From Voice Memo to Marketing Board, Automatically

A marketing executive at a mid-market financial services company — about 150 employees — reached out because she knew my background and wanted to see if AI could solve a problem she'd been living with for a while. Her company moves fast, and so does she — ideas surfaced constantly, but almost always while she was in motion: walking between meetings, mid-conversation, on the go. The only way to actually do anything with an idea was to stop, sit down, and type it all out: expand it, draft it, log it properly on the team's board. That stopping and sitting was the real cost. It ran 30-35 minutes per idea, and she was hitting that wall 2-4 times a day — easily several hours a week spent just processing ideas rather than acting on them.

She'd considered just hiring someone to shadow her and capture things in the moment, but that felt like a lot of overhead for what was really a workflow problem. She wanted to know if an agent could do the sit-down work instead, so she didn't have to.

The core flow was one I'd already built for myself: an agent watching a dedicated Slack channel, turning raw ideas into organized, drafted content automatically. But watching how she actually worked reframed the build. The bottleneck wasn't the idea, it was the forced context switch — stop moving, sit down, type. So instead of asking her to change her pace to fit the tool, I built the tool to fit her pace. Now she just talks: a quick voice-to-text drop into a Slack conversation with the agent, without ever breaking stride.

From there, the agent takes over the part that used to require sitting down. But a raw voice-to-text transcript is messy — it's a stream-of-consciousness brain dump, not something you can just drop into a Notion field. So I built a skill that takes that transcript and makes sense of it: understanding what she actually meant, then structuring it into the correct tone, format, and terminology her team's board expected, in the fields they'd already agreed on. (That parsing-and-structuring skill turned into its own project — more on that in a separate case study.)

Her team was already living in Slack for communication and Notion for tracking and planning marketing ideas, so nothing new to learn — I plugged the agent into what already existed. It updates the board directly, which notifies the team, and they pick up right where the agent left off, already 80% down the ideation path.

The build started as a tool for her alone. Now that she's seeing it work, she's looking at opening it up so the rest of the marketing team can drop ideas into Slack the same way and have the agent warm them up for everyone.

Reclaims 4+ hours a week that used to go to stopping, sitting down, and manually processing ideas

• Per-idea time dropped from 30-35 minutes to about 3 minutes
• As a side effect, she can now capture an idea the moment it happens instead of waiting until she's back at a desk — a nice bonus, though the real win was getting the hours back
• Drafts land in the board's correct tone, structure, and terminology — not a raw transcript dump
• The team picks up ideas already 80% down the ideation path, with zero manual handoff
• What started as a personal tool is now being scoped for team-wide rollout

This is how I approach these kinds of problems: watch how someone actually works before building anything. Sometimes that means fitting the system to their current process as-is — but here, the better move was building toward how she actually wanted to work, not the bottlenecked way she'd been stuck in. She didn't need faster typing. She needed to never have to stop and type in the first place.

What makes someone good at their job is the idea, the judgment behind it, and the final polish they put on it — not the Notion table it gets logged into or the formatting it takes to get there. If an agent can absorb the tedious documentation in between and leave someone free to focus on the parts that actually make them successful, that's the win.