Cole Downing

The Problem

Seamless had a clear wedge — it was one of the best tools in the market for finding leads. But that wedge had a ceiling. Customers would pull their data, export it, and immediately leave to run their outreach in a separate system. The workflow was broken at the seam. Every sale ended with a handoff to a competitor.

What I Did

I joined as the founding PM with a mandate to expand the platform into sales engagement — email, calling, social, activity tracking, and automated campaigns. With a six-month runway to MVP, I made the call to launch the basics first and layer on AI and automation in phase two. No over-engineering the first version.

We launched an early beta with roughly equal investment in email and calling. Usage data immediately told a different story: roughly 80% of users were using email exclusively. Rather than preserve roadmap symmetry, I pushed to fully build out the email runway first and defer calling. We went deep on one thing instead of thin on two.

The Chrome extension was also a key unlock — we moved from the old extension format to the Google sidebar model, giving users a more seamless in-context experience that let them engage from anywhere on the web without breaking their workflow.

The Outcome

Connect activated roughly a third of Seamless’s monthly active users and is projected to become a significant ARR driver for the business. More importantly, it reframed what Seamless is: a platform that takes a customer from lead discovery through execution, rather than a data tool that hands off to competitors at the critical moment.

The Problem

Connect proved that users wanted an integrated sales workflow. The next question was harder: could the platform execute that workflow autonomously? Sales teams were still making dozens of manual decisions per day — building lists, selecting contacts, sequencing outreach, categorizing leads. The work was repetitive, high-volume, and exactly the kind of task an agent should own.

What I Did

I extended my mandate from Connect into the Agentic AI platform — a layer spanning the full stack from lead generation through engagement through analytics. The organizational challenge was as hard as the product challenge: we had essentially no native AI in the platform, and I was driving a roadmap that would make AI the primary surface for most user workflows.

That meant building internal conviction before building product. I led competitive research, defined the agent architecture with engineering, and structured a phased rollout that showed leadership what customers could do at each stage — not just where we’d end up. The goal: a system where an agent could find leads, build lists, create campaigns, and fire personalized outreach with minimal human intervention.

The Outcome

The company-wide AI roadmap is active and in execution. The strategic bet — that AI-native automation is the core differentiator for the next generation of sales platforms — is now Seamless’s primary product thesis.

The Problem

Large enterprises were managing territory assignment, quota planning, and budget allocation in a patchwork of specialized tools and spreadsheets. The category wasn’t new — but the incumbent solutions were either too complex, too siloed, or both. Salesforce’s bet was that an all-in-one, less intimidating platform built natively on CRM data could own this space.

What I Did

I was one of three founding PMs on the product, involved from the first whiteboard session through launch — hypothesis validation, requirements, design collaboration, and execution across engineering teams. The core MVP bet was territory planning and quota allocation, the highest-pain workflows in the category.

Two things made this hard. We hit a leadership transition at roughly the midpoint of the roadmap, which required rebuilding alignment without losing execution momentum. I also conducted R&D on generative and predictive AI features at a time when enterprise SaaS was only beginning to take AI seriously — knowing when to stop researching and ship was itself a judgment call.

The Outcome

SPM launched as a new Salesforce product line addressing a multi-billion dollar market opportunity. The territory planning core became the foundation the team continued building on after my tenure.

The Problem

Territory Planning was built on raster-based geographic data — an older format that lacked the granularity for complex territory modeling and only supported a small subset of countries. Enterprise customers in international markets couldn’t use the product. The technology constraint was also a revenue constraint.

What I Did

I ran the data team responsible for both Salesforce Maps and Territory Planning. The project required more than sourcing new geographic data — it required rearchitecting the data layer. We migrated from raster to vector format, built internal ETL pipelines to process and validate new data sources, and designed a data-agnostic ID system so the infrastructure could serve both products without tight coupling to either.

The hypothesis was straightforward: there was demand we couldn’t capture because we didn’t exist in those markets. Getting the data correctly and quickly was the only unlock.

The Outcome

Coverage expanded to 255 countries, opening international markets that had been structurally inaccessible. The TAM expansion was measurable within two months of launch.

The Problem

Salesforce Maps relied on a tightly coupled, source-dependent backend — a single database without proper environment separation, automated pipelines, or the flexibility to onboard new data providers. Every new data source required custom integration work, and the architecture couldn’t scale to meet enterprise demands for coverage and reliability.

What I Did

I took what was essentially one database and built it into a full-blown enterprise-level backend data architecture. That meant standing up three environments, building automated ETL processes and scripts to fetch, download, transform, and load data, creating deployment processes to push to production, and adding unit testing on the database layer.

The biggest accomplishment was moving from a source-dependent backend to a data-agnostic one. I redesigned the data architecture and schemas to support our own custom unique identifiers, which meant we could ingest data from any provider without rewriting integration logic. That unlock let us add over 20 new data sources for the mapping tool.

The Outcome

The data infrastructure went from fragile and source-locked to a scalable, data-agnostic platform supporting tens of thousands of enterprise customers. The architecture served both Salesforce Maps and Territory Planning with combined ACV exceeding $50M.

The Problem

When epidemiologists needed to update disease reporting guidance — the standardized value sets that define how conditions are classified and reported — those changes had to reach all 50 state health departments quickly and correctly. During Ebola and Zika response, this wasn’t an academic concern. If a state was reporting against outdated value sets, the data flowing back to the CDC was incomplete or miscategorized. The pipeline that distributed these updates was slow by design. A business analyst would manually update value sets through a GUI, then wait for scheduled overnight jobs to promote changes through development, staging, and production environments. At best, the process took three days. In an active outbreak, three days is a long time to be reporting incorrectly.

What I Did

I learned the system end-to-end — both the product surface and the backend architecture — and identified that the bottleneck wasn’t just slow process, it was structural dependency on legacy scheduled jobs and a fragile GUI layer. Working with the dev team, I designed an alternative path: a workflow that took updates from a spreadsheet directly to a database, validated through automated unit testing, and deployed to staging and production without touching the old GUI or waiting for overnight jobs.

The government contractor environment made this harder than it sounds. Zero margin for error, strict compliance requirements, and a culture that moved carefully by design. Building the case for a new deployment path required both technical credibility and stakeholder trust across CDC program managers and the dev team.

The Outcome

Value set publishing time dropped from three days to one hour, distributed across all 50 state departments of health. During active Ebola and Zika response, that meant updated surveillance guidance could be in the hands of state reporters within the same workday — and the data flowing back to the CDC reflected it.