The following was originally published on LinkedIn in July 2025.
By Matthew Ko, DeepScribe CEO and Co-founder
In my time building DeepScribe and working shoulder-to-shoulder with oncology groups across the country, one reality has become abundantly clear: the practice of oncology is an unforgiving intersection of high-stakes decision-making and relentless operational demand.
Of course, every medical specialty has its own profound challenges. But oncology represents a uniquely dense burden of logistical, administrative, and operational complexity.
This is the paradox of cancer care. The work is sacred, but the system fails the people who do it every day.
In this article, I’d like to briefly examine parts of that system, and offer ways that ambient AI technology remedies the excessive work (and costs and risks) that overburdens oncology teams.
Breaking Down What’s Broken
At its core, oncology is both science and choreography. It’s an intricate coordination of people, decisions, records, and technology. With every new therapy, new guideline, and new patient need, the choreography becomes more complex.
Large oncology groups can have 10 support staff for every one oncologist, managing everything from clinical coordination to a mountain of payer-related work. For perspective, primary care averages about three to five support staff per physician.
The oncologist and their team are left to assemble all the pieces, filling in gaps from their memory, PDF files, and even hallway conversations.
We’ve seen firsthand how the current oncology infrastructure wasn’t built to handle this level of complexity and detail. EHRs are designed for record-keeping, not contextual decision-making. Practice management systems focus primarily on scheduling and billing, not burden reduction.
These are multiple systems capturing snippets, not symphonies. And increasingly, the oncologist and their team are left to assemble all the pieces, filling in gaps from their memory, PDF files, and even hallway conversations.
This is not a data problem. This is a workflow problem. And it requires a shift in how we think about infrastructure.
The Costs and Risks of the Current System
Oncology practices are uniquely burdened by a value chain that spans vast clinical variation, multidisciplinary coordination, evolving guidelines, and seemingly endless documentation requirements. It all chips away at clinician and financial wellbeing, while opening the door to missteps or missed information.
If we want to preserve the humanity of this work, we need a new infrastructure.
The administrative and operational tasks that make up the patient journey account for roughly 80% of overhead at oncology practices. Two of the most significant categories, referral and scheduling, and patient monitoring and follow-up, each make up about one-fifth of overhead.
If we want to preserve the humanity of this work — and make it scalable — we need a new infrastructure. One that reduces the cost of complexity. One that integrates rather than fragments. One that centers on ambient technology.
Mapping the Oncology Value Chain
To understand the potential opportunities, we must understand the work. The oncology value chain spans two critical domains: the patient’s clinical journey, and the practice’s operational engine.
While these are obvious to those working in oncology, it’s important to visualize the volume of work and the interconnectedness across parts.
For each link of the chain, I’ve included the opportunities that ambient technology offers, the attainable promise beyond AI documentation.
The patient journey: Required jobs and ambient opportunities
Referrals and Scheduling
The Work
Intake teams juggle incoming referrals, insurance verification, external records, and multi-specialty coordination. This is far more than just data entry and management — it’s a constant negotiation of time, availability, and urgency.
The Opportunity
Let’s start with the patient conversation, the foundation of what we call the Ambient Operating System at DeepScribe. As the system listens to real-time clinical conversations, it detects the need for a referral or an order, and then automates the next step. The system could generate the order, or surface appointment or chair availability based on predictive scheduling.
During the encounter and at call centers, ambient technology can also assess acuity and triage high-risk cases by automatically flagging urgent referrals.
Initial Consultation and Diagnostics
The Work
Clinicians synthesize patient histories that are scattered across PDFs, pathology reports, scanned faxes, and EHR notes. The goal: a coherent diagnostic picture.
The Opportunity
Ambient AI can bring the patient’s journey into organized focus. It can automatically surface relevant diagnostics from multiple sources, contextually summarize those findings, and create structured inputs for downstream clinical decision-making.
Treatment Planning
The Work
Oncologists must harmonize a wide swath of information into a single, executable care plan: tumor board decisions, payer requirements, guidelines, genomics, and, perhaps most critically, the patient's voice.
The Opportunity
As part of the flow of care, the Ambient Operating System can capture and integrate these multi-stakeholder inputs in real time, recommend evidence-based pathways, and automatically initiate prior authorization.
As an example, during tumor boards, the system can document and curate insights directly from discussions. This is more than just precision note-taking; it’s preserving context and decision rationale as a key component of the treatment planning process.
Therapy Administration
The Work
From chemo chair logistics to infusion orders and emergent symptom response, this is where safety, timeliness, and precision intersect.
The Opportunity
In addition to documenting visit conversations, ambient technology can flag any order discrepancies, and detect signals in conversation that indicate acute complications, generating alerts in real time.
When integrated with patient-accessible chatbots, ambient AI can also escalate issues automatically, greatly improving the chance at early intervention and a reduction of avoidable hospitalizations.
Monitoring and Follow-Up
The Work
Teams are assessing treatment response, tracking labs and imaging, monitoring for side effects, and adjusting care plans based on changing needs.
The Opportunity
Ambient AI can detect deterioration signals, escalate concerns, and ensure continuity across handoffs. It can even prompt oncologists to initiate end-of-life discussions with clinical sensitivity.
Supportive Care and Survivorship
The Work
This is the management of a unique combination of needs that must take the whole patient into account: Comorbidities, psychosocial concerns, symptom burden, and the long-term effects of treatment. This stage may also involve coordination with primary care, nutrition, pain management, mental health, or financial counseling. It also includes documenting survivorship care plans and navigating transitions in care.
The Opportunity
Ambient systems ensure these touchpoints are all captured, documented, and actioned — without additional clinician effort.
Operational workflows
While oncology’s clinical demands are highly visible, behind-the-scenes operational workflows are just as essential — and often just as strained.
Many of these functions are deeply connected to both clinical outcomes and financial sustainability. Yet, they remain marked by fragmentation and manual processes, a combination that leads to missed opportunities.
Revenue Cycle Management (RCM)
In oncology, submitting claims, managing denials, and optimizing billing codes are rarely straightforward. The challenges of coding chemotherapy regimens and care plans, along with proper staging documentation, create ample room for error.
Teams must effectively track pre-authorizations — often a literal queue of them, especially for new patients — correct rejected claims, and ensure coding specificity, all while managing payer requirements. Errors and delays affect far more than revenue; they can halt care and disrupt operations.
The Opportunity
This is already in use by DeepScribe customers: The Ambient Operating System listens to the encounter and then automatically suggests ICD-10, HCC, and E/M codes in real time. It can even pre-empt denials by flagging any gaps or undercoding before submission — effectively reducing denial rates and shortening the revenue cycle.
Ambient documentation also provides evidence from the visit transcript for any claims adjudication.
Clinical Research and Pharma Partnerships
Oncology practices are vital engines of clinical trial activity and pharmaceutical innovation. But recruiting eligible patients, tracking trial outcomes, and coordinating with sponsors or clinical research organizations (CROs) is labor-intensive.
Research coordinators must parse through siloed data sources to identify potential candidates and then fulfill reporting obligations, often relying on manual chart review or ad hoc communication with clinical teams. The result is slowed enrollment, less access to trials, and potentially limiting a practice’s ability to scale its research footprint.
The Opportunity
Think of ambient AI as a real-time eligibility engine: detecting biomarkers, disease stage, or patient interest, and then surfacing potential trials in the moment of care. Of course, this doesn’t replace auto-matching platforms and patient-facing registries; it complements and aids them by embedding trial awareness directly into the clinical encounter.
Payer Compliance and Quality Reporting
Practices must meet various value-based care benchmarks, regularly submit measures (HEDIS, STAR, MIPS), and respond to audits. It typically requires pulling data from multiple systems and interpreting evolving regulatory criteria, with significant administrative oversight.
Compliance officers and quality teams spend countless hours to ensure documentation is complete and codified correctly — often retroactively. This leads to administrative fatigue, but also risk exposure if gaps or inconsistencies are found.
The Opportunity
Ambient systems can surface quality prompts at the point of care, supporting one of the most critical elements of cancer care delivery: consistency.
The technology can also automatically capture metrics, keeping teams compliant without the weight of the administrative burden.
This is far from a complete analysis. But even with this view, the pattern across the chain is clear. Each link is a job to be done, so often bogged down by disconnected systems and the need to do many things manually. Yet, much of this work can be done without overtaxing resources.
The challenge is obvious and the technology is available to meet it.
Applying Ambient Technology Across the Oncology Value Chain
Ambient AI offers a new approach to the value chain: sitting alongside clinicians, listening to the clinical encounter, and integrating deeply into the workflows that matter most. It’s not just transcription. It’s infrastructure.
We’re just in the early stages of what’s possible. As my colleague Dr. Dean Dalili says, “We’re only in the third inning.”
When we talk about an Ambient Operating System for oncology— its current and potential applications — we’re talking about a platform that understands key areas of the value chain and can intelligently support each step.
It’s important to remember that we’re just in the early stages of what’s possible with the system. As my colleague Dr. Dean Dalili says, “We’re only in the third inning.”
The Ambient Operating System doesn’t replace humans. It elevates them. It creates structure from conversation. It detects signal within noise. And, most critically, it integrates seamlessly into the cadence of care.
Oncology Needs Its Own Ambient System
Unfortunately, the inefficiencies of the oncology value chain have become the norm, and each comes with a cost: financial, human, or both.
The two-way interaction between ambient technology and the EHR ensures data that delivers benefits beyond documentation.
At DeepScribe, we’ve learned that, from a technology standpoint, success in oncology requires not just intentional, collaborative design, but also deep integration.
Through our preferred partnerships with Epic, Ontada, and Flatiron Health, DeepScribe oncology clinicians get direct integrations with MyChart, iKnowMed, and OncoEMR systems. The two-way interaction between ambient technology and the EHR ensures a precise, intelligent clinical note — and, just as important, data that delivers benefits beyond documentation.
With direct access to ambient conversations, EHR data, and practice management systems, we can:
- Align visit summaries with financial workflows
- Nudge staff to address documentation needs within the conversation during high-complexity encounters
- Code with relevance and accuracy, and identify potential gaps ahead of time
- Surface trial eligibility criteria in real time
These are only some of the initial values of the Ambient Operating System for oncology. With technology integration moving vertically rather than horizontally, even more integrated high-impact solutions will be created across the scope of oncology work.
Final Thoughts: The Opportunity Ahead
Every oncology leader I’ve spoken with wants the same things: better care delivery, less administrative burden, and a financially viable path forward. Achieving that requires a shift from piecemeal processes to centralized systems.
From ambient AI as a feature to ambient AI as the foundation.
Designing a system with the full oncology value chain in mind can reduce the ongoing and mounting cost of complexity. We can elevate the clinician experience while unlocking new levels of operational efficiency.
All the while, we maintain the north-star goal of all healthcare technologists: ensuring more connected time with patients and the quality of care they receive.
As I mentioned earlier, this look at the oncology value chain is far from a complete review. It's intended to inspire thought as to how one of the most rapidly adopted technologies in healthcare delivery today can make a real difference in oncology beyond its current common use case.
In the weeks to come, I'll delve deeper into the categories of work mentioned above and the application of ambient technology solutions to their inherent problems. (I'll add links to those posts here.)
For now, in the world of oncology, the Ambient Operating System is both a current presence and a future ideal. It’s already here and continuing to emerge.
text
