Date: 2026-10-15 (11:45 AM EDT, ~5 min) Attendees: Michael Houck (michael@ryevalley.co), Emil Barr (not on the Granola participant list; identified from the title and transcript) Granola: https://notes.granola.ai/d/25438063-0172-4671-a905-e9126de5569c
Context: GovGrid is Emil’s business; not a Rye Valley decision. Follows the Oct 13 catchup. Revenue target is in Owner tier. Speaker attribution below is approximate: the whole transcript is on one microphone channel with no speaker labels.
Summary
- Problem (relayed from Will): the deeper GovGrid asks per agency, the more combative agencies get. A request for a few thousand emails gets a conversation; a request for a few million gets a hard no.
- Will’s proposed pivot: go wide instead of deep. Rather than exhaustive data from one agency (example: Ohio tax agency manuals), collect a limited scope from 30 to 34+ state agencies and package a composite: tax workflows and processes across 30 to 48 states, how they differ, with supporting emails.
- Take on the composite: more optimistic than the caller expected. A cross-state comparison has incremental value, but only if the depth within each state is still enough to reconstruct that state’s workflows and processes. The broker calls so far said workflow reconstruction is where the value is. Nobody will know whether the depth is sufficient until data comes back and can be tested.
- Counterpoint raised: buyers are paying for context, so a few emails from many places is less valuable than a lot from one place. Offsetting that: wide collection is far less painful than exhaustive single-agency collection, and the composite could be sold as its own bundled product.
- Goal stated: not just pilots, a large sale closed by end of year, plus a repeatable window.
Decisions
- Run both approaches simultaneously: keep pursuing deep single-agency data and also do a quick wide pull across 30+ states at shallower depth, then assess how many states (20 of 30? 5 of 30?) have usable depth before deciding packaging and sales. See decision page.
Action items
- Michael: tell Will there is optimism in the composite (wide) approach and to run both in parallel
- Unassigned (whoever plays with the data, likely Emil as technical lead) [confirm]: once the wide batch comes back, assess how many of the 30+ states have enough depth to rebuild workflows
- Unassigned: packaging and sales motion for a composite product, decided after that assessment
Notable transcript excerpts
“when Will has sent an email to a few like, hey, we’re interested in a couple thousand emails versus a couple million emails like, the tone shifts have been really dramatic from like, yeah, let’s have a conversation about it to like, fuck off.”
“what if we actually hit up 34 states and got like, had a more limited scope, but collected it across 30 different state agencies and packaged and offered that up.”
“I’m fine with a wider, not deeper approach, as long as the depth we get lets us sort of rebuild that stuff. And we’re not going to know for sure until we start getting some data back that I can play around with.”
“These AI models need a good context. They’re paying for a context window basically. A few emails from a ton of different places isn’t nearly as valuable as a ton of stuff from one place.”
“Do we have enough depth from all 30+ states that we’re going to get? Is it like 20 of the 30? Is it like 5 of the 30? And then we can start to make decisions from there.”
“Okay, so yeah, maybe we’ll do both simultaneously, then I’ll talk to Will.”
Full transcript in Granola (link above).