‍

Healthcare documentation is going through its biggest shift in twenty years. Two distinct approaches are competing for budget, attention, and clinician time: ambient AI scribes and agentic documentation. They sound similar. They are not the same thing. And picking the wrong one for your practice can cost months of wasted implementation effort.

This post breaks down what each approach actually is, where each one shines, where each one struggles, and how to choose the right fit for your size, specialty, and workflow.

The Short Version

Ambient AI scribes listen to a clinician and patient conversation and produce a clinical note. That note goes into the EHR after the clinician reviews and signs it. Examples include Abridge, Nuance DAX Copilot, Suki, and Augmedix.

Agentic documentation does more. It listens to the same conversation, drafts the note, and then takes additional actions: pulls orders, drafts prior authorizations, codes the encounter, queues follow-up appointments, sends patient education, and flags care gaps. The note is one output among several.

Both approaches reduce documentation burden. They reduce it in different ways, and they fit different practice profiles.

What an Ambient AI Scribe Actually Does

An ambient scribe sits in the room (physically as a phone or laptop, or virtually as a video call participant). It captures the conversation, transcribes it, and converts the transcript into a structured clinical note in your preferred format (SOAP, H&P, progress note, procedure note).

The clinician reviews the draft, makes edits, and signs it. Time saved comes mostly from not having to type or dictate the note. Reported time savings cluster around 60 to 90 minutes per clinical day, with the highest savings in conversation-heavy specialties like primary care, behavioral health, and internal medicine.

Most modern ambient scribes also handle a few adjacent tasks: tagging patient instructions, summarizing for the patient portal, drafting the assessment and plan in a structured way. But the core capability is documentation.

The trade-offs:

• High value, fast to implement

• Limited scope, fixes one (large) problem

• Workflow is largely unchanged outside of dictation and typing

• Clinician still drives every other downstream action

What Agentic Documentation Actually Does

Agentic documentation starts with the same audio capture and transcription. From there, it does more. It reasons across the encounter, the chart history, and the practice's clinical and administrative workflow, and it takes actions across multiple systems.

Here is what an agentic flow can look like during a single primary care visit:

• Listens to the visit and drafts the note

• Identifies that the patient's last A1C was over a year ago and queues a lab order for clinician approval

• Notices a discussed referral to cardiology and drafts the referral letter with relevant chart context

• Recognizes a controlled substance prescription and starts the prior authorization

• Identifies that the patient is overdue for a colonoscopy and adds a care gap reminder

• Drafts patient education on the new diagnosis and queues it for portal release

• Codes the encounter at the appropriate level and flags any documentation gaps that would lower reimbursement

Each of these still goes through clinician review, but the clinician is reviewing finished work rather than starting from scratch. Time savings are larger but more variable, and adoption is more sensitive to workflow design.

The trade-offs:

• Higher value when it works

• Slower to implement

• More integration depth required (EHR, orders, coding, scheduling, patient portal)

• More compliance and safety considerations

• Workflow changes for clinicians, billers, and care coordinators

‍

Where Ambient Scribes Win

Ambient scribes are the right answer for most practices today, especially in these situations:

Solo and small group practices. The implementation overhead of agentic systems is hard to justify with five or fewer clinicians. Ambient scribes deliver most of the documentation relief with a fraction of the integration work.

Specialties with relatively simple downstream workflows. Behavioral health, dermatology, ophthalmology, and similar specialties have less complex order and referral patterns. Most of the win is getting the note written.

Practices on lighter EHR contracts. If your EHR does not offer deep API access or your contract limits write-back, a scribe is feasible while agentic documentation may not be.

Teams not ready for workflow change. Ambient scribes preserve the existing workflow. Clinicians keep doing what they did, just without typing. Adoption is faster because nothing else changes.

Fee-for-service environments where the bottleneck is throughput. If you bill volume-based and the constraint is documentation drag on patient throughput, an ambient scribe directly fixes the binding constraint.

Where Agentic Documentation Wins

Agentic documentation earns its complexity in environments where the downstream workflow matters as much as the note:

Value-based care contracts. When reimbursement depends on closing care gaps, capturing diagnoses accurately, and managing populations, an agent that surfaces gaps, drafts referrals, and ensures coding accuracy is worth far more than a scribe.

Large multi-specialty groups. Coordination across specialties is itself a workflow. An agentic system can surface relevant chart context, queue specialist referrals with the right information, and reduce the back-and-forth that wastes clinician time.

Practices struggling with revenue cycle. If denials, missed charges, or coding errors are eating margin, the coding and pre-bill scrubbing capabilities of an agentic system pay for themselves.

Health systems with prior authorization burden. Prior auth volume in cardiology, oncology, orthopedics, and similar specialties is brutal. An agentic system that drafts and submits PAs based on the encounter is a different category of value than note-taking.

Integrated care models. Care management, chronic disease programs, and team-based care benefit when the documentation system can fan out tasks to the right team members rather than dropping everything on the physician.

Where Each Approach Falls Short

Honest weaknesses on both sides.

Ambient scribes:

• Limited to documentation. Everything else still has to be done by humans.

• Note quality varies by specialty and accent. Surgical and procedural specialties get less benefit.

• Some practices report the time savings shift to "review time" rather than disappearing entirely.

• They do not help with coding accuracy, denials, or downstream tasks.

• Vendor lock-in is real. Switching is painful once clinicians are trained on a specific tool.

Agentic documentation:

• Implementation timelines stretch when EHR integration is harder than expected.

• More moving parts mean more failure modes. A broken integration can stop multiple workflows at once.

• Higher compliance burden. Every action the agent takes is auditable, and every action needs human approval pathways.

• Adoption is harder. Clinicians, billers, and coordinators all need to change how they work.

• Pricing is less standardized. Quotes vary widely between vendors and across feature sets.

• Off-the-shelf agentic systems are still maturing. Custom builds carry their own risk and cost.

A Practical Decision Framework

Use this as a starting point for your evaluation.

Choose an ambient scribe if:

• You have fewer than 25 clinicians

• Your main pain is documentation time, not coding or downstream workflow

• You bill primarily fee-for-service

• Your EHR integration options are limited

• You need a win in the next quarter, not next year

Choose agentic documentation if:

• You have 50 or more clinicians, or a multi-specialty footprint

• You operate under value-based contracts or risk-bearing arrangements

• Coding accuracy, denials, or care gap closure are in your top three problems

• You have engineering or partner resources for a 6 to 12 month implementation

• You can fund a longer payback period in exchange for higher long-term value

Consider a phased approach if:

• You want the documentation win immediately

• You can plan for agentic capabilities to layer on later

• You can pick an ambient scribe vendor with a roadmap into agentic capabilities, or design a build that can evolve

The Phased Path Most Practices Should Consider

Many practices benefit from sequencing rather than choosing.

Phase 1: Ambient scribe (months 1 to 6). Implement the scribe, drive clinician adoption, and measure documentation time savings. Build trust in AI-generated content while the stakes are limited to a note that the clinician reviews.

Phase 2: Add coding and chart prep (months 6 to 12). Layer in computer-assisted coding, pre-visit chart summaries, and documentation gap detection. These are adjacent to documentation but begin moving toward agentic behavior.

Phase 3: Add agentic workflows (year 2). With clinical trust established, begin adding orders, referrals, prior authorizations, and patient communication. Each new agent capability gets its own pilot, success metrics, and approval pathway.

This staging avoids the all-or-nothing risk of a big-bang agentic deployment, and it spreads the change management cost across two budget cycles.

What Both Approaches Require to Work

A few things separate successful deployments from the ones that quietly get shelved.

Deep EHR integration. Surface integration limits both approaches. FHIR-based access, write-back permissions, and structured data extraction are non-negotiable for serious value.

Human-in-the-loop design. No clinical action should happen without clinician approval. The agent drafts, the clinician decides.

Clear success metrics. Documentation time, after-hours work, denial rate, coding accuracy, clinician satisfaction. Pick the metrics, baseline them, and measure.

Strong clinician onboarding. Even the best tool fails without proper training and a feedback loop.

Compliance scaffolding. PHI handling, audit logs, BAA coverage, and access control are required regardless of approach.

A retraining and improvement cycle. Models drift. Workflows change. Clinical guidelines update. Your system needs to evolve.

Common Mistakes Practices Make

A few patterns show up across deployments that go sideways.

• Buying based on demo polish rather than specialty fit. A polished demo on a generic primary care visit tells you very little about how the tool handles your dermatology procedures or your cardiology consults.

• Skipping the baseline measurement. Without before-and-after data on documentation time, denials, or after-hours work, you cannot prove value or course-correct.

• Underestimating change management. Tools that change clinician workflow need real onboarding, real feedback loops, and real executive sponsorship.

• Treating it as an IT project. Documentation tools are clinical tools. They need clinical owners, not just technical owners.

• Locking in a single vendor too early. Negotiate exit terms and data portability up front, especially for agentic systems where vendor lock-in is deeper.

Conclusion

Ambient AI scribes and agentic documentation are not competing answers to the same question, they are answers to different questions. If your problem is documentation burden, an ambient scribe is the simpler, faster, and lower-risk path, and it works well across most outpatient practices today. If your problem is the entire encounter workflow, including coding, referrals, prior auth, and care gap closure, an agentic system can deliver substantially more value, with substantially more implementation work to match. Most practices benefit from starting with an ambient scribe to build trust and adoption, then layering agentic capabilities once the foundation is solid. Whichever path you pick, the deployments that work share the same fundamentals: deep EHR integration, human-in-the-loop design, clear success metrics, and a vendor or partner who treats healthcare seriously rather than as a vertical they added last quarter.

How We Think About This at Cabot

The honest answer to "ambient or agentic" depends on your specialty mix, your EHR, your contracts, and how your clinicians actually work today. That is why we do not lead with a product pitch when practices ask us this question. We start by mapping the workflow, the pain points, and the realistic constraints, and then recommend the lightest-weight thing that solves the actual problem.

Sometimes that means a clean ambient scribe deployment with a thin layer of EHR write-back, live in a few weeks. Sometimes it means a phased agentic build that starts with documentation and grows into coding, referrals, and prior auth over a year. Sometimes it means we tell a practice an off-the-shelf vendor will serve them better than a custom build, and we walk away from the project. Healthcare is the only industry we work in, so the answer that protects the relationship long-term matters more to us than any single engagement.

When a custom build is the right answer, the way we work tends to look the same: HIPAA-aligned architecture from the first commit, FHIR and HL7 integration handled by people who have done it before, human-in-the-loop checkpoints on every clinical action, and outcomes instrumented so the practice can see what is actually changing month over month. The ISO 27001 certification, the Microsoft partnership, and the near-shore plus offshore delivery model are all in service of that. They are the boring infrastructure. The interesting work is figuring out the right answer for your practice.

If you are weighing this decision, we are happy to talk through it.

‍