Research to Insight

Input

• Company: Narwhal Health — a B2B digital health startup offering chronic condition management tools for employer health plans
• Research questions on the table: Why are care navigators abandoning the task assignment workflow mid-session? Is the onboarding flow the reason new employer accounts take 6+ weeks to activate? Do care navigators trust the AI-generated care recommendations? Should we add a mobile app for navigators working remotely?
• Constraints: 3-week sprint, 1 researcher, 6 care navigator participants already recruited, 2 employer admin contacts available
• Business context: Q3 OKR is to reduce time-to-value for new employer accounts; PM suspects onboarding is the culprit but leadership wants abandonment data first

---

Output (abbreviated)

Research-to-Insight: Narwhal Health Care Navigator Study

---

Step 1 — Research Prioritize

Ranked Research Questions

Priority	Question	Rationale	Recommended Method
1	Why are care navigators abandoning task assignment mid-session?	Direct signal loss; tied to retention and navigator satisfaction	Interviews + session log review
2	Is onboarding the reason new employer accounts take 6+ weeks to activate?	Directly supports Q3 OKR; employer admin contacts already available	Interviews (employer admins)
3	Do navigators trust AI-generated care recommendations?	Influences adoption of core differentiator; can fold into Q1 interviews	Interviews (piggyback on Q1)
4	Should we build a mobile app for remote navigators?	Generative, low urgency; no evidence base yet — defer	Future round

Researcher decision point: Pursue Q1 + Q3 via care navigator interviews (n=6); Q2 via employer admin interviews (n=2). Q4 deferred. Proceed with /interview-plan for both cohorts.

---

Step 2 — Interview Plan

Care Navigator Sessions (n=6)

Objectives

• Identify the specific moments and triggers that cause navigators to abandon task assignment
• Understand how navigators evaluate and act on AI care recommendations

Hypotheses

• H1: Abandonment occurs when navigators can't find a matching care resource within 2–3 search attempts
• H2: Navigators distrust AI recommendations when no rationale or source is shown

Discussion Guide (abbreviated)

1. Walk me through the last time you assigned tasks for a member — what did that look like start to finish?
2. Tell me about a time you stopped partway through. What was happening?
3. When you see an AI-suggested care plan, what goes through your mind before you act on it?
4. What would need to be true for you to trust that suggestion without checking it yourself?

Employer Admin Sessions (n=2)

• Focus: Step-by-step account activation experience, points of waiting, who owns each step internally

---

Step 3 — Interview Synthesis (per-session nuggets, 3 of 6 shown)

Session 3 — "Maya," Senior Care Navigator, 4 years tenure

Nugget	Type	Quote
Abandons when care directory returns >2 screens of unfiltered results	Friction point	"I just give up and schedule a manual callback. It's faster."
Uses AI recommendation as a starting point, then calls the member's PCP to verify	Trust workaround	"I don't not trust it — I just can't explain it to the member if they ask why."
Wants to see the clinical source or guideline behind each recommendation	Unmet need	"If it said 'based on ADA 2024 guidelines,' I'd feel a lot better."

Session 5 — "Derek," Care Navigator, 18 months tenure

Nugget	Type	Quote
Task assignment abandonment tied to session timeout — loses work	System failure	"It logs me out and I have to start over. That's when I just quit."
Not aware AI recommendations exist — thought they were auto-generated alerts	Discovery gap	"Wait, that's AI? I thought that was just the system flagging things."

Employer Admin Session 1 — "Sandra," Benefits Manager, Kelsey Foods

Nugget	Type	Quote
Activation stalled 3 weeks waiting for SSO configuration from internal IT	External dependency	"We didn't know we needed IT involved until week 4. Nobody told us."
Onboarding checklist PDF was outdated — referenced a legacy portal	Documentation gap	"We were following the wrong instructions for half the setup."

---

Step 4 — Research Synthesize

Cross-Session Themes

Theme 1: Task Assignment Abandonment Has Two Distinct Causes

Confidence: High (5/6 navigators)

Two failure modes emerged — not one. Directory search overwhelm (too many unfiltered results) and session timeout data loss are independent triggers that both end in abandonment. They require separate fixes.

Supporting pattern: 4/6 navigators cited search quality; 3/6 cited timeout. 2/6 cited both in the same session.

Theme 2: AI Recommendation Distrust Is a Transparency Problem, Not a Confidence Problem

Confidence: High (6/6 navigators)

Navigators are not skeptical of the AI's accuracy — they're unable to explain recommendations to members or supervisors without a cited source. Trust is a communication need, not a model quality need.

Implication: Adding a "based on [guideline]" citation may resolve adoption friction without model changes.

Theme 3: Employer Onboarding Delays Are Caused by Undisclosed IT Dependencies

Confidence: Medium (2/2 admins — small n)

Both employer admins hit the same wall: SSO configuration requires internal IT involvement, but nothing in the onboarding flow signals this until week 3–4. The delay is structural, not behavioral.

Risk: n=2 is low; recommend 2 additional admin interviews before building a solution.

Theme 4: AI Feature Awareness Is Near Zero Among Newer Navigators

Confidence: Medium (3/6 navigators, tenure <2 years)

Navigators with under 2 years of tenure did not know AI recommendations were AI-generated. This is a discoverability and onboarding gap, not a trust problem for this cohort.

---

Step 5 — Research Readout

Executive Summary

Narwhal's care navigators are abandoning task assignment for two fixable, distinct reasons — and they're not using AI recommendations at their full potential because they can't explain them. New employer accounts are stalling in onboarding due to an undisclosed IT dependency. All three findings are actionable in Q3.

---

Finding 1: Two Root Causes Behind Task Abandonment

What we learned: Abandonment splits into (a) search overwhelm when directory results exceed ~10 unfiltered items and (b) session timeout causing data loss mid-task.

Implication: A single "fix abandonment" initiative will underdeliver. These require separate product responses — search filtering/ranking improvements and autosave/timeout extension.

Recommendation: Address timeout/autosave first (lower effort, higher frustration weight). Roadmap search quality improvement for next sprint.

---

Finding 2: AI Recommendation Adoption Blocked by Explainability Gap

What we learned: Navigators trust AI recommendations privately but won't act on them without a citable source. They are managing liability and member communication, not expressing model distrust.

Implication: Adding clinical guideline citations to recommendation cards would unblock adoption for experienced navigators. A separate awareness + onboarding fix is needed for newer navigators who don't know the feature exists.

Recommendation: Add source citation to recommendation UI (design spike in current sprint). Update new navigator onboarding to include explicit AI feature introduction.

---

Finding 3: Employer Onboarding Delay Is One Missing Sentence

What we learned: Both employer admins lost 3+ weeks because SSO setup requires internal IT, and nothing in the onboarding flow mentions this until it becomes a blocker.

Implication: Time-to-activation can likely be cut by 2–3 weeks with a single checklist update — low-cost, high-OKR-impact.

Recommendation: Add IT dependency callout to onboarding step 1. Validate with 2 additional employer admins before building