Service Design · Enterprise Insurance

The service had been failing for years.
Nobody had ever drawn the whole thing.

1,500+ insurance professionals. A platform built for data entry, not service delivery. The interface was the smallest part of the problem.

Diagnosis, not brief: When I mapped the full service ecosystem, I found a spreadsheet operating as a shadow contact database, an email channel that had migrated entirely off-platform, a supervisor role with no designed touchpoint, and an IT ownership model that turned a 30-second task into a 3-day wait. None of these were interface problems. All of them were surfacing as interface friction.

Role
Design Strategist & Service Lead
Domain
P&C · Mgmt Liability · Workers' Comp
Actors
6 types · 4 user roles
Was 3 days
30s
Contact update — backstage ownership shift
Was 14 steps
8
Journey steps — 6 workarounds removed
Was 0% baseline
85%+
Task success across all actor roles
Lost daily
3–5h
Supervisor time recovered per day
Service Blueprint Ecosystem Mapping Customer Journey Mapping Physical Evidence Co-Design Frontstage / Backstage Analysis Contextual Research Service Prototyping

From UX problem
to service problem

The brief arrived as a UI redesign. The first thing I did was reframe it — not to be clever, but because the wrong diagnosis would have produced the wrong solution.

— Original framing (UX lens)
Problem Statement
"The interface is hard to use and there are too many steps"
Research Focus
Task completion rates, usability errors, navigation issues
3-day contact update
Treated as workflow inefficiency → add a faster form
Siloed communication
Treated as inbox UX problem → improve notifications
No role-based visibility
Treated as a display/filter UI problem
— Service design reframe
Problem Statement
"The service architecture doesn't reflect the operational reality of the people delivering it"
Research Focus
Who depends on whom, where handoffs break, what the backstage is doing to the frontstage
3-day contact update
IT owned a task that belonged to operations — ownership problem, not a speed problem → move the ownership
Siloed communication
Email had become the real coordination layer, invisible to the platform → bring the channel inside the service
No role-based visibility
Four actors depending on the same service, none seeing what they needed — actor design gap, not a filter problem

Mapping the actors, not just the users

Before designing anything, I mapped who was actually touching this service and how they depended on each other. The ecosystem exposed four undesigned relationships, one completely missing actor touchpoint, and two channels operating entirely outside the service boundary.

"There was no designed touchpoint between Claims Handlers and Insurers. Every piece of communication was happening through personal Outlook accounts — nothing in the system, no record kept, nothing that survived when someone left the team."

Key ecosystem finding
Claims Management System CLAIMS HANDLER Primary frontstage actor Manages claims end-to-end 4 ROLES TEAM SUPERVISOR Oversight + escalation NO DESIGNED TOUCHPOINT ROLE GAP OPS MANAGER Performance oversight NO SYSTEM VIEW Depended on verbal reports IT / ADMIN Held contact ownership 3-DAY BOTTLENECK WRONG OWNER INSURER External actor ENTIRE RELATIONSHIP UNDESIGNED CLIENT / CLAIMANT End beneficiary Completely invisible NOT DESIGNED FOR Designed relationship Broken / undesigned channel Wrong ownership

Exposing the backstage breakdowns

The blueprint was built in three passes: as-is, failure map, to-be. Physical Evidence, Frontstage, Backstage, and Support Processes as distinct swim lanes. Read the full as-is story first, then see what the to-be replaced.

As-is — what we found
To-be — what we designed
Stage 📍 Physical Evidence 👤 Frontstage — Actor actions ⚙️ Backstage — Staff processes 🔗 Support — Systems & tools ⚡ Service failure
— Claim intake
Intake Printed claim form. Personal spreadsheet. Sticky notes for contact names. Handler receives claim. Searches spreadsheet for insurer contact. 15–30 min to find right person. No triage logic. All claims enter one queue regardless of type or urgency. Shared spreadsheet (not linked to platform). No single source of truth. Contact not found → manual search. Claim stalls.
Shadow layer
— Contact management
Contact update IT ticket printout. Email thread with IT. 3-day wait confirmation. Handler needs to update insurer phone. Raises IT ticket. Waits 3 days. Claim stalls. IT receives ticket. Updates backend record. No SLA visibility for handler. IT ticketing system (separate tool). 3-day SLA. 3-day wait to fix a 30-second task.
Wrong owner
— Claim communication
Notify insurer Personal Outlook inbox. Email printouts. No in-system record. Handler emails insurer from personal Outlook. No record in platform. No traceability. No structured comms protocol. Each handler improvises. Supervisor has no visibility. Personal email (Outlook). Entirely off-platform. Knowledge lost when handler leaves. Communication invisible. Lost on handoff.
Off-platform channel
— Supervisory oversight
Status check Phone. Personal notes. Verbal conversations. No system artefact. Supervisor interrupts handler for update. 3–5 hours per day lost across team. No system visibility for supervisors. Oversight done verbally or by email. Phone, email. No aggregate view. Each check is manual. 3–5h lost daily. Supervisor blind.
Actor: role gap
— Claim handoff & task routing
Handoff Email. Verbal briefing. No structured artefact or record. Handler manually updates fields, then emails supervisor separately. No confirmation. No structured handoff protocol. Each handler decides individually. Personal email + verbal. No record of what was communicated. Handoff invisible or lost. No context for next handler.
No protocol
— Claim resolution
Resolution Personal Outlook sent items. No closure artefact in platform. Handler closes claim. Sends closure email from personal Outlook. No standard process. No defined resolution protocol. No institutional learning. Personal email. No closure record in platform. No pattern data captured. Inconsistent quality. No learning loop.
No learning infrastructure

Blueprint insight: Of the 14 as-is steps, 6 existed only as workarounds for backstage failures — not genuine process complexity. The Physical Evidence row exposed the organisation's artefacts (sticky notes, printed emails, spreadsheets) as a visible map of every undesigned service layer.

Stage 📍 Physical Evidence 👤 Frontstage — Actor actions ⚙️ Backstage — Staff processes 🔗 Support — Systems & tools ✅ Service fix
— Claim intake
Intake Digital claim record. Role-filtered queue view. Inline contact card. Handler opens claim in role-filtered queue. Insurer contact surfaced inline. No search needed. Rule-based triage routes claim by type + urgency on arrival. Governed contact database integrated into platform. Self-service, audited. Contact resolved in seconds. Claim enters workflow immediately.
Fixed: ownership shifted
— Contact management
Contact update Inline edit field in contact card. Audit log entry. Instant confirmation. Authorised handler edits contact directly in platform. Live in 30 seconds. Audit trail auto-logged. Permissions model enforces who can edit. All changes versioned and attributable. Contact DB is part of CMS. IT dependency removed entirely. 30 seconds. No ticket.
Fixed: self-service
— Claim communication
Notify insurer Claim communication thread in platform UI. Message templates. Supervisor dashboard notification. Handler composes message inside claim record. Or sends from Outlook — auto-synced back. All claim comms attached to claim record. Searchable. Supervisor sees full thread. Outlook integration (two-way sync). Message templates for standard scenarios. Full audit trail. Knowledge stays after handoff.
Fixed: channel reintegrated
— Supervisory oversight
Status check Supervisor dashboard screen. At-risk claim flags. Portfolio view. Supervisor opens team dashboard. At-risk claims surfaced automatically. No interruptions. System applies risk indicators to surface priority claims in real time. Role-based supervisor view — first designed touchpoint for this actor. Self-serve oversight. Handlers uninterrupted.
Fixed: actor touchpoint added
— Claim handoff & task routing
Handoff Structured handoff form in platform. Auto-notification receipt. Claim timeline event. Handler completes structured handoff form. System auto-notifies next handler + supervisor. Handoff creates event in claim timeline. Next handler sees full context. Task queue with structured assignment. Automated notifications. Full audit trail. Handoff is a designed moment. Context preserved.
Fixed: structured protocol
— Claim resolution
Resolution Resolution summary screen. Feedback prompt. Closure confirmation email generated from platform. Handler completes templated resolution flow. Structured closure comms sent from platform. Resolution data feeds service review cadence. Pattern analysis identifies recurring failures. Resolution templates + structured feedback. First institutionalised learning loop. Consistent quality. Learning loop active.
Fixed: learning infrastructure

The human at the end of the service chain

Internal inefficiencies don't stay internal — they cascade to the claimant. Mapping the end-to-end claimant experience showed how every backstage failure became a frontstage moment of anxiety. The recovery arc is causal, not cosmetic.

Phase 01 — FNOL
Submission & Silence
Touchpoint: Web Portal
High anxiety
Claimant submits their initial claim. No meaningful confirmation of next steps. The system logs the entry — and then goes silent.
"I submitted it but I don't know if anyone saw it. What happens next?"
Phase 02 — Evidence
The Black Box
Touchpoint: Email / Post
Frustration peak
Claimant sends documents but hears nothing. Behind the scenes: adjusters manually downloading PDFs, re-uploading to EPIC — "swivel-chair integration" at its worst.
"I sent everything they asked for. Why is nobody responding?"
Phase 03 — Status
Conflicting Information
Touchpoint: Call Centre
Trust erosion
Claimant calls. The rep navigates three systems while on hold — reading from data 24 hours out of date. Rep and adjuster have different views of the same claim.
"I was told something different last time."
Phase 04 — Decision
Waiting for an Outcome
Touchpoint: Phone / Silence
Low trust
Claimant waits for a resolution decision. The adjuster handling the claim has changed — but because handoffs were invisible, the new handler has no context. Any follow-up call means re-explaining from scratch.
"I spoke to someone last week. Now there's a different person and they don't know anything about my case."
Phase 05 — After redesign
Informed & Resolved
Touchpoint: Portal + Email
Trust restored
Structured handoffs → faster updates. Single source of truth → call centre reps and adjusters see the same claim. Templated resolution → consistent closure.
"I got a clear summary. It felt like they actually knew my case."

The recovery arc is causal, not cosmetic: Phases 1–4 failures all had the same root — a fragmented internal service. Fixing the backstage directly improved every claimant touchpoint. The claimant experience was not redesigned separately — it improved because the service did.

What the research actually surfaced

I spent time with handlers, supervisors, ops managers, and IT — watching how they actually worked, not how the system was supposed to work. The methodology was deliberate: contextual observation first, interviews second, cross-role synthesis third.

01
Ecosystem · Ownership gap
The spreadsheet was a shadow service layer
Insurer contact data lived in a spreadsheet because the platform had never been designed to hold it. Not a data problem — an undesigned service layer the organisation had built around its own system's limitations over years.
02
Backstage · Ownership misalign
IT held ownership that belonged to operations
The 3-day contact update delay wasn't a technical constraint — it was a service ownership misalignment. Redesigning ownership was the solution, not redesigning the form.
03
Frontstage · Channel failure
Communication had migrated entirely off-platform
Handlers had stopped using the system for coordination. Personal Outlook had become the real service channel — invisible to supervisors, irretrievable when people left.
04
Actor · Role gap
Supervisors had been designed out of the service
The platform was built for handlers. Supervisors had no designed touchpoint — their oversight function happened entirely through verbal check-ins. 3–5 daily hours that a single visibility layer could replace.
05
Journey · Workaround inflation
14 steps were not all interface problems
Of the 14 steps, 6 existed only as workarounds for backstage failures. Reducing to 8 required removing systemic workarounds. The interface simplification was the consequence — not the cause.
06
Support · Learning gap
Nobody knew if the service was actually working
When a claim closed — nothing captured. Service quality depended entirely on who happened to be doing the work, not on any standard the organisation had designed.

Who had the power to block or enable change

Before any co-design or blueprint work began, I mapped all six stakeholders by power and interest. This determined sequencing — who I engaged first, how, and what I needed from each. The matrix made visible a crucial dynamic: the highest-power stakeholder (IT) was also the one most at risk of blocking change.

Power →
Keep Satisfied
Ops Leadership
High power, lower interest
Compliance
High power, low day-to-day
Manage Closely
IT / Admin ⚠
HIGHEST RISK — block or enable
Claims Supervisor
Medium power, high interest
Monitor
Insurer (ext.)
External, low direct power
Keep Informed
Claims Handler
Lower power, highest impact
Low InterestHigh Interest

Engagement sequence — how the matrix shaped the project

Week 1–2 · IT / Admin: Engaged first — highest risk. Co-built the as-is blueprint together to neutralise resistance before any solution was proposed. By helping surface the failure themselves, the solution felt like a shared discovery.
Week 2–3 · Ops Leadership: Presented service failure in business cost language (3–5h daily × 1,500 people). Secured approval for the IT ownership transfer before designing solutions.
Week 3–5 · Supervisors + Handlers: Role-play walkthroughs and contextual research. Surfaced specific friction points that no survey or interview would have caught.
Week 4 · Compliance: Consulted specifically on the permissions model. Reframed the risk — the danger wasn't in capturing emails on-platform, it was in the current state where emails were entirely untracked. They approved.

How four actors experienced the same service differently

The map only made sense when all four rows were visible at once. What looked like minor friction from the handler's chair was the thing that paralysed the supervisor.

Actor / Stage Claim Intake Contact & Notify Processing Handoff Resolution
Claims Handler Receives claim, searches spreadsheet for insurer Emails insurer from personal Outlook — no record in system Checks status by memory or asking colleagues Manually updates fields. Emails supervisor separately Closes claim, sends closure email from personal Outlook
Handler's thought "Is the spreadsheet up to date?" "I need to remember to log this email later" "I'm losing track of mine vs reassigned" "Did my supervisor see this?" "I'll just do what I did last time"
Team Supervisor No visibility into new claim volume Unaware of which insurers are being contacted Spends 3–5h/day on status calls Learns of handoffs only if handler remembers No view of resolution quality or patterns
Pain Peaks HIGH — 15–30 min contact search HIGH — No comms traceability MED — Cognitive overload HIGH — Invisible, error-prone MED — No institutional learning
SD Opportunity Integrated contact DB removes shadow layer Outlook integration brings comms on-platform Role-based dashboard eliminates status calls Structured handoff protocol + auto-notify Resolution templates + feedback loop

The principles that governed every decision

These four principles came directly from what the research had shown — not design philosophy, but specific lessons this service had already taught us. Each one was tested at least once when a stakeholder pushed back.

01 ——

Own the service layer, not just the interface

If a UX problem has a backstage root cause, fix the backstage first.

Tested when: Stakeholders proposed improving the contact update UI before addressing IT ownership. The principle held — redesigning the form without moving ownership would have produced a faster path to the same 3-day wait.
02 ——

Design for all actors, not the loudest one

Every actor in the ecosystem needs a designed touchpoint.

Tested when: Engineering pushed to descope the supervisor dashboard as "phase 2." Without a supervisory visibility layer, the 3–5h daily interruption overhead would have remained. The dashboard stayed in scope.
03 ——

Visibility is a service condition, not a feature

When actors can't see what they need, the service fails — regardless of what the interface looks like.

Tested when: A single unified "all claims" view was proposed as simpler to build. The principle forced the question: simpler for whom? It would have reproduced the exact visibility failure we were solving.
04 ——

Channels are part of the service, not outside it

Email doesn't stop being part of the service because it leaves the platform.

Tested when: Compliance raised concerns about email records on-platform. Reframed: the risk wasn't in capturing emails, it was in the current state where they were entirely untracked. Compliance approved.

What we redesigned at each layer

Six interventions — each targeting a specific service layer failure. The screen redesign was the last thing we built. The first was organisational: changing who owned what in the service.

Organisational Layer — ownership & accountability
🗄️
Self-Service Contact Database
Transferred insurer contact ownership from IT to operations. Eliminated the spreadsheet shadow layer and the 3-day IT dependency. 30-second self-service update replaces a 3-day ticket.
Backstage Layer — process & routing
Automated Task Routing
Codified the triage rules that already lived in people's heads and put them in the system. Consistent intake logic replaces supervisor discretion.
🔄
Structured Handoff Protocol
Handoffs were the most fragile moment and the least designed. A structured form replaced email + hope. Context preserved. Auto-notification on assignment.
📋
Resolution Templates + Feedback Loop
First institutionalised learning loop the service ever had. Resolution data feeds pattern analysis — so failures don't repeat silently.
Frontstage Layer — actor visibility & channels
📊
Role-Based Visibility Dashboard
Supervisors had never had a designed place in this platform. This was their first touchpoint — and replaced 3–5 hours of daily phone calls with a morning dashboard.
📧
Outlook Integration
The email channel wasn't going away — so we brought it inside. Claim-related emails attach to the claim record automatically. Visible to supervisors, traceable on handoff.

The interface redesign came last. The 14→8 step reduction didn't come from better IA. It came from removing 6 steps that only existed because something backstage was broken. Fix the root cause, and the complexity largely dissolves.

Designing with them, not for them

The most useful thing I did had nothing to do with design methods. It was getting IT, Compliance, handlers, and supervisors in the same room — in several cases genuinely for the first time. Not to reach agreement. To look at the same problem together.

Who was in the room
H
2 × Claims Handlers
Primary frontstage actors
S
1 × Team Supervisor
Oversight & escalation role
IT
1 × IT Lead ⚠
Highest-risk stakeholder
O
1 × Ops Manager
Resource & performance view
C
1 × Compliance Rep
Governance & audit lens
A
Facilitator (Ashish)
Observer + note-taker

Desktop Walkthrough · Litigation Claim Scenario · 8 participants · 90 minutes · 6 weeks into project

01
Briefing · 10 min
Each person receives a role card — their goal and the tools they normally use. No script.
02
Claim arrives · 10 min
Handler receives a complex litigation claim. Searches spreadsheet for insurer contact.
03
Live walkthrough · 35 min
Every actor moves through their touchpoints simultaneously in real time. Rule: do what you'd actually do — not what you should do.
⚡ Key moment — min 32: Supervisor asks for a status update. Handler navigates 3 systems. 20 minutes later gives the answer. IT watches silently.
04
Debrief · 35 min
Each actor names what surprised them. Three findings emerged that no individual interview had surfaced.
Finding 1: Supervisor had never watched a handler work. "I do this 5× a day, don't I."
Finding 2: Problem reframed — not "supervisors need more info" → "stop interrupting creators"
Finding 3: IT saw — firsthand — the cost of their ownership model. Resistance dropped immediately.

Impact at service level, not just screen level

The questions I was measuring against: does a supervisor's day actually look different? Does a handler have what they need when they open a claim? Does a handoff stay intact when someone's off sick?

Was: 0% baseline
85%+
Task success validated through end-to-end service scenarios, not isolated screen tasks — across all 4 actor roles
Was: 3 days per update
30s
Contact update time — a 3-day IT ticket replaced by 30-second self-service. Ownership shift, not UI fix.
Was: 14 steps (6 were workarounds)
8 steps
Journey steps reduced by fixing the service — not by polishing the UI. 6 workaround steps removed.
Was: lost to status calls daily
3–5h
Per supervisor, per day recovered. Status call overhead eliminated — supervisors shifted to system-mediated oversight.

Organisational capability, not just a better product

The deepest impact was structural: this was the first time the brokerage had a service blueprint for its claims operation. The design work created organisational infrastructure that outlasts the interface itself.

A service blueprint for the claims operation (first time ever)
Documented actor roles with designed touchpoints for all 6 actors
Clear backstage ownership model for all service tasks
Replicable handoff and resolution protocols
Integrated communication channel with full audit trail
First institutionalised feedback and learning loop
"Fix the backstage first. The frontstage will follow."
A principle that held at every pushback point in this project