B2B · Lead Principal Content Designer · Current
Ingram Micro Xvantage
Xvantage is a B2B ecommerce platform that gives resellers and IT partners a single place to browse, quote, and purchase technology products and services from Ingram Micro's global catalog. As the founding content designer, I define and drive the vision, strategy, systems, and tooling to evangelize content design across the enterprise.
Xvantage voice
Problem
Ingram Micro had three established brand voice characteristics: always honest, somewhat geeky, and extra empathetic. The only guidelines for writing in this voice were built for marketing, which serves a different purpose than UX content. That left a gap in how these characteristics should flex for in-app experiences.
- Starting with research: I led and worked with our UXR team to design a content design study across the different user personas. Each session showed participants different versions of the same content across real platform scenarios. From there, we moved into interviews to dig into how each version made them think and feel. I only kept themes that repeated across participants, so every guideline traces back to what users told us. My goal was for every voice decision to be defendable with data.
- Bringing the voice to life: I turned those findings into guidelines for writing our voice in the platform. For each characteristic, I defined how it shows up overall and how it flexes for associates and customers. Associates pass what they read on to partners, so every X4A message needs to be complete enough to repeat, simple enough to read mid-task, and specific enough to act on. Customers care more about outcomes, so their copy stays simpler with the same level of transparency. I paired each characteristic with Do this / Not that examples, so designers can apply the voice without having to interpret it.
- Building the style guide: This voice became the foundation of our UX content style guide. It covers design system guidelines for content components with formulas and reusable patterns, how the voice shows up for AI agents, editorial guidelines, accessibility and inclusivity (including writing alt text), and writing English copy that translates well for our global audience.
- Scaling with tooling: I'm currently introducing Figma skills and agents trained on all our guidelines to work across Figma Make and Figma Design to check, suggest, and directly replace content in mocks.
Impact
This is how we solve for quality and consistency at scale across 35 designers and over a dozen domains. To measure it, I'm tracking content debt through bi-annual audits, using my initial CX audit as the baseline, and tracking adoption of our reusable patterns to see how consistently teams are putting the guidelines into practice.
AI
Problem
AI agents in Xvantage were being built in silos across more than a dozen domains, so no two agents spoke or behaved consistently. It's no wonder trust was the biggest barrier to adoption, both internally and with customers.
Solution
My first big bet in this role was standardizing AI agent personas across Xvantage, defining one consistent personality and voice for all agents to use.
- First, I defined how Ingram Micro's existing voice and tone tenets should flex specifically for AI and conversational experiences.
- From there, I created two layers of guidelines: one in the guardrails and one at the evaluation level. I built a system prompt specific enough for AI to understand and follow, yet generic enough to work across every agent and use case, and lightweight enough not to drive up latency. This was the first system prompt the UX team had ever created, and much of the work was educating on why this was the right approach, as it lives outside the normal release cycle, so content standards can be updated instantly across every agent without waiting on a code release or re-deploy.
- I pitched the approach to the AI Council of Excellence, who signed off and gave Content Design access to sync the prompt as a constitutional layer across every agent, and to formally add it to the release gates in the Agent Development Lifecycle (ADLC).
- As part of the release gate, I built five content-based evaluation items, giving teams concrete test prompts and ideal outputs to measure agents against.
Impact
This work made content design a required, non-optional stage in AI agent development at Ingram Micro, and established the single voice and tone standard now governing all 15 agents in production. It also gave us the credibility to start getting UX more involved in the ADLC overall.
Content Design playbook
Problem
As one content designer supporting a UX org of 35 designers and ten UX researchers, I couldn't be in every room or review every piece of content myself. Designers were eager for support, so I quickly needed to document the long-term strategy for building the program and what it would actually look like day to day for them.
Solution
My first job was figuring out what it would actually take to stand up the discipline, then turning that into a year-one roadmap. I built the strategy for completing that roadmap in three stages:
- First, I defined the mission, vision, and eight guiding principles behind every content decision.
- I then mapped the full scope of what Content Design owns at Ingram Micro, from in-app copy and content systems to AI/LLM content, accessibility and localization, system communications, help and support, and product storytelling, giving stakeholders a clear reference for what to bring to Content Design and what not to.
- I documented the operational details that removed guesswork, including prioritization, intake and triage, and expectations for designers, all aligned with leadership so the practice could run consistently from day one.
Impact
The playbook became the vehicle for evangelizing content design as a strategic function across the enterprise, giving leadership and stakeholders a clear line of sight into the practice's scope, standards, and operating model. It remains the source of operational truth for how content design runs at Ingram Micro.
Note: All internal links are not usable.
