Ascertain is a NYC-based technology company whose mission is to provide healthcare enterprises with best-in-class automation quickly, safely, and effectively. Launched out of New York’s largest healthcare system (Northwell) by AI and automation experts with experience from Amazon, Microsoft, and Google, Ascertain is bringing AI Agents to support and augment medical professionals, starting in the domains like revenue cycle management, pharmacy, and transitions of care.

AI agent–also referred to as autonomous agents–are based on large language models (LLMs), but add additional features that allow these LLMs to power tasks related to automation. In contrast to the “chatbot” interface that many users of solutions like OpenAI’s “ChatGPT”, Google’s “Gemini”, or Anthropic’s “Claude,” AI agents use LLMs to do a wide variety of complex tasks, such as process documents, generate API calls, query databases, and even send faxes or make phone calls.

Ascertain offers an AI platform that primarily supports healthcare professionals–like case managers, social workers, pharmacy administrators–in their work creating, processing, and submitting documentation for things like prior authorization requests, summaries for care transitions, FMLA requests, and much more. This application is highly customizable for specific workflows; by the end of 2024, it will have been deployed across at least five different workflows.

This applicability to multiple administrative workflows is possible because underlying technology in Ascertain’s application gives it a very high level of configurability. First, we have created a proprietary framework called the Unified Agent Framework (UAF) which allows us to build, scale, and deploy AI agent configurations quickly and effectively in Enterprise Healthcare. Second, we’ve invested in developing a set of differentiated, highly performant agents that tend to be the most common and most critical components of automating administrative workflows (typically data ingestion and reasoning/adjudication against rules).

Effective multimodal models that natively support vision and text are enabling new types of use cases. Relevant for healthcare, where several legacy systems are still a part of many major workflows, these agents can be used to integrate through the user interface to automate components of workflows where legacy systems are a part. They also can be useful for other interfaces, including interfacing with web applications and portals.

We can usually optimize an existing agent graph even when working to automate a workflow in a new environment or use case. In short, we take a relevant graph structure that encodes a workflow (or part of a workflow) similar to one we’re trying to automate, we optimize prompts in each node in the chain for performance, and then finally employ parameter-efficient fine tuning for repetitive tasks if needed.

Ascertain does not need comprehensive access to your data warehouse. Ascertain is able to deploy its agents provided it understands the workflow and establishes the necessary integrations to the systems that are current dependencies for the given workflow.

Our tech stack was chosen to ensure robust security, scalability, and rapid development capabilities. Compliance requirements like HIPAA and SOC2 dictate much of the technology choices. We leverage Terraform to provision infrastructure allowing for cloud-agnostic or multi-cloud setups . We ensure robust scaling with Kubernetes as an industry-standard solution.

Ascertain has developed the ability to deploy LLM systems locally, putting users in control of their technology stack while ensuring data are kept private, reducing the risk of breaches. While others will increasingly have to rely on rate limiting, caching and request prioritization, thereby degrading the user experience, Ascertain will be capable of serving requests promptly and consistently ensuring a good user experience even during peak usage.

Following executive buy-in and formal approval, Ascertain worked with members of Northwell Health’s central care team to understand their workflow for gaining authorization for post-acute transitions of care. It then proceeded to build virtual employees that:

• Exhibit agency and take work off employees’ plates to demonstrate meaningful productivity gains.
• Ask for help and adjust according to feedback to ensure correct outcomes.
• Show their work for maximal observability and control.

In all instances, Ascertain works with the team to develop a proof-of-concept, carries out validation testing and user acceptance testing, and launches a pilot to track KPIs and return on investment. Our lifecycle is tuned for the healthcare enterprise and assures our partners are involved throughout the process to ensure successful delivery.

We’re excited to connect with you!

Contact us at hello@ascertain.com

Open-source frameworks are a great resource for researchers and early tech developers, but unfortunately are not often robust enough in production environments. Other proprietary agent frameworks are emerging, but they’ve been surprisingly limited in their functionality to date, and require some element of “lock-in” with a specific cloud platform or model.

We believe our technical approach–built from the ground up for healthcare enterprise customers–is best in class. It is distinguished by its enterprise-grade security and trustworthiness, ease of development, and a plug-and-play ecosystem enabling rapid deployment of features and use cases. For example, we support distributed execution natively, which ensures scalability and efficiency with the increasing demand, as well as high availability, high fault tolerance, high performance with reduced latency.

We are not just a “wrapper” around GPT. Besides several points of differentiation around how we implement our AI agents, we leverage multiple models, which include GPT as well as several others. We believe a robust, enterprise grade solution will provide the appropriate model with the appropriate input data at the right place and at the right time. This coordination is critical for successful use of AI in automation solutions; otherwise, the ultimate result is likely to be slow, costly, inaccurate, or some combination of these three.

Ascertain has developed approximately a dozen families of agents for various tasks, including database retrieval, rules adjudication, summarization, phone tree navigation, clinical entity validation, and medication reconciliation. We provide the capability to both extend existing agents and as well as configure new ones.

Due to our origin as a joint venture with Northwell Health, we've prioritized security and robustness from the very beginning. Consequently, Ascertain has implemented a comprehensive security infrastructure that includes:

• Electric AI MDM: Mobile Device Management solution.
• Perimeter 81 VPN: Virtual Private Network solution for secure remote access.
• Malwarebytes: Advanced malware protection.
• CrowdStrike Falcon Enterprise: Endpoint Detection and Response solution for enhanced threat detection and mitigation.

We comply with HIPAA standards and are currently undergoing SOC2 certification and HIPAA attestation.

Not at all! While this FAQ is meant to describe our technical underpinnings and some of the advantages and points of differentiation we have, we built our applications to be simple to integrate and use in a wide variety of settings, on any cloud, with a wide variety of systems that are often used in healthcare (e.g. your EHR, CRM, and other source systems).

Ascertain’s application is being used to automate the back-office operations associated with:

• Post-acute prior authorizations
• Specialty pharmacy prior authorizations
• 340B compliance and auditing
• Transitions of care medication education

We’re particularly focused on autonomously extracting relevant information from complex clinical documents to meet the needs of each of the use cases above.

Our largest deployment is with Northwell Health’s centralized transitions-of-care administration team, which currently handles more than 800 post-acute authorizations per month.

That’s great–we’re looking for innovative partners that want to expand what our technology is capable of! Typically, we work very closely with our partners to gather relevant inputs, conduct team interviews to understand workflows, and develop a proof-of-concept to quickly demonstrate value. So far we’ve been able to move as fast as most healthcare organizations have been able.