How a scalable content system replaced 40+ individualized workflows,
reduced monthly support tickets by 63%, and drove $10M in attributable
incremental spend — and is still running today.
The root issue: content was treated as a service request, not a system. Every account got a handmade answer to a question that had a scalable solution.
Every guide a CSM built became a liability the moment Amazon Business updated its UX or added functionality. Customers were following instructions for interfaces that no longer existed.
CSMs spent significant time recreating personalized guides for each account — pulling them away from high-value advisory work. This was not scalable.
~17,000 monthly CS tickets. Customers confused, frustrated, reaching out for help that should have been preemptively addressed. Cost center, not growth engine.
I wasn't handed a brief. I was the one experiencing the problem firsthand as a Senior Customer Success Manager — watching the same broken loop play out across my accounts, and then across my team's accounts too.
I wrote the PR/FAQ — Amazon's internal mechanism for proposing new programs — and sold this to my manager and sales partner first, then to CSM leadership, then to sales and marketing, then to upper leadership for production investment. Every stage required new data, new framing, new stakeholders.
That progression mattered. It meant I understood the business case at every altitude — from the ground-level CSM capacity problem to the executive-level revenue opportunity.
Most organizations treat customer education as a cost center — a reactive function that absorbs support tickets. The Atlas thesis was different: if we could build a content marketing system that was evergreen, centralized, and findable, it would function as a scalable growth mechanism. Every new account that self-served through Atlas was an account that didn't need a CSM hour to onboard.
Materials were organized around what the product does, not who bought it. A healthcare customer could find a government compliance video without it being "for them." Discoverability by need, not persona bucket.
The core technical insight: if the URL never changes, the content behind it can. Sales reps shared links that remained valid forever — we updated the video, not the link. This eliminated the guide-expiry problem entirely.
One content library, three audience paths. End users, administrators, and finance users received different newsletter messaging and Salesforce-curated portfolios — but drew from the same underlying asset library.
Atlas specifically unlocked small and medium business accounts that couldn't qualify for dedicated customer advisors. Self-service embedded in help pages and the Amazon Business homepage — discoverable at the moment of need.
Atlas required alignment across every team that touched the customer — from product to RevOps. Each function brought something the others couldn't.
Source of truth for workflow and use case accuracy.
Voice of customer + sales need to close deals.
Use case specificity + what customers resonated with.
Region-specific insights for UK + Italy localization.
Surfaced gaps in training across the account portfolio.
Upleveled creative quality and scaled video throughput.
Kept us honest — metrics, cohort analysis, attribution.
Embedded video links + provided ticket pattern feedback.
Branding, messaging, segmentation, and content tracking.
Atlas wasn't a content library. It was a content marketing infrastructure — with customer-facing channels and internal tooling built to reinforce each other simultaneously.
Central training library — searchable, evergreen, always current.
Segmented by role: end user, admin, finance. One send, three paths.
Self-serve at moment of need — intercepts confusion before it becomes a ticket.
Standardized assets for reps to deliver Atlas content in live customer meetings.
Portfolio links curated per account. Reps deliver without creating. Evergreen URLs.
Contextual help links inside the AB platform, surfaced at the relevant feature.
Built and owned centrally — every CSM sending the same Atlas-linked message to accounts.
Atlas content indexed for internal reference — CSMs and AMs could find and share quickly.
Live view for sales reps showing which videos were driving activation and results per account.
Social layer for internal updates, new content announcements, and cross-functional coordination.
Results in weeklies, MBRs, and QBRs — keeping leadership continuously informed.
UK + Italy via in-market CSM partners — existing content adapted, not rebuilt.
No CSM, no Account Manager — served entirely through marketing channels and CS. Atlas was the full customer relationship for this segment, driving faster activation, higher repeat spend, and greater product expansion than non-Atlas peers in the same tier.
Watched vs. not — logged-in tracking
Account creation → first use
Speed of usage spread within the account
Repeat spend + additional product activation
CS contacts per account pre/post Atlas
The program worked. But there's one thing I'd change — and it shaped how I approach every content program I've run since.
I built the system before I had the stakeholders. The pilot data was undeniable — but I had to compile it manually, wrangle multiple data sources, and waited for RevOps bandwidth to run comparative analysis just to make a case that should have been easier to make. If I'd brought revenue ops and senior leadership into the framing earlier, I would have had faster budget unlock, better cross-functional infrastructure from the start, and probably a full-time content marketing manager a year sooner.
Know who needs to see what, at what stage, to move the program forward. The internal sell matters as much as the external product.
The measurement system should be designed before the first piece of content ships. Attribution you can't prove is impact you can't build on.
Internal visibility — in weeklies, MBRs, QBRs — should be built into the operating model, not bolted on later. Leadership needs a continuous signal, not a quarterly surprise.
Atlas wasn't a campaign. It was content marketing infrastructure. It scaled a business function that couldn't otherwise scale. It eliminated cost. It created growth. And it kept working after I left.
That's what I build.
Entrepreneurs, educators, & advocates driving social or cultural impact.
Change Management
Emotional Intelligence
Change-makers are the ones redefining what’s possible, redirecting the master narrative, and protecting communities.
I’m here to help you transform bold ideas into sustainable impact.
I thrive at the intersection of emotional intelligence and empathy-driven change management, helping you navigate the complexities of transformation with care and clarity.
Whatever you’re working toward, I would be honored to amplify the movement.
Artists, filmmakers, writers, & designers looking to scale their vision.
Strategy
Analytics
Marketing Yourself
…these things tend to give even the most successful creators a case of the spookies.
Let us translate your genius and back it up with data and narrative.
We’re the team you call when the creative sparks fly but the details start to weigh you down.
From capturing those behind-the-scenes moments to brainstorming bold ideas or locking in the opportunities that take you to the next level, we’ve got you covered.
Nonprofits, start-ups, mission-focused brands, & socially responsible companies.
Whether you need:
a creative partner
a strategist
an executor
all of the above
—I’m here for it.
I’ve worked with professionals at every level, from C-suites and executive directors to board members, strategists, managers, and individual contributors, making sure everyone understands their role and feels empowered to contribute to the project’s greater purpose.
Thank you for taking the time to explore my portfolio.
Every project here represents a relationship built on trust—trust from mission-driven organizations and individuals with big ideas and the courage to pursue them.
That trust is something I deeply respect.
Each campaign I’ve created, every strategy I’ve designed, and all the content I’ve crafted have been opportunities to amplify voices that matter, highlight meaningful work, and connect people through shared purpose.
For me, marketing isn’t about selling—it’s about serving. It’s about showing up with creativity, clarity, and commitment to help others bring their vision to life.
Thank you for being here. I hope my work inspires you to imagine what’s possible for your own purpose, and I look forward to the possibility of supporting your journey.
With gratitude,
Tess
What is Federated Learning?
Federated learning is an AI training approach that enhances privacy and security by keeping data localized on users’ devices instead of centralizing it in one location. This decentralized method allows models to learn from data across multiple devices or servers while only sharing insights—rather than raw data—back to a central system.
In the context of AI responsibility, federated learning minimizes data exposure, reduces the risk of breaches, and supports compliance with data protection regulations like GDPR and CCPA. It also promotes ethical AI development by preserving user control over personal information and enabling more inclusive and privacy-focused AI systems.
What is Dynamic Workload Scheduling?
Dynamic workload scheduling is an energy-efficient computing strategy that adjusts when and where AI workloads are processed based on real-time conditions, such as renewable energy availability, electricity prices, and server capacity.
In the context of sustainable AI computing, it means shifting AI training or inference tasks to times and locations where renewable energy sources (like solar and wind) are abundant to reduce carbon emissions and energy costs.
How Does Dynamic Workload Scheduling Work?
Aligning AI Training with Renewable Energy Peaks
Load Balancing Across Data Centers
Taking Advantage of Variable Electricity Pricing
AI-Optimized Scheduling Systems
Google’s Carbon-Aware Computing:
Microsoft’s Project Forge Global Scheduler:
What Is an AI Accelerator?
An AI accelerator is a specialized hardware component designed to speed up artificial intelligence (AI) and machine learning (ML) workloads more efficiently than traditional processors like CPUs (Central Processing Units) or even GPUs (Graphics Processing Units). These accelerators are optimized for parallel processing, lower energy consumption, and high-performance AI computations.
How Do AI Accelerators Work?
Unlike general-purpose CPUs, which handle a wide variety of computing tasks, AI accelerators are custom-built for specific AI operations such as:
Matrix multiplications & tensor processing (core operations in deep learning).
Neural network training & inference (faster model execution).
Optimized data flow (reducing memory bottlenecks).
These accelerators reduce the energy and time required to train AI models and process real-time AI applications, making them crucial for sustainable computing strategies.
Examples of AI Accelerators
1. Google Tensor Processing Units (TPUs)
What it is: Custom-built by Google for deep learning workloads.
Why it matters: Uses less power than GPUs while accelerating AI model training.
Example: Google’s TPUs power Google Search, Google Photos, and AI-driven healthcare research.
2. AWS Inferentia (Amazon Web Services)
What it is: A custom AI chip designed for machine learning inference (running trained AI models efficiently).
Why it matters: Uses lower power and costs less than GPUs for AI-powered applications.
Example: Powers Alexa, AWS AI services, and real-time recommendations for e-commerce.
3. NVIDIA Grace Hopper Superchip
What it is: A hybrid CPU-GPU superchip designed for high-performance AI applications.
Why it matters: Reduces energy consumption while handling massive AI models like large language models (LLMs).
Example: Used in supercomputers, autonomous vehicles, and generative AI models.
A specialized processor designed for parallel processing, originally developed for rendering graphics. GPUs have thousands of smaller cores that can process multiple tasks simultaneously, making them ideal for AI, machine learning, gaming, and high-performance computing. Unlike CPUs, GPUs are optimized for large-scale data computations, enabling faster processing of complex mathematical operations.
