How direct agent research exposed a flaw in the platform strategy and led to a household-centered information model.
Read Time: Summary 3-7 Min. / Deep Dive 8-12 Min
The business needed agents to move to a new CRM, but the planned experience did not reflect how they ran their offices or served customers. I introduced direct agent evidence into the migration strategy and translated it into a household-centered information model that stakeholders and users could validate.
Reframing a single-workflow CRM into two platforms that agents actually needed.
Problem
A large insurer was modernizing how agents managed customer relationships, but the initial platform direction treated their work as a single workflow. Research revealed two distinct needs: serving customers during conversations and managing day-to-day agency operations. Without a clearer structure, the new system risked slowing agents down and limiting adoption.
Action
Led discovery and experience strategy for the enterprise CRM modernization. Secured and helped facilitate multi-day workshops with agents, synthesized the findings, and translated them into a customer-data hierarchy, prioritized workflows, and platform concepts showing what information agents needed at each point in a customer conversation.
Result
Established an evidence-based direction for a platform supporting both customer service and agency operations. The work gave product and business teams a shared understanding of agent needs, informed the system's information architecture and experience priorities, and created a clearer foundation for service, engagement, and adoption.
Reframed a single workflow into two
distinct needsÂ
Customer conversations and agency operations
sharpening product and design decisions.
Gave product and business teams a common, evidence-based view of agent needs across the platform.
Informed the system's information architecture and experience priorities for faster service, stronger agent engagement, and adoption.
STRATEGIC REFRAME
The initial direction treated agent work as a single workflow. Research reframed it into two distinct needs — customer conversations and agency operations — giving product teams a clearer path to prioritization and design.
ROLE
Design strategist and information architect leading the agent-facing experience within a multi-team enterprise CRM modernization.
CHALLENGE
The business needed independent agents and their office teams to move from a legacy CRM to a new enterprise platform. But the proposed experience was being shaped by internal feature owners and product requirements that had not been tested against the workflows of the people expected to use it.
STRATEGIC INSIGHT
Agent resistance was not simply resistance to change. The planned platform organized information around internal products and individual policies, while agents managed relationships at the household level. The migration would remain difficult to justify until the product reflected that difference.
STRATEGIC RESPONSE
I introduced direct agent research into the product-definition process and translated the findings into a Hierarchy of Data framework that reorganized customer information around households, service needs, and life events.
VALIDATED OUTCOME
The household-centered direction was validated by agents and stakeholders and gained support over a competing lift-and-shift approach. I transitioned off the project before implementation, so post-launch adoption and performance were not outcomes I directly observed.
Deep Dive
To go deep and learn more about my thinking, judgment, and reasoning expressed on this engagement, continue reading and dive deep.
Est. Reading: 8 -12 Minutes
My Scope Within a Multi-Team Modernization
The engagement ran roughly two years, ending when I transitioned to a different role, not when the work concluded. Multiple teams were involved: the author's own design track, at least one peer design team pursuing a different technical approach, product management, and the feature product owners feeding them specs. The author's scope was the customer-facing side of the modernization specifically, the household data and service experience agents used day to day, not the platform's full technical build, and not a separate, related back-office tool the business also had in flight for agents' own business management (a distinct effort, outside this case study's scope).
Worth stating plainly: this account covers what happened up to that departure. It is not a launched-and-measured story, and it doesn't pretend to be one.
Independent agency office workers aren't a captive audience. They run their own businesses, and the CRM is the tool they use to serve their own customers. Force a migration onto a system that doesn't reflect how they actually work, and there are exactly two outcomes: the rollout stalls, or agents route around it in ways that eventually reach the customers those agents serve. Neither is acceptable when the thing being modernized is the tool people use to manage another person's insurance, their bills, and the life events attached to both.
That's the stake that made this more than an internal tooling decision. A failed migration here doesn't just cost the business money. It puts the actual service someone receives from their insurance agent at risk.
On paper, this looked like a documentation gap. During early demos, it became clear that the specs didn't have enough detail on how the work actually flowed; screens were being built to match a described process that didn't match the real one closely enough to hold up. The initial fix was more detail: refine the wireframes, tighten the prototypes, and go back to the same spec source and ask for more.
The real gap wasn't in the level of detail in the specs. It was in where the specs came from. Product managers had been collecting requirements from feature product owners inside the company, people responsible for a product line, not people who spent their day inside an agent's actual workflow. Nobody in that chain had asked the agents themselves how they worked, what they needed in front of them first, or what a household actually looked like from their side of the desk.
That assumption wasn't unique to this work stream. Around the same time, I was separately asked to map a customer-service scenario, because, again, no one else had documented how that particular interaction actually worked. Two different requests, the same missing piece: an organization that had built a spec-review process with no seat at the table for the people the specs were about.
Once that was visible, the fix wasn't better documentation of the old plan; it was a new source for the plan. The team ran direct end-user workshops, usability testing, and focus groups with independent agency office workers: not to validate a design that already existed, but to map how their business actually ran, what a household meant in their world, and which scenarios the platform needed to solve for that no spec had captured.
The reframe that came out of it: build the platform for how these agents do their job and run their business, not for how corporate assumed it should work, and not for how a product roadmap had already assumed it could be implemented.
The framework that came out of that research was Hierarchy of Data, a principle for presenting end-user information in stages that match real need, instead of surfacing everything a system holds about a customer at once. Applied to this work stream, it became the household view: instead of an agent working one policy at a time, they'd see a household's full product relationship together, with the details that mattered most- a life event, a lapsed payment, a coverage gap- surfaced first rather than buried under fields that didn't matter to that conversation.
This wasn't a framework built for the case study after the fact. It's the name I used for the framework at the time, and it's a principle the competing peer-team approach, preserving the old system's look and feel through a "lift and shift," didn't share.
The central tension wasn't technical. It was that another design team was already committed to lift-and-shift, keeping the existing interface's shape to minimize disruption, while this workstream's direct evidence pointed toward a structural rebuild. Both sides believed they were representing the end user. Both approaches were described as user-centered, but only one had been tested directly with the agents expected to use the platform.
Resolving that took more than a design decision. It took proving the recommendation against evidence the other approach didn't have: workshop findings, usability sessions, and a direct account of how agents' businesses actually ran. That evidence, not seniority or preference, is what eventually moved stakeholders toward the Hierarchy of Data direction.
One organizational constraint shaped what "the new platform" could even mean: the business didn't want to move off its core CRM platform outright, given the cost of a global platform change. Any recommendation had to work within that constraint, not against it. The household view had to be something the existing platform could actually support, not a clean-slate reimagining that ignored what the business had already committed to.
That's a real tradeoff, not a footnote: the strongest version of an idea is rarely the version that ignores what the organization has already paid for.
Customer Impact: The validated design would allow agents to see a household's full relationship with the company in one place, with life events surfaced rather than buried, meaning conversations with customers could be proactive instead of reactive. This reflects the validated design direction; I did not observe it in live use, since the transition off the project came before launch.
Business Impact: The business avoided the likely alternative outcome of a forced migration built on unvalidated specs, a path that risked either a stalled rollout or agents working around a system that didn't fit their business. Adoption outcome not measured; author transitioned off the project before implementation.
Organizational Impact: Stakeholders and end users validated a working example showing how direct end-user research produced requirements and design decisions that more closely reflected agent workflows. Whether that changed how the organization sourced specs going forward is not something the author was in a position to confirm.
The clearest lesson from this project wasn't about data hierarchies. It was about where trust actually gets built in a migration. Not in the interface, and not in the announcement. In whether the people being asked to move can see themselves in what they're being asked to move to.
What was genuinely surprising, in hindsight, wasn't that the specs were incomplete. Incomplete specs are common. What was surprising was how confidently a competing team believed their own assumption-based approach was already user-grounded, right up until it was tested against people who'd actually go ask. That's not a criticism of the other team's intent. It's a reminder that "we understand the user" and "we asked the user" are not the same claim, and only one of them holds up under a workshop.