Hims & Hers (NYSE: HIMS) is one of the largest telehealth brands in the world, generating $1.5B+ in revenue annually. As a telehealth brand they provide prescription medications, over-the-counter medications, and personal care products.

Felix Lacroix is Head of Growth at Hims & Hers Canada and runs the company’s entire paid acquisition as a solo media buyer. Felix had been building his own integrations with ad platforms inside a Cursor-based agent to automate his analysis and media buying workflows. With the launch of Converge’s MCP and Agentic Media Buying product, that fragmented stack setup was replaced by a purpose-built agent with the guardrails, security, and learning loop needed to automate performance analysis and move budget autonomously.

Hitting the ceiling of what one person can manually scan

Challenge

A one-person team manually reviewing every channel, campaign, ad set, and creative every day eventually runs out of hours

Felix built earlier AI and automation tools but ran into two walls: AI outputs he could not trust, and off-the-shelf platforms with black-box reasoning he could not audit. Felix wanted to leverage AI to automate his analysis and ultimately media buying in general. The next layer of leverage, agentic media buying, was only going to work if it sat on top of data Felix already trusted.

Result: Without a granular, trustworthy data layer underneath, agentic media buying was a non-starter. The AI was only ever going to be as good as the data it pulled from.

Solution

Automated performance reviews and analysis with the Converge MCP and Cursor

With Converge already capturing clean and granular data across all channels, Felix plugged the Converge MCP into a Cursor-based agent. The agent pulls First Touch and Last Touch CPA from Converge, runs a self-calibrated Meta creative-testing framework that decides which creatives to pause or scale. Every recommendation comes back with reasoning and data sources Felix can audit.

Impact:

10+ hours saved per week analyzing marketing performance.
More frequent and deeper analysis Felix did not have time for before is now available with a single prompt.
Daily reports and recurring analysis happens automatically.
Felix went from a fragmented setup that used many different MCPs and APIs to one system that has full context and doesn’t require maintenance or more time from Felix.

Agentic media buying

Challenge

A custom-built agent kept losing context and could not be trusted to move budget

Even with the Converge MCP plugged into Cursor in combination with the Meta CLI, the agent setup had real limits. It misinterpreted data and hallucinated periodically, forgot instructions it had been given less recently, and was coded without guardrails or security. Felix could not trust it with the ability to actually move budget, so he was still manually verifying and implementing every recommendation.

The agent could also not learn on its own. Every time Felix wanted to change its behavior or teach it a new pattern, he had to write or tweak skill files by hand. A single workflow like Kill Creative Losers took many iterations to get right.

Result: Media buying remained manual, and the tool that was supposed to automate it became another thing Felix had to maintain.

Solution

True agentic media buying through Converge

Converge’s Agentic Media Buying product replaced Felix’s Cursor setup with an agent that makes the same media buying decisions Felix would. Four things made the difference in practice.

It learns automatically from every interaction. When Felix gives the agent a rule (a kill threshold, a gating condition, a scaling trigger), it writes that rule to its memory and applies it consistently on every subsequent run. Felix did not have to re-explain his logic across sessions, and the agent did not silently drop instructions over time, which is what kept happening in Cursor.
It uses Converge as the source of truth for marketing performance. Every recommendation is grounded in the same server-side events and multi-touch attribution Felix already uses to read performance. The agent and the operator are looking at the same numbers, so when the agent proposes a change, Felix can verify the reasoning against the same Converge views he would have used to make the decision manually.
Guardrails sit between the agent and the ad account. The agent surfaces proposed changes with the underlying data and reasoning behind it. Felix reviews and approves before the agent applies changes in Meta. This is the layer Felix could not rely on in Cursor, and it is what makes it acceptable to let the agent operate on a multi-million-dollar budget.
It runs with full account context. The agent reads Felix’s account history, current pacing, channel allocation, and product-level targets before proposing any change, so its recommendations sit inside the operator’s full picture rather than reacting to one metric in isolation.

“I had been writing skills in Cursor for hours trying to get a Kill Creative Losers workflow working and it kept making mistakes,” Felix says. “After my very first interactions, I had already become more confident in the Converge agent than what I had built myself because I noticed my logic was being written to memory so it ran correctly on every subsequent run. 80% of the time I used to spend on media buying is now spent on larger growth initiatives, allowing us to move much faster.”