AI sidekick for physicians

An EHR companion for physicians that highlights red flags, missed care opportunities, potential coding gaps, low cost alternatives, accurate diagnosis & strengthen charting

Doctors spend most of their time working within EHRs, but often lack visibility into what’s happening across their care team. Luma serves as a smart sidekick that lives right inside the EHR, giving doctors an instant window into care activities, updates, and patient progress without ever leaving their workspace. It also provides high risk alerts & other care recommendation that improves documentation accuracy and reduce denials. By seamlessly connecting to Nova’s capabilities, Luma keeps doctors informed, coordinated, and in control—without breaking their clinical flow.

Client
–
Roles & Responsibilities
  • Market Research
  • Contextual Inquiry
  • Interaction & Visuals
  • A/B testing
  • Prototyping & Conversational Testing
Users
  • VBC / FFS physicians
Devices
  • Responsive Web

What is sidekick?

Well, its an AI powered side drawer that synthesises patient’s profile from across the sources & brings most crucial information at point of care. Helps physicians & care teams do accurate diagnosis & documentation without need of context switching.

Why we need a sidekick?

One word. Data fragmentation. Patient meet different doctors for different needs. Some clinic use siloed tools for patient engagement vs care management. Interop is still taken for granted. During a 30 mins encounter, physicians keep multiple tabs open to make sense of patient’s wholistic data. 

Context switching
5
+
app switches every hour
Out of sync care
5
mins
pre-visit sync per patient
Claims denial
30%
+
missed coding / care gaps

Doctors spend ~50–60% of their day inside EHRs yet lack visibility into what their care teams are doing. Coordination breakdowns lead to duplicated work, care gaps, and billing denials.

Shadowing with physicians

We conducted think-out-loud sessions while shadowing their real life usage in simulated dummy encounter for a chronic care patient.

# Observation Context What We Saw & Heard (Qualitative Insight) User Behavior Pattern Emotional Signal Opportunity for AI Sidecar
1 Note-taking & Charting During observations, doctors toggled between multiple screens — notes, labs, messages. Many stayed back after clinic hours to complete charting. Fragmented multitasking Frustration + Resignation Auto-draft summaries, inline note capture, contextual care thread view
2 Missed Screenings / Labs Doctors often learned about missed screenings or overdue labs only when patients revisited. Most relied on manual reminders from care teams. Reactive discovery Anxiety about missing care Many physicians skipped detailed coding mid-visit, deferring to billing staff, causing revenue delays.
3 Coding & Billing Many skipped detailed coding mid-visit; deferred to billing staff. Deferred admin work Avoidance + Dependency Suggest HCC/risk codes contextually before closing case
4 Denial Management Revenue managers shared that most denials came from documentation inconsistencies between physicians and care teams. Error correction loop Inefficiency fatigue Validate documentation live; AI check for coding/compliance before submission
5 Prescribing & Cost Doctors lacked instant visibility into patient affordability or formulary coverage during prescribing. Blind prescribing Empathy + Helplessness Show guideline-based affordable alternatives inline during prescription
6 SDOH Context SDOH data (housing, safety, transport) existed in EHR but buried in other modules, unseen during encounters. Hidden context Concern + Powerlessness Surface SDOH flags & actionable prompts for social follow-ups
7 Patient Education Patient education typically occurred post-visit or reactively after a message, missing high-engagement moments. Missed timing Overload + Regret Contextual patient education cards triggered during encounter
8 Emotional Fatigue Many physicians described emotional fatigue — feeling like they’re managing tasks, not patients. Disengaged caregiving Disillusionment + Burnout Automate repetitive admin; restore focus on patient interactions

Making blindspots accessible

Doctors aren’t asking for another dashboard — they want quiet clarity inside the one they already live in. They needed access to following touch points.

1

Social determinants

Financial, mobility, transportation issues etc
2

Behavioural determinants

Poor sleep quality due to a new born at home
3

Care gaps

Missed / due screenings, assessments & other H&M
4

Potential coding gaps

Missing ICD codes to apply
5

Recommend cost effective alternatives

Effective alternates covered in health plan
6

Recommend education opportunities

Diagnosis specific education

Intial sidekick explorations

From descriptives to actionable experiences, we explored 5-6 variations of AI sidekick.

Linear actionable view vs Insightful analytics view

Actionable view with quicker access to care sections

At a glance visit summarisation

Alternate exploration for quick navigation

Arriving at shippable version

We scored a balance between tech complexity & care essentials. While this did had a trade-off, it helped us find a balance between shipping the first version & collecting real usage feedback.

Access to quick assignment of care journey

Introducing practice specific widget customisation which makes this experience scalable. For instance, we can enable Prognosis for Nephrologist.

The final outcomes

AI sidekick reduced the need for context switching, improved documentation accuracy, reduced claims denial and improved overall CSAT  of physicians as well as patients got increased face time during encounters.

Reduced Context switching

60

%↓
Less cognitive fatigue

Reduced Claims denial

30

%↓
Better documentation accuracy