Social Impact AI Lab – New York (SIALNY)

Marty Elisco is the Founder and CEO of Augmented Intelligence, a software and services company that seeks to transform human services by providing innovative models that improve practice and operations with artificial intelligence and cognitive computing solutions, helping social services organizations evolve from case management to care management. Prior to Augmented Intelligence, Marty held leadership positions in healthcare, government and social services technology sales, and in telecommunications engineering. He is also a longstanding member of his village’s community relations board. Marty has a BS in Engineering from Duke University and an MBA from the Kellogg School of Management at Northwestern University. He currently resides in the Chicago suburbs.

Social services organizations – government and nonprofit child welfare, behavioral health, housing services, employment services, etc. – collect massive amounts of information. 80% of this data is unstructured data – case notes – the primary data for understanding needs of vulnerable populations. However, making sense of unstructured data is extremely difficult, preventing staff from obtaining sufficient understanding about individuals and families they care for.

Social Impact AI Lab – New York (SIALNY) builds AI and Natural Language Processing products that mine case notes to provide insights and alerts to staff to drastically improve their ability to deliver care and monitor services delivered by their agency.

Last year, we piloted our products, proved their feasibility and validated their impact. Should we be selected for an MIT Elevate award, we would have the resources to scale our products and expand their reach throughout our organizations and to child welfare organizations across NY.

Social services organizations – government and nonprofit – have a key problem that inhibits their ability to deliver services and assess investments in underserved communities. They are extremely limited in their ability to consume to data to inform decisions about (1) individuals receiving care and (2) programs and infrastructure that may benefit communities they serve, specifically in the areas of child welfare, behavioral health, housing services, employment services and other community-based services. There are approximately 700,000 caseworkers who provide care for 60 million people in every community in the US, and nearly all experience this issue.

These organizations collect massive amounts of information. Over 80% of this data is unstructured case notes – thousands of pages for individuals and millions across an agency. It is the primary data for understanding individual and community needs. However, making sense of it is extremely difficult, preventing staff from obtaining sufficient understanding about those they care for, and limiting the ability of leadership to oversee services delivered. This is a longstanding, monumental problem.

Technology has not improved to address this. Nearly all decision support technology strictly uses structured data – only 20% of agency data – to guide decision making, ignoring narrative data altogether.

We are building, piloting and scaling AI products that process vast amounts of case notes and supporting data to improve decision making. First, we draw data from government case management systems and electronic medical records systems (EMRs). Then, our products methodically provide insights and alerts to users.

We use natural language processing (NLP) – a key AI discipline where computers are used to understand the written word. Using our customers’ historical data, we train language models to understand social services, then apply those models to live data.

We have two initial products that are developed and now need to scale. The first is our case dashboard used by front-line staff to gain understanding of a case. We present the social worker with risks, strengths, social determinants of health, and social interaction data, as well as the progression of these items over time.

The second product is our supervision dashboard used by supervisors, quality improvement specialists, and leadership to monitor services delivered across the organization, improve compliance and gather data to make policy and investment decisions. This dashboard presents a library of NLP queries that probe narrative for quality, compliance and performance issues – the key input required to make decisions.

Across the US, approximately 700,000 caseworkers support the lives of approximately 60 million Americans in underserved communities. These communities are urban, rural and everything in between, and all experience significant deficiencies regarding access to employment, physical and mental health services, education, transportation, sources of food and many other factors commonly known as social determinants of health (SDOH).

Our project supports the social services organizations that directly support these communities. These organizations provide a wide range of services to address the factors listed above, whether it is through individualized care or through the investment in programs and infrastructure. By providing AI/NLP products to front-line staff to help them make data-driven decisions in ways not previously possible, the organizations improve the quality of services they provide, their capacity to provide services, and their capability to make investment decisions for their communities.

Through the improvement in quality, capacity and decision-making ability, a measurable improvement in outcomes of individuals who receive these services – employment, physical and mental health, child and domestic violence prevention, etc. – can be achieved en masse for individuals, families and communities.

Our project aligns directly with the dimension “Elevating opportunities for all people, especially those who are traditionally left behind” because our products are used by the social services organizations that provide services to vulnerable individuals in underserved communities, which is another way to describe “those who are traditionally left behind.” Our project elevates opportunities for individuals in these communities by improving the quality and capacity of the social services organizations that directly support them, and by helping these organizations to improve conditions in their communities by making data-driven investments to improve employment, education, transportation and many other key social factors.

Our project is the result of collaboration and partnership building over the past four years. In 2016, the organization Stewards of Change – a national social services think tank – researched use cases and viability of AI in social services. To do this, they convened focus groups around the country consisting of practitioners at federal, state and local levels. From this work, the need to apply AI/NLP to case notes was identified. In September 2018, Stewards of Change worked with Marty Elisco to form Augmented Intelligence to address this need.

In parallel, three of the largest child welfare nonprofits in NYC – the New York Foundling, SCO Family of Services and MercyFirst – began exploring how AI could be applied in their agencies. In Summer 2018, they executed proof-of-concept projects to assess processes in their organizations. This work provided valuable insight and proved they could derive value from AI.

In January 2019, Augmented Intelligence was connected with these agencies via Stewards of Change. It was immediately clear that all four agencies were addressing this need from different perspectives, and they formed the SIALNY partnership. Since then, SIALNY has been working to develop, pilot and scale AI applications for social services.

The social services industry is massively underserved from a technology perspective but is paradoxically the industry where technological advances can have the most immediate impact on the quality of life for underserved communities. Cutting edge technologists may be trepidatious of this market due to false perceptions of long sales cycles, small budgets and lack of desire to innovate.

All of us at SIALNY are either former social workers, have family who were social services practitioners, or were recipients of social services themselves. We understand innately how tech can address longstanding problems in this space. More importantly, we dedicate our lives to reducing child abuse and domestic violence while improving the health and well-being of children and families in our communities. We each bring our unique skills to SIALNY – sustainable business formation, technology conceptualization and development, feasibility assessment, voice-of-the-customer, etc. – to translate resources into sustainable social impact in our communities.

In other words, we are passionate about applying our professional abilities and our lives’ work to reliably develop and commercialize technology from the ground up to meet the needs of the social services industry, an industry that is, more often than not, neglected by cutting-edge technology providers.

AI projects in social services have primarily been one-off or research projects to answer specific questions or support specific workflows. Our partnership, alternatively, delivers scalable products for a wide array of use cases, and builds the business to expand and sustain products at social services organizations the country. Our team possess the skillset to execute this.

Marty Elisco – CEO, Augmented Intelligence -- Marty brings deep technology and operational experience from leadership positions at major technology companies, including Motorola and Zebra.

Besa Bauta – Chief Data Officer, Mercy First – Besa runs data integration and analytics. She has a health services research PhD, masters in clinical social work and masters in global public health from NYU.

Arik Hill – Chief Information Officer, The New York Foundling – Arik brings significant international consulting expertise driving digital transformation at many organizations.

Tom Castelnuovo – Director of Data Management, SCO Family of Services – Tom chairs NY’s child welfare information group and has 30+ years of child welfare systems experience. He has a masters of city planning from MIT and an MBA from Yale.

Jim Lindstrom – VP of Product, Augmented Intelligence – Jim has led many startup and AI product teams in many industries including the child welfare space.

Logan Courtney – Chief Data Scientist, Augmented Intelligence – Logan received his PhD in Engineering with a focus on machine learning and data science. He is the lead inventor of multiple patents and has extensive experience teaching deep learning at the graduate level.

Access to data is a major obstacle in almost any AI project. In the social services space, where government is a key stakeholder, there are a wide range of technical and bureaucratic roadblocks that must be overcome to acquire access to data.

In our project, the narrative data that we seek is contained in electronic medical records systems and government case management systems. To obtain this data, we have invested significant resources to build our infrastructure to be HIPAA compliant and abide by all government security requirements. We have therefore been able to acquire access to key electronic medical records systems. Through this, we were able to build initial versions of our products.

However, we have not yet obtained access to government case management systems, and this remains a key obstacle for us. To overcome this, over the past year, we have been engaging with key government stakeholders at state and local levels them to pursue processes to gain access to this data. This is not yet complete. Though lengthy, during this process, we have generated positive awareness for our products with government stakeholders and have learned that state and local government entities may be potential customers as well.

Earlier Marty Elisco’s career, he led engineering teams at Motorola, developing public safety radio communication equipment. In his free time, he recognized that even though the mission-critical radio communications products that Motorola built were designed for public safety personnel, those same products could be used by social services caseworkers to help them maintain their own safety and the safety of the children and families they care for when meeting with them in their own homes. However, Motorola did not serve this market.

Therefore, while working his day job as an engineer, Marty spent his free time developing a business plan for Motorola to address the social services market. He then spent a year pitching this plan to executives throughout Motorola, which included resource requests for product management, engineering, sales and other related business functions. Even though he was rejected by many executives along the way, he did not stop until his plan was approved.

Once his plan was approved, Marty transitioned from engineering to lead Motorola’s first social-services focused business unit, which continues to operate to this day.

Social Impact AI Lab – New York (SIALNY) is a partnership of three major NYC-based child welfare nonprofit agencies (The New York Foundling, SCO Family of Services and MercyFirst) and an AI software development organization (Augmented Intelligence).

Our work addresses the longstanding problem of social services practitioners not having the tools they need to maximize their ability to deliver care. Narrative data – the data that describes the experiences of individuals receiving social services – has not been utilized to inform care delivery, even though this data is widely recognized as paramount. This has been due to technology limitations, but which now can be overcome due to advancements in AI. This is the key innovation of SIALNY – to provide social services agencies with unprecedented insight and decision making ability borne through the application of AI and NLP to narrative data, which had not previously been possible prior to our work.

Our work has a straightforward and direct impact on humanity because the users of our tool provide direct service to individuals underserved populations to improve their quality of life, and to improve dire conditions in the communities they live.

We drive change because the products we develop help staff better understand the individuals they care for, better manage agency resources, and make smarter investment decisions. Due to this, the capacity and quality of care of social services organizations improves. In turn, the social outcomes and quality of life for vulnerable individuals improves, and in parallel, the conditions in the communities we serve improve.

Our work supports the social services practitioners who support vulnerable individuals in their communities. The social services practitioners who currently use our products serve 260,000 individuals. In one year, should we be able to expand more broadly in New York, we will be able to serve 1.5 million individuals. In five years, we expect to work with one out of every five states. Through this, we will reach 12 million individuals in vulnerable populations.

Our primary three social impact goals align directly with the goals of the social services organizations who use our products, and that is to improve safety and well being of the individuals and families they care for. There are specific sub-goals that these agencies and the government track, such as reduction in abuse, reduction in neglect, and improvement in permanency of children in foster care through adoption or through reunification with birth parents.

Our goal is to help agencies improve each of these metrics by 3%. We show that we achieve these goals by measuring the improvement in many discrete factors that roll up to the above goals. Some of these factors include the ability of agencies to reduce abuse, reduce neglect, support permanency, improve access to education, and improve success of evidence-based therapy sessions.

Our two key barriers are financial and market barriers that slow the adoption of our products. Regarding financial barriers, we have built our business from the ground up to meet the needs of social services customers, and that includes pricing that can be supported by operational budgets. In other words, we are built to be affordable by social services organizations. However, due to the perceived riskiness of adoption cutting edge technologies like ours, agencies are leery of purchasing our technology for the first time because they think it might fail, wasting their money.

Regarding market barriers, social services customers may worry that the AI software, as it processes case data, may fail to detect certain safety issues that are described in case notes, which may create a blind spot for the social services worker, increasing the possibility that risks aren’t addressed, which may increase the likelihood of future abuse. In other words, because the output is not perfect, agencies may believe it is too risky to adopt our products.

To address the financial barriers above, for any new customer, we spend five months piloting the software to validate that it works, and we provide the customer with cancellation terms in our contract should the pilot not be successful. Additionally, our product generates financial return that pays for the costs of purchasing it. Our products can be used to monitor compliance of documentation. When this is applied to billing documentation, we can improve agencies’ reimbursement rates from payors. This improvement in reimbursement is typically more than the entire cost of the project, meaning that in addition to the social returns we generate, we also generate financial returns.

To address the market barriers above, we share success stories from our current customers and, again, we pilot the software to make sure that the product works and the quality of output meets our customers’ needs. This typically addresses the worry that our AI might fail to identify certain case factors. Additionally, our products provide access to the entire case data, including the insights provided by the AI. So, should the AI fail to identify an issue, the same data that has always been available to the caseworker – the raw progress notes – continues to be available.

We partner with other organizations from go-to-market and knowledge sharing perspective. For example, we partner with the organization Inorupt to facilitate business opportunities and with COFCCA (Council of Family and Child Caring Agencies) to acquire input from subject matter experts regarding the functionality of our product and what capabilities are necessary to integrate with government case management systems. We are continuously seeking out new partnerships beyond our own SIALNY partnership to improve the value of our product, deliver the value to the social services industry, and support other impact organizations with similar goals to the best of our ability.

We deliver our software using an enterprise software-as-a-service business model – there is an annual cost to use our software. This model streamlines the ability of agencies to adopt our technology and enables our customers to operationalize the costs. It also enables them to align costs with the benefits that are received from our product, enable return-on-investment to be measure in a straightforward way.

Our primary means for funding our work is through grants and through the license fees associated with selling our products. Additionally, our revenue model is a hybrid of embedded and external. It is embedded because our products are used by our own organizations and it is external because we are selling the same products to other social services organizations. Our goal is to create a sustainable business where all positive cash flow is reinvested into product innovation.

In 2019, our initial product development was funded with $320,000 by our own agencies. We have been able to successfully develop, pilot and release our two AI products using these funds. The result of this investment is a product that has been internally validated by our own agencies and externally validated by other social services organizations that are evaluating the purchase of these products. However, additional funds are needed to improve product capabilities so that they are scalable across the industry.

As described above, we are funded through customer contracts and do not intend to raise money through debt or equity. However, to enable our business to scale more quickly so that the social benefits of our software can be acquired by other social services agencies, we are seeking grant and similar funding, as well as in-kind support associated with these funds, which is why we are applying to the MIT Elevate Prize.

Our expenses for 2020 are approximately $550,000 which primarily consist of salaries and cloud computing expenses.

We are applying to The Elevate Prize for many reasons. First, we believe our product directly aligns with the goal of the Prize to help organizations who directly support “those who are traditionally left behind.” This is the focus of our products and of our lives’ work. Therefore, in addition to the potential financial support, we believe that the MIT associates and affiliates, who would support us if we were fortunate enough to win, would be passionate in helping us grow by coaching us, connecting us with key people who can help grow our business, and by aligning with us as we develop marketing campaigns.

Given the alignment of our goals with the MIT Elevate Prize goals, we believe that program representatives would be uniquely qualified to support us in establishing connections and providing mentorship to grow our go-to-market partnerships, scale our business and expand internationally.

We are interested in partnering with the Health IT software development organizations – specifically electronic medical records vendors – that build the systems that house the data to which we apply our AI, so that we can collaborate with them from technical and go-to-market perspectives. Similarly, we are looking to partner with vendors of government case management systems to, again, collaborate from technical and go-to-market perspectives.