Thomas Meerschwam | Product Marketing · Enterprise Software & AI

I help complex products find their market.

I've spent six-plus years working across product management and commercial operations at Braze, ASAPP, and Artsy. The work that has interested me most sits between the product and the market: understanding customer needs, identifying where a product can win, and explaining why a customer should care.

At Braze, I built the market-selection criteria and analysis behind an RCS launch across 11 US and European markets, then worked with product marketing and customer-facing teams on positioning and launch readiness.

I'm now moving into product marketing, bringing hands-on experience with enterprise launches and with building AI applications myself.

Bringing new channels to enterprise customers.

Both launches put a new messaging channel in front of large enterprises. In each, the product question came with a market question: where to launch first, why a customer should adopt it, and whether the economics worked.

Braze Product Manager · SMS & RCS

Launching RCS across 11 US and European markets

11 US and European markets at launch

40 enterprise customers began deployments within 90 days

The market question

RCS readiness isn't uniform from country to country. Which markets should go first, and what would each customer-facing team need to sell it?

What I did

Developed market-selection criteria and analysis to recommend launch sequencing. Contributed to go-to-market planning and cross-functional readiness with marketing, sales, customer success, and external partners.

Worked with product marketing on positioning, messaging, sales enablement, collateral, announcements, and conference materials.

Result

40 enterprise customers initiated RCS deployments within 90 days of launch.

The broader SMS/RCS portfolio grew ARR approximately 50% year over year. This was a portfolio-level result with multiple contributors, not a direct measure of the RCS launch.

ASAPP Product Manager · Enterprise AI

Adding WhatsApp as an enterprise customer-service channel

2M+ end users reached in the first 30 days

The market question

Why would an enterprise add WhatsApp to its customer-service mix, and did the channel economics justify building it?

What I did

Developed the customer case for WhatsApp as an enterprise service channel. Worked with pricing and finance to understand the channel economics and inform the build decision.

Navigated Meta platform requirements, integrations, and launch readiness to bring the new channel to enterprise customers.

Result

More than 2 million end users reached within the first 30 days.

Enterprise software has been the throughline.

I started in sales and revenue operations at Artsy, then moved into product at ASAPP and Braze.

At Artsy, I rebuilt revenue systems and worked closely with sales and other customer-facing teams. It taught me how a product gets sold and supported, not just how it gets built.

At ASAPP and Braze, the work I cared most about was the part that touched the market: which customers a product was for, why they would adopt it, and what the teams selling it needed to know.

Understanding AI products from the inside.

I build AI applications to understand what emerging technology can actually do, where it fails, and what a customer would need to trust it. These are engineering projects—not marketing campaigns—but they give me firsthand experience with product capabilities, limitations, and enterprise buyer questions.

Flagship Project

LLM application · controls · evals · business case

Synthetic Case Ecommerce Support

Support Agent | Delivered-Not-Received Workflow

I built this so the LLM has one job: turn a messy customer message into structured case data. The rest of the workflow is explicit.

When a customer says a package was marked delivered but never arrived, the system retrieves the evidence, applies policy, checks authority, and either issues an idempotent refund or routes the case to a human.

This is a synthetic case. The retailer, data, policies, integrations, and economics are simulated. The application is real and runnable. I used the case to build and test controls, failure handling, evals, and rollout logic.

Problem

A missing-package claim can end in a refund. The system needs to make a good decision without giving the model uncontrolled access to money or customer state.

What I Built

Customer message → structured case → evidence lookup → policy checks → authorization → refund or human review.

What I Tested

Workflow behavior, authorization, duplicate actions, failure recovery, model extraction, tracing, and a staged rollout plan.

Modeled Business Case

What would make this worth buying?

In the synthetic base case, the workflow covers 70% of eligible cases. It frees roughly 1,200 hours of support and review work each month and produces about $46K in monthly net value in the model. These are modeled results. With a real customer, I would validate the assumptions before making the case.

The model interprets the evidence. The system owns the process.

The LLM can classify the case. Policy and authority live elsewhere. Money moves only after the workflow passes those checks. That separation is what makes an AI system something a risk-averse enterprise can say yes to.

LLM extraction Policy rules Tool permissions Authorization Idempotency Human review Tracing Evals Rollout plan Economics

Try Demo Project README View Code

Public demo uses scripted extraction and makes no provider calls. Free hosting may take about 30–60 seconds to wake after inactivity.