The organization tests a product in a limited setting with predefined endpoints, safeguards, ownership, and a decision date. A pilot should answer a specific question. It should not become an indefinite substitute for an enterprise decision. Leaders should define success, failure, and insufficient evidence before the work begins.

Clinical or operational validation partnership

The health system and startup collaborate to generate evidence, refine workflow, or validate performance. This may involve research oversight, data governance, intellectual property, publication rights, and regulatory considerations. The parties must distinguish quality improvement, product evaluation, and human-subjects research rather than treating the labels as interchangeable.

Strategic commercial partnership

The organization may receive preferred access, co-development rights, regional arrangements, or shared commercialization opportunities. These relationships can create deeper alignment but may also limit future choice. Exclusivity should be narrow, time-limited, and supported by measurable obligations.

Direct equity investment

The health system purchases an ownership interest. Equity can align incentives and create financial upside, but it also introduces valuation, governance, liquidity, conflict-of-interest, and concentration risks. An equity stake should never lower the standard used to evaluate the product as a customer or clinical partner.

Venture fund participation

The organization invests as a limited partner or through a dedicated corporate venture structure. This approach can provide diversified exposure and market intelligence. It can also distance the organization from direct clinical priorities if fund incentives focus primarily on financial return. Leaders need explicit expectations regarding strategic access, information rights, and conflicts.

Incubation or spinout

The health system supports internally developed intellectual property, talent, or workflows through a new company. This model may unlock value from innovation created inside the organization. It requires disciplined handling of ownership, licensing, inventor rights, institutional resources, and the separation between clinical duties and commercial interests.

The engagement model should follow the strategic objective. If the goal is to solve a workflow problem, a commercial contract or structured pilot may be enough. If the goal is to develop a new regulated technology, a validation or co-development agreement may be appropriate. If the organization seeks diversified financial exposure, a fund structure may be more suitable than a series of direct minority investments.

4. Establish an investment committee with clinical authority

Healthcare startup decisions should not sit exclusively with innovation, information technology, or finance. Each function sees a different part of the risk. The operating model must bring those perspectives together while preserving clear accountability.

An investment committee may include the chief executive, chief operating officer, chief financial officer, chief medical officer, chief nursing officer, chief information or digital officer, compliance and privacy leadership, legal counsel, strategy, procurement, and the executive responsible for the affected service line. The committee does not need every participant at every meeting, but its charter should define required review based on risk.

Clinical authority is essential. A technology that influences diagnosis, treatment, prioritization, staffing, patient communication, or care transitions requires clinical review even when it is marketed as administrative software. The committee should identify who can approve the clinical use case, who can pause deployment, and who owns surveillance after launch.

The committee charter should define:

• the size and type of commitments it may approve;
• when board approval is required;
• the difference between investment approval and purchasing approval;
• conflict-of-interest disclosure and recusal requirements;
• clinical, regulatory, privacy, cybersecurity, and equity review triggers;
• evidence standards by level of patient and enterprise risk;
• reporting expectations for active investments and pilots;
• criteria for additional capital, scale, pause, redesign, or termination;
• responsibility for documenting dissent and unresolved risk.

Good governance does not eliminate speed. It eliminates repeated uncertainty. When leaders know the standards, decision rights, and evidence requirements, the organization can move faster without improvising every review.

5. Use a ten-domain diligence framework

Traditional venture diligence evaluates market size, team, technology, competition, business model, and financial potential. Healthcare systems need those dimensions plus a deeper review of clinical use, workflow, regulation, data, security, reimbursement, and operational resilience.

1. Problem significance

Confirm that the problem is important enough to justify change. Review baseline performance, affected volume, variation, current cost, patient impact, and strategic urgency. Avoid inflated market narratives that are not connected to the organization’s own data.

2. Clinical evidence and safety

Determine what evidence supports the product’s intended use. Review study design, comparator, population, setting, endpoints, limitations, and external validation. Ask whether the evidence demonstrates technical performance, clinical utility, operational value, or patient outcomes. These are not equivalent.

3. Regulatory status

Clarify whether the product is a medical device, clinical decision support tool, administrative platform, or another regulated offering. Verify current clearances, authorizations, registrations, and intended use. The FDA notes that artificial intelligence-enabled device software may be reviewed through pathways including 510(k), De Novo classification, or premarket approval. A company’s description of its regulatory position should be confirmed against authoritative records and legal review.

4. Workflow fit and human factors

Map how work changes for patients, clinicians, staff, and support teams. Identify new steps, alerts, handoffs, exceptions, training needs, documentation, and downtime procedures. A product can demonstrate technical accuracy and still fail because it adds friction or shifts work to an already constrained team.

5. Data governance and interoperability

Specify the data required, its source, quality, permitted use, retention, location, and deletion process. Determine whether the company will train models, create derived data, use subcontractors, or combine information across customers. Review interface standards, application programming interfaces, identity management, auditability, and exit requirements.

6. Cybersecurity and resilience

Evaluate architecture, access controls, encryption, vulnerability management, incident response, business continuity, recovery objectives, software dependencies, and third-party risk. Require prompt notification and coordinated response obligations. Consider what happens to care if the startup’s service fails, the company loses funding, or a key cloud dependency becomes unavailable.

7. Reimbursement and economic model

Identify who pays, what activates payment, and whether the model depends on uncertain coding, coverage, or utilization assumptions. CMS has continued to develop models that connect technology-supported care with measurable outcomes. The ACCESS Model, for example, introduces outcome-aligned payments for qualifying chronic conditions. A startup’s economic thesis must match the actual payer, contracting, and care model relevant to the health system.

8. Company and team durability

Review leadership experience, clinical expertise, implementation capacity, customer concentration, financing runway, burn, governance, key-person dependence, insurance, litigation, and succession. A promising product can become an operational liability if the company cannot support enterprise deployment.

9. Enterprise economics

Model the full cost of adoption, not only the contract price. Include integration, interfaces, cybersecurity, devices, data work, training, workflow redesign, clinical governance, support, change management, downtime planning, compliance, and opportunity cost. Estimate benefits conservatively and identify who can verify them.

10. Scale and exit path

Define what enterprise scale would require and how the organization can exit safely. Confirm data portability, transition support, patient communication, knowledge transfer, device recovery, and continuity obligations. A health system should not enter a pilot without knowing how it would end.

The committee should score these domains, but it should not allow a total score to hide a critical failure. Certain deficiencies should operate as gates. A product with strong market potential but unacceptable clinical risk, unclear data use, or inadequate resilience should not advance until the issue is corrected.

6. Separate product promise from implementation readiness

Many startup evaluations focus on what the technology can do under favorable conditions. Executives must also evaluate what the organization must do for the technology to work in routine care. That includes staffing, training, workflow ownership, data quality, integration, patient engagement, communication, exception management, and performance monitoring.

Implementation readiness should be assessed on both sides of the relationship.

The startup should demonstrate:

• a realistic implementation plan and named accountable leader;
• sufficient customer support and clinical expertise;
• transparent technical and data requirements;
• escalation procedures and service expectations;
• training materials appropriate to the users;
• a method for monitoring performance and unintended effects;
• financial capacity to support the agreed deployment.

The health system should demonstrate:

• an executive sponsor and operational owner;
• a clinical owner when care may be affected;
• committed implementation and technical resources;
• access to baseline and outcome data;
• capacity for training and workflow redesign;
• a defined adoption strategy;
• the authority to make the next decision.

When either side lacks readiness, the project should be delayed, narrowed, or redesigned. Starting a pilot without committed operating capacity does not test the product fairly. It tests whether an unsupported project can survive inside a complex organization.

7. Design pilots as decisions, not demonstrations

A pilot should reduce uncertainty that matters to the next decision. It should not be a prolonged product demonstration, a public-relations event, or a favor to an internal champion. Every pilot should have a written charter.

The charter should include:

• the problem and target population;
• the intended use and excluded uses;
• the clinical and operational owners;
• the baseline and comparison approach;
• primary and balancing measures;
• patient and workforce safeguards;
• data access and governance;
• implementation responsibilities;
• duration and minimum sample requirements;
• decision thresholds;
• the date and authority for the scale, redesign, pause, or stop decision.

Measures should reflect the full operating impact. A tool that reduces documentation time but increases correction work, patient confusion, or cybersecurity exposure may not create net value. A care navigation platform that improves engagement but attracts only digitally confident patients may widen inequity. A diagnostic support tool that performs well overall may create unacceptable variation across subgroups or settings.

Balancing measures make these tradeoffs visible. They should examine unintended consequences, workload shifts, access differences, false positives or negatives where relevant, patient complaints, clinician overrides, downtime, and the cost of managing exceptions.

The organization should also identify the minimum evidence needed to make the next decision. Not every pilot can establish long-term clinical outcomes. It can still test adoption, workflow reliability, technical performance, data quality, or economic assumptions. Leaders must be precise about what the pilot can and cannot prove.

8. Build an enterprise economic case

Startup business cases often emphasize revenue growth, avoided cost, productivity, or market differentiation. These claims should be translated into the health system’s own financial and operating model.

The economic case should include four layers.

Direct financial impact

Estimate contract cost, implementation expense, capital requirements, staffing changes, reimbursement, revenue, avoidable utilization, and cash timing. Distinguish accounting savings from released capacity and theoretical savings. Reducing minutes in a workflow creates value only if the time can be redeployed or the burden reduction produces another measurable benefit.

Clinical and operational value

Quantify changes in access, safety, quality, throughput, length of stay, care transitions, staff workload, patient experience, and reliability. These outcomes may have financial implications, but they should not be reduced to dollars when the connection is weak.

Strategic option value

Some investments create learning, market access, talent, intellectual property, or future flexibility. Leaders may choose to invest before the full economic case is proven when the strategic option is important. The committee should label this value explicitly rather than disguising it as a near-term return.

Risk-adjusted downside

Model implementation failure, slower adoption, financing instability, regulatory delay, data incidents, clinical harm, reputational loss, and replacement cost. A base case without a downside case is advocacy, not analysis.

The committee should test assumptions through scenarios. What happens if adoption reaches only half the expected level? What if integration takes six months longer? What if reimbursement changes? What if the company needs another financing round before the health system can scale? What if the organization must replace the product in two years?

These questions protect both parties. A startup benefits when the customer understands what successful adoption requires. A health system benefits when enthusiasm is converted into explicit operating assumptions.

9. Protect independence when equity and procurement intersect

Equity ownership can create real or perceived conflicts. The organization may become financially interested in a company whose product it evaluates, purchases, studies, or recommends. Executives and clinicians may hold advisory roles, receive compensation, or participate in intellectual property arrangements. These relationships require more than disclosure.

The health system should separate the investment decision from the clinical, procurement, and research decisions to the extent practical. An equity position should not guarantee purchasing volume, preferred clinical treatment, access to patients, or favorable study conclusions. Product evaluation should use the same or stronger evidence and safety standards applied to nonportfolio companies.

Policies should address:

• personal and institutional conflicts of interest;
• recusal and independent review;
• valuation and related-party transactions;
• use of the organization’s name and reputation;
• access to clinicians, patients, and data;
• publication rights and scientific independence;
• intellectual property and licensing;
• gifts, advisory compensation, and board service;
• communication with patients and employees;
• board reporting.

The appearance of compromised judgment can damage trust even when the underlying decision is defensible. Leaders should design the relationship so that an independent observer could understand how patient interests remained primary.

10. Govern data, artificial intelligence, and learning systems over time

Digital health products increasingly use models that change, learn, or depend on shifting data. Evaluation cannot end at go-live. The organization needs ongoing surveillance.

For artificial intelligence and algorithmic products, leaders should document intended use, model version, training and validation context, user responsibility, performance thresholds, subgroup analysis, monitoring cadence, override pathways, and change control. They should know whether the system produces a recommendation, prioritizes work, automates a decision, or acts without direct review.

The FDA’s Digital Health Center of Excellence emphasizes responsible, high-quality digital health innovation. That direction is consistent with an executive duty to connect adoption with governance. A model may be technically sophisticated and still be unsafe in a new population, poorly integrated into workflow, vulnerable to data drift, or misunderstood by users.

Health systems should require notification of material model, software, data, subcontractor, or hosting changes. They should preserve the right to test updates, suspend use, and investigate performance. Contracts should define how incidents, complaints, corrections, and recalls will be handled.

The CMS health technology ecosystem also places growing attention on interoperable, patient-centered digital infrastructure. Strategic investors should therefore favor companies that can participate in an open, secure, standards-based environment rather than creating avoidable data lock-in.

11. Manage the portfolio, not only individual deals

A series of reasonable individual investments can still create an incoherent portfolio. The executive team should review the combined exposure, resource demand, strategic coverage, and learning value across all startup relationships.

Portfolio review should examine:

• concentration by technology, founder, fund, clinical domain, and risk;
• cumulative financial commitments and follow-on expectations;
• total demand on integration, cybersecurity, legal, clinical, and change teams;
• overlapping products and duplicative pilots;
• stage distribution from exploration to scaled adoption;
• equity exposure and liquidity horizon;
• performance against strategic priorities;
• number and age of pilots without a decision;
• dependencies on companies with limited financing runway;
• enterprise learning that can improve future decisions.

Leaders should classify each relationship as explore, validate, adopt, scale, hold, or exit. This simple discipline reveals when the organization is accumulating pilots without moving them forward or closing them. It also prevents a well-connected sponsor from keeping an underperforming project alive without evidence.

Capital allocation should consider implementation capacity as a scarce resource. A health system may be able to fund more pilots than it can responsibly support. When integration, clinical leadership, or change capacity becomes the constraint, adding projects reduces the probability that any of them will succeed.

12. Use a balanced executive dashboard

The board and C-suite need a concise view of portfolio performance. The dashboard should not reward activity alone. Counts of meetings, pitches, pilots, or startup relationships can create the appearance of innovation without showing value.

A balanced dashboard may include:

Strategic alignment

• percentage of investments tied to approved enterprise priorities;
• distribution across priority domains and populations;
• concentration and dependency risk.

Evidence and safety

• projects by evidence stage;
• unresolved clinical, regulatory, privacy, or cybersecurity issues;
• performance against safety and balancing measures;
• material incidents, complaints, or pauses.

Adoption and operations

• active-user and workflow adoption measures;
• implementation milestones;
• training completion and support demand;
• integration reliability and downtime.

Value

• outcomes against the approved business case;
• verified financial impact;
• patient, workforce, and access effects;
• strategic learning and option value.

Capital and company health

• committed, deployed, and reserved capital;
• follow-on obligations;
• portfolio company runway and material financing risks;
• valuation and liquidity indicators where applicable.

Decision discipline

• pilots with defined decision dates;
• time from initial review to decision;
• scale, redesign, pause, and stop rates;
• projects operating beyond their approved period.

The dashboard should prompt inquiry rather than produce a decorative score. A high adoption rate may be good, or it may reflect use by a narrow group while high-risk patients remain excluded. A low stop rate may signal strong selection, or it may indicate reluctance to acknowledge failure. Metrics require case review and context.

Executive conclusion

Healthcare startup investment can strengthen access, care delivery, workforce performance, and strategic resilience. It can also consume capital, create operational fragmentation, introduce clinical and cybersecurity exposure, and distract leaders from problems that require process discipline rather than new technology.

The difference is not enthusiasm. It is governance.

C-suite leaders should treat venture engagement as an operating capability. Start with the enterprise problem. Define a narrow investment thesis. Choose the engagement model that fits the objective. Apply healthcare-specific diligence. Separate equity from clinical and purchasing judgment. Design pilots as decisions. Measure total adoption cost and verified value. Govern data and algorithms over time. Review the portfolio as a whole. Stop weak projects with the same discipline used to fund promising ones.

The executive promise should be clear: the organization will not finance novelty for its own sake. It will invest where disciplined partnership can produce safer care, better work, measurable value, and a stronger capacity to serve patients over time.