eTech Insight – Improving AI Success with Accelerators and Incubators - Cover

eTech Insight – Improving AI Success with Accelerators and Incubators

The Problem: Identifying and Supporting AI Projects with High Potential Value:

The healthcare industry has launched hundreds of AI solutions that will likely fail to achieve long-term viability and use. In many cases, the AI solutions are well intended but poorly designed to effectively resolve the operational issues that provider organizations will face or are currently facing. Poorly designed AI algorithms result in AI models that are not generating output that is believable or trusted by providers.

Some of the key AI failures are the following:

  • Brittleness – not recognizing an image that has been changed for a known object, like a new cardiology image.
  • Embedded bias – data algorithms created on stereotypes related to race or socioeconomic factors have been proven to result in less optimal outcomes for some patients.
  • Catastrophic forgetting – new training data sets abruptly forget information they previously knew with updates to the training algorithm.
  • Explainability – the ability to effectively explain the workings of the AI algorithms and models (e.g., the black box) to generate trust with clinicians.
  • Quantifying uncertainty – designing AI algorithms to calculate uncertainty and associated probabilities. This is especially important for AI algorithms that are used for supporting medical diagnoses.
  • Common sense – the ability of the AI model to reach acceptable, logical conclusions based on a vast context of everyday knowledge that people usually take for granted.
  • Math – AI is surprisingly not good at mathematics. Many AI algorithms score low on math problems compared to computer- or human-calculated results.

The Solution: An AI Accelerator Supported by Providers and Big Tech Experts:

The Mayo Clinic has begun an accelerator for health AI start-ups called the Mayo Clinic Platform_Accelerate. The four start-ups in the first cohort will participate in a 20-week program run by the Mayo Clinic with tech experts from Google and Epic. This AI-accelerator collaboration should help alleviate several of the previously mentioned AI failures. The Mayo Clinic AI experts should be able to help reduce or eliminate embedded bias, explainability, and common-sense challenges. Google and Epic should be expected to improve the AI models relative to brittleness, catastrophic forgetting, quantifying uncertainty, and math.

The power of this accelerator is providing both end-user and technical expertise that can be focused to help AI solutions provide more timely benefits for their targeted benefits that will drive higher levels of clinician adoption. Current incubator solutions include the following:

  • Quadrant Health – medical record and patient-messaging analytics to coordinate communications and even predict patient harm ahead of time.
  • ScienceIO – using AI to create organizational tools to cut down on paperwork and administrative processes for doctors.
  • cliexa – focusing on using AI to provide care plans for people with chronic or cardiovascular diseases to maintain their health.
  • Seer Medical – supporting at-home epilepsy diagnostics and management models by finding digital biomarkers to predict seizures.

AI incubators/accelerators can also provide needed capital/operating support that enables the AI vendor to achieve a higher level of maturity that can improve the probability of acquiring additional funding levels needed for ongoing growth.

The Justification: Supporting Healthcare AI Start-ups with Expertise and Operational Stability:

The AI incubator environment will be very valuable for healthcare as AI development continues to mature. The incubator environment can provide expertise for the environments the AI algorithms are designed to support while also providing operational stability for the emerging vendors. In some healthcare organizations, the innovation centers act as incubators for creating solutions that the healthcare organizations desire that are either not commercially available or not at an acceptable maturity level.

AI solutions developed in incubator environments that are supported by both provider organizations and technical companies (e.g., Google, Amazon, Microsoft, Apple) will likely be successful as the technical companies can improve technical designs and providers can offer real-world expertise. 

The Players: Accelerators and Incubators Driving Emerging Healthcare Solutions:

Accelerator and incubator companies generating successful emerging technology companies include the following:

  • Mayo Clinic Platform_Accelerate – a collaboration of Mayo Clinic, Google, and Epic to support emerging AI solutions with high potential.
  • Matter Health – a start-up incubator that includes a community nexus and oversight from providers to focus on the innovation.
  • Cedars-Sinai Accelerator – helps entrepreneurs bring their innovative technology products to market.

Success Factors:

  1. Provider innovation centers looking to expand their support capabilities for emerging technology solutions should partner with existing incubator/accelerator companies or seek alliances with technology companies with expertise in the technologies they want to develop.
  2. Incubator/accelerator solutions based on AI technologies must include provider resources that are experienced in data analytics and people experienced with AI for either machine learning or neural networks.


AI continues to occupy the “Inflated Expectations” portion of the Gartner Hype Cycle for healthcare. While we all see the vast potential of AI for resolving key operational, diagnostic, and predictive challenges that providers face as we transition to value-based care, few AI solutions have emerged that have proven to be accurate or trusted.

As the industry continues to pursue effective AI solutions, the ability to create an effective AI design with incubators/accelerators that generate proven AI outcomes/results will be enhanced with the oversight of experienced providers that can monitor the technical designs and data models used for the AI algorithms and models. Technical expertise provided by technology companies who use AI across many industries will also contribute to the design of AI algorithms. Incubators/accelerators that use this combination of professional resources will likely create successful emerging technology companies that provide solutions that are trusted and achieve higher provider-adoption rates. This approach should significantly reduce the amount of time the AI solutions occupy the “Trough of Disillusionment” portion of the Gartner Hype Cycle for healthcare.

Photo Credit: maxsim, Adobe Stock