Billy AI

The US healthcare system is one of the most expensive and complicated in the world, in large part due to administrative complexity. The healthcare system as a whole is overburdened with intricate procedures for managing medical claims and denials, which are frequently executed through manual means.  These claims handling processes create immense financial waste and take time away from patient care, especially in underserved communities. This puts provider sustainability at risk, affecting providers’ bottom lines and workforce capacity. 

Market research indicates that: 

- US hospitals lose over $262 billion per year, or 10% of total claims value, due to denied claims. The average 350-bed hospital loses $5 million annually due to denied claims.

- 65% of denied claims are never resubmitted, resulting in significant lost revenue opportunity while appeals processes are time-consuming and resource-intensive.

- Claim errors, such as missing documentation, incorrect codes, or failing to follow payer-specific rules, are a leading cause of denials. 30-40% of denials are attributed to registration and technical errors. 

- Manual claims processing is slow, costly, and prone to human error. Medical billing and coding staff spend countless hours on data entry and fixing mistakes.

This problem disproportionately impacts under-resourced healthcare providers and communities:

- Safety-net hospitals that primarily serve low-income, uninsured, and vulnerable populations operate on thin margins. Denied claims can threaten their financial stability and ability to provide care. 

- Rural hospitals have fewer resources to handle billing complexities. Delayed and unpaid claims strain already limited budgets.

- Uncompensated care from claim denials can lead providers to limit services for Medicaid patients and other underserved groups. Meanwhile, patients may get into medical debt because of the denials.

In summary, Billy AI aims to increase capacity and resilience of the US health system. Specifically, Billy AI reduces administrative complexity and prevents lost revenue for healthcare providers by improving the efficiency and accuracy of processing medical claims. Improved processes prevent financial waste, improving provider sustainability and expanding access to care, especially for underserved populations. 

For the specific solution we provide, Billy AI is an AI-powered software platform that automates and streamlines the medical claims filing and denial management process for healthcare providers. Our solution uses advanced and secure Large Language Models to reduce errors, improve reimbursement rates, and save time for billing staff.

Here's how it works:

1. Data Ingestion: Billy AI securely integrates with a provider's electronic health record (EHR) system and practice management software to extract relevant patient demographic, insurance, and encounter data.

2. Claims Generation: Our fine-tuned Large Language Models analyze unstructured clinical notes to identify billable services, procedures, and diagnoses. This information is automatically populated into claim forms with the appropriate medical codes.

3. Error Detection: The system will then present a list of pre-populated claims to medical coders/billers and flag potential errors, omissions, or inconsistencies in claims for coders/billers to review prior to submission. This includes checking for missing information, incorrect coding, and compliance with payer-specific requirements.  

4. Payer-Specific Optimization: Our models are trained on millions of historical claims to predict the likelihood of denial by each payer. Claims are optimized in real-time to maximize the probability of first-pass acceptance.

5. Denial Management: When claims are denied, Billy AI automatically reviews the reason for denial and determines the next best action. This could include correcting and resubmitting the claim, filing an appeal, or requesting additional documentation. Our models continuously learn from successful appeals to improve recommendations.

6. Analytics and Reporting: Providers can access real-time analytics on their billing and denial trends, staff productivity, and reimbursement rates. Customizable reports and dashboards aid in performance improvement and revenue forecasting.

By automating the most manual and error-prone aspects of revenue cycle management, Billy AI enables billing staff to focus their time on more complex issues while reducing costly denials. Our platform is HIPAA-compliant, EHR-agnostic, and easy to implement either on cloud or on device depending on security needs. 

We believe that every provider should be paid fairly and promptly for the care they deliver. Billy AI's intelligent automation empowers providers to improve financial performance, reduce administrative burden, and ultimately direct more resources to patient care.

See our 1-minute demo video here.

Billy AI's solution serves healthcare providers, particularly those in underserved communities, who struggle with the financial and administrative burden of medical billing and insurance claims management. Our AI-powered platform directly improves the lives of several key populations:

1. Physicians and nurses at community health centers and safety-net hospitals: These providers serve predominantly low-income, uninsured, and vulnerable patients. They operate on razor-thin margins and rely heavily on Medicaid reimbursement. Claim denials and delays threaten their financial viability and ability to maintain services. By reducing denials and accelerating payments, Billy AI alleviates financial strain and allows these providers to focus on patient care. We estimate that our solution can help a typical community health center recover $500K-$1M in annual revenue, funds that can be reinvested into patient programs and staff.

2. Billers, coders, and revenue cycle staff: These unsung heroes of the healthcare system are bogged down by manual, repetitive tasks and ever-changing payer rules. High stress leads to rapid burnout and turnover. Billy AI automates the most error-prone aspects of their job, reduces tedious data entry, and provides intelligent guidelines to decision-making. We aim to cut claim touch time by 50% and reduce stress and overtime. By making their jobs more fulfilling and impactful, we hope to improve job satisfaction and retention.

3. Patients, especially those in underserved communities: When providers face financial hardship from denied claims, patients suffer. Practices may cut back services, limit Medicaid patient slots, or even close sites. Patients face access barriers, longer wait times, and rushed visits. By strengthening providers' revenue, Billy AI helps preserve critical access points for underserved patients. Financially stable practices can invest more in social determinants programs, care coordination, and patient education - all of which improve outcomes. Coding and reimbursement also affects the bills that patients receive. Inaccurate claims often leave patients responsible for incorrect bills or “surprise bills.” Patients with lower financial and health literacy have less resources to fight or correct bills, leading to medical debt. While patients may never know Billy AI by name, they will feel the downstream impact on the quality and accessibility of their care and fiscal responsibility. 

4. Payers and health plans: While payers may seem like an odd beneficiary for our provider-focused solution, they too stand to gain. Claim denials are expensive to process, with payers spending $3-5 per claim on review. By reducing the overall denial rate, Billy AI can help plans save millions in administrative cost - savings that can be passed on to patients via lower premiums. More accurate coding and documentation also supports value-based contracts and risk adjustment. Plans can reinvest these savings into member benefits and population health initiatives.

Fundamentally, Billy AI's solution is about making the healthcare revenue cycle more efficient, transparent, and equitable. By eliminating waste and friction in the billing process, we aim to redirect more resources - both dollars and time - to patient care and population health. Our vision is a healthcare system where providers can focus on care, not coding, where patients can access services without fear of surprise bills and where payers and providers can collaborate on value. While our platform alone won't solve every challenge, we believe it is a critical piece of the puzzle towards a more resilient, equitable health system for all.

Our team at Billy AI is uniquely positioned to deliver our AI-powered medical claims management solution due to our deep proximity to and representation of the healthcare provider communities we aim to serve. 

As a co-founder, Alisha has seen firsthand the financial strain that claims denials place on healthcare organizations. Her best friend is a primary care physician at a community health clinic that struggles to get paid by insurance companies for services rendered to underserved patients. Witnessing her constant battles with payers to keep the clinic afloat fuels Alisha’s drive to fix this broken system. Another co-founder Chris previously worked at Grady, the largest public hospital in the Southeast, which nearly went bankrupt from unexpected drops in reimbursement. In addition, Mackenzie, also has deep insights in this field by serving as a public health policy consultant in Deloitte for a long time. Hence, the team intimately understands the pain points and needs of providers because we have lived them.

Furthermore, our team composition reflects the diversity of the healthcare ecosystem we serve:

- Yiko, our product designer, brings the patient perspective as individuals who have navigated complex medical bills and experienced the downstream effects of provider financial instability on quality of care and also worked for a hospital’s in-house billing department to revamp their outdated claim processing system’s user interface. 

- Philippe, Lotus and Alisha, with backgrounds in data science, computer science, machine learning and engineering, have worked closely with hospital revenue cycle teams to streamline their processes. They understand the day-to-day workflows and challenges of billers and coders and how billing and coding affects the healthcare system and population health. 

In developing our solution, we are taking an intentionally community-driven approach. We have established an advisory board of billers, coders, clinicians, and revenue cycle leaders to guide our product roadmap and go-to-market strategy. Through interviews, shadowing, and focus groups, we continuously gather their input on the most critical pain points to solve, key workflow integration points, and metrics for success.

Diversity, equity, and inclusion are also core to Billy AI's DNA. We believe our solution will only succeed if it is accessible and valuable to providers serving underserved communities, not just large health systems. To that end, our target pilot sites are small primary care clinics and community health centers in low-income urban and rural areas. We will try to partner with organizations like the National Association of Community Health Centers to understand and address the unique billing challenges faced by these providers.

By combining lived experience with deep healthcare industry expertise, we are building a multi-disciplinary team culture that mirrors and uplifts the resilience of the providers we serve. Our proximity enables us to recognize both the assets and challenges in their communities, and to co-create a solution that measurably improves health access and outcomes by strengthening the financial sustainability of frontline care.

We do not currently have any users but are working on an MVP that we are validating with potential customers and medical billers and coders. We recently came in second place in the Harvard Hackathon, where we continued to iterate our initial working version of Billy AI.

At Billy AI, we are applying to MIT Solve because we believe the program's unique combination of funding, mentorship, and network access can help us overcome critical barriers to scaling our AI-powered medical claims management solution and maximizing our impact for underserved healthcare providers and communities.

While the $10,000 grant prize would certainly be valuable in extending our runway, we are most excited about the potential for non-monetary support through Solve's 9-month program and cross-sector partner community. Specifically, we hope Solve can help us tackle the following key challenges:

1. Navigating healthcare data privacy and security regulations: As we build out our data infrastructure and partnerships with health systems, ensuring compliance with HIPAA and other evolving regulations is both essential and complex. We would benefit immensely from legal guidance and mentorship from Solve partners with deep healthcare regulatory expertise. 

2. Accessing social impact funding: As an early-stage startup committed to serving underserved providers, we sit at the intersection of tech and social impact. Solve's network of impact-minded investors and entrepreneurs could unlock critical sources of values-aligned capital to fuel our growth.

3. Building trust and credibility with healthcare stakeholders: The revenue cycle management space is crowded with legacy vendors promising a better mousetrap. We need help telling our story, proving our impact, and elevating the voices of the providers and patients we serve. Solve's media partnerships and showcase events offer powerful platforms to build our brand authentically.

4. Developing our impact measurement muscle: While we have strong hypotheses and anecdotal evidence around our solution's impact on key metrics like denied claims rate and bad debt, we need to build more rigorous impact measurement practices to continuously monitor and improve. Participating in Solve's Impact Measurement & Management program would be game-changing.

5. Pressure-testing our health equity strategy: Equity is core to our mission, but we know we have blindspots as a founding team. Solve's diverse, global community can push our thinking and connect us to partners who've successfully built inclusive, culturally competent healthcare solutions.

Beyond these specific barriers, we are drawn to Solve's collaborative, solutions-oriented ethos. The opportunity to connect with and learn from other Solver teams tackling systemic challenges in innovative ways would be invaluable as we chart our own path. We believe no one organization alone can fix the tangled web of our healthcare system and it will take a web of changemakers bound together by a shared vision for health justice, for which Solve offers a unique platform to co-create that future.

Billy AI innovates through streamlining a complex process, using new technology in an outdated ecosystem. Advanced AI-Powered Automation will significantly reduce human error and accelerate the claims process, two notoriously difficult pieces in healthcare administration. Training on historical claims data to create payer-specific optimization is a first, which will maximize first-pass acceptance and revolutionize back-office procedures. Setting a new standard for healthcare efficiency could shift towards greater transparency, accuracy, and collaboration, leading to improvements in healthcare delivery and outcomes. The slow adoption of technology in the US healthcare system makes AI adoption revolutionary, and Billy AI revents an opportunity to leapfrog outdated systems.  Overall, improved revenue cycle management frees up providers and health system leadership to innovate in healthcare in other ways, including improving equity and access by reducing financial strain and improving sustainability. This expands access to care and supports initiatives addressing social determinants of health.

Activity: Billy AI automates the claims filing and denial management process using AI-powered technology. 

Short-Term Outcomes: By automating and streamlining claims processing, Billy AI reduces human error, speeds up the claims filing process, and decreases the rate of denials, thus getting providers paid what they are owed, and paid on time. 

Longer-Term Outcomes: Over time, Billy AI's solution helps healthcare providers improve their financial stability and sustainability. By reducing financial strain, providers can maintain or expand services, especially in underserved communities. This strengthens the healthcare system overall and improves access to care for vulnerable populations, prevents patients from receiving incorrect bills, and reduces administrative complexity that makes healthcare in the US inefficient and inexpensive.

Goal: Reduce Claim Denial Rates by 50% for healthcare providers using Billy AI.

Measurement: Track denial rates before and after implementing Billy AI, monitoring improvements over time.

Goal:  Help providers recover 10% more of annual revenue. 

Measurement: Compare providers' revenue from claims before and after using Billy AI, measuring the increase in recovered revenue.

Goal: reduce claim touch time by 50% for billing and coding staff.

Measurement: Monitor the average time spent processing claims and track reductions over time.

Goal: strengthen the financial stability of safety-net and rural hospitals by reducing denied claims.

Measurement: interviews with providers and investigate perceived financial stability and financial goals once reimbursement is improved (i.e. is access to care expanded).

Central to Billy AI's platform are sophisticated LLMs that have undergone extensive fine-tuning using vast amounts of medical claims data. Through the utilization of transfer learning, it is possible to modify open-source general-purpose language models such as Llemma and Mistral in order to address the unique challenges of healthcare revenue cycle management. Our fine-tuning process involves several key steps:

1. Data Preparation: We start by curating a large corpus of anonymized medical claims spanning multiple years, payers, and provider specialties. This data is cleaned, normalized, and annotated with key metadata such as diagnosis codes, procedure codes, and denial reasons.

2. Model Pre-training: We initially pre-train our LLMs on a broad set of publicly available healthcare datasets, such as MIMIC-III and PubMed, to establish a strong baseline understanding of medical terminology, clinical concepts, and documentation norms.

3. Domain-Specific Fine-Tuning: We then fine-tune our pre-trained models on our proprietary claims dataset using a combination of supervised and unsupervised learning techniques. This allows the models to develop a deep understanding of the nuances and patterns specific to medical billing and coding.

4. Task-Specific Adaptation: Finally, we adapt our fine-tuned models to specific tasks within the revenue cycle workflow, such as claim generation, error detection, and denial management. This involves training the models on task-specific input/output formats and reward functions to optimize their performance.

A suite of highly specialized LLMs is produced as a final product, capable of intelligently triaging denied claims for appeal or write-off, identifying and correcting errors prior to submission, and automatically generating clean, compliant claims from raw clinical documentation.

1. Natural Language Understanding: Our models can comprehend a wide range of clinical language and documentation styles, accurately extracting key information like patient demographics, diagnoses, and procedures.

2. Code Classification: By learning the complex mappings between clinical concepts and standardized coding systems like ICD-10 and CPT, our models can automatically assign the most appropriate codes to each claim.

3. Error Detection: Our models are trained to spot common billing and coding errors, such as missing modifiers, incorrect place of service, and medical necessity issues, and suggest corrections in real time.

4. Payer-Specific Optimization: By fine-tuning on historical claims data from each payer, our models learn the unique documentation and coding requirements of each insurer and can adapt claims accordingly to minimize denials.

5. Appeals Generation: For denied claims, our models can automatically generate compelling appeal letters by extracting relevant clinical evidence from the medical record and crafting persuasive arguments based on payer guidelines.

In order to safeguard the confidentiality and integrity of sensitive patient information, we implement cutting-edge methodologies such as differential privacy and federated learning throughout the model training phase. By doing so, we are able to gain insights from a wide range of provider data without ever having to access or centralize protected health information directly.

Philippe: worked as a lead AI Engineer at a software company and a current student at HBS.

Lotus: an Engineering student at Brown University and worked as a frontend engineer at UofT’s research labs.

Yiko: an AI Product Designer at Harvard Engineering School, and worked at IBM’s AI Research lab. 

Chris Giardina: an MD/PhD and post-doc at Harvard Med School performing AI research with clinical data models

Mackenzie: a Master’s in Public Health student at Harvard, and healthcare consultant that’s focused in revenue cycle management.

Alisha: worked as an Engineer at NASA and Apple and a current student at HBS.

4 months

Diversity, equity, and inclusion are deeply ingrained values at Billy AI. Not only should our solution be helpful to large health systems, but it should also be accessible to providers serving underserved communities. To that end, we will make sure more than 60% of our trial locations will be at community health centers and small primary care clinics located in low-income urban and rural regions. In addition, we partner with organizations like the National Association of Community Health Centers to understand and address the unique billing challenges faced by these providers.

We are proud of the gender, racial, ethnic, and professional diversity that exists within our founding team. From diverse immigrant, Asian American and Hispanic origins, three of our six founders are women. Chris brings the patient's perspective as a physician, while Mackenzie represents public health expertise. Meanwhile, Yiko and Lotus offer product and engineering prowess, and Alisha and Philippe contribute business and entrepreneurial experience. 

We have decided against using hierarchical titles in order to promote inclusivity and instead give everyone the opportunity to shine in their area of expertise. Important decisions are made through consensus, and there are frequent open forums where people can voice their concerns and offer suggestions. Furthermore, psychological safety is a key priority, with team members trained on active listening, unconscious bias, and how to be an upstander.

Moving forward, we are implementing a "Rooney Rule" to guarantee that we conduct interviews with a broad pool of candidates for all open positions, with an emphasis on attracting female sales leaders and engineers. To further our goal of creating a pipeline of diverse talent, we are also looking into potential collaborations with organizations such as Hack.Diversity, Smarter in the City, and Black Girls Code. We are investing in a comprehensive benefits package that includes generous parental leave, flexible hours, and location-agnostic remuneration because we know that diversity is about more than just numbers.

Our ultimate goal is to have an AI workforce and models as diverse as the communities and patients we help. By making DEI a priority from the beginning, we aim to strengthen our company and make a tiny contribution to a more equitable and inclusive health IT sector. Although we anticipate making some blunders, we are dedicated to gaining knowledge, refining our approach, and keeping ourselves responsible in our pursuit of creating an inclusive environment where all individuals can flourish.

By doing market research, we discovered that the average U.S. hospital loses $5M annually to denied claims, and 65% of denials are never even appealed. Our combined expertise in hospital operations, data science, and machine learning led us to see a huge untapped potential for artificial intelligence in solving a pressing yet sometimes neglected issue.

Over the past four months, we've conducted 30+ interviews with revenue cycle leaders, billers, coders, and practice managers. They unanimously validated the need for a more intelligent, automated claims management solution. Quotes like "I spend half my day on hold with insurance companies" and "We don't have time to appeal every denial" made it clear that the status quo is broken.

Users are enthusiastic about Billy AI's ability to streamline their workflows, lessen the load of administrative tasks, and boost revenue collection. Thus, we are providing a solution that all the pain points observed will be solved.

Specifically, Billy AI operates on a SaaS model, charging healthcare providers a monthly subscription fee based on the volume of claims processed. We offer 3 pricing tiers:

- Physician practices: $250/provider/month for up to 1000 claims; $0.10 per additional claim

- Billing companies: $10K/month for up to 10K claims; $0.50 per additional claim

- Hospitals/health systems: Custom enterprise pricing starting at $100K/year

We also charge a one-time implementation fee (typically $5-50K) for data integration, workflow configuration, and staff training.

At scale, we estimate the market opportunity as follows:

- Physician practices: 250K doctors x $3K/year = $750M ARR

- Billing companies: 400 firms x $120K/year = $48M ARR

- Hospitals: 5000 hospitals x $250K/year = $1.25B ARR

We aim to capture 5% of the market within 5 years, which equates to $100M+ in revenue.

Our plan for becoming financially sustainable will be modeled after a proven business model structure that has been successfully employed by our competitors and which has been well-received by customers as evidenced in customer interviews.

To date, we have demonstrated some traction and market validation for our solution:

• Validate the solution with 50+ billers/coders by showing them the prototype

• Closed three pilot contracts and will have them pay when the product is ready to use

Looking ahead, we forecast an annual revenue run rate of $2M within the next 12 months as we expand our customer base and solution footprint. At an average revenue per user (ARPU) of $2,000 per month, this equates to roughly 80-100 providers on the platform. We expect to be cash flow breakeven in Q4 of 2025 as we cross the $500K MRR mark.

To fuel our next stage of growth, we plan to bootstrap the company without turning to VC firms for funding as we strongly believe that our profit will be sufficient for us to grow the company quickly.

Ultimately, our vision is to become the trusted LLMs-based AI partner for healthcare financial operations - a $100B+ addressable market. By driving unprecedented efficiency and transparency in medical billing, we aim to reduce administrative waste, improve provider economics, and unlock resources for reinvestment in patient care.