Ultranative · Venture Firm
Ultranative’s AI operating stack, built for trust before speed
Ultranative, an Austin venture firm running multiple entities, wanted AI in the middle of its daily operations: the founder’s inbox, the meeting flow, the books. The engagement shipped a working stack in under three months. What made it work was a rule applied everywhere: AI drafts and ranks, a human approves. Nothing runs on autopilot.
Client
Ultranative
Industry
Venture Capital · Multi-Entity
Engagement
Three-month build, extended
Stakeholder
Founder and operations team
The starting point
Ultranative operates several entities with a founder and a small operations team. Email triage was manual. Institutional memory lived in scattered tools and in people’s heads. The bookkeeper reconciled card charges across entities by hand, with weak ground truth on what each charge was for.
The brief was not "show us something impressive." It was: put AI into the workflows that actually run this firm, in a way the team will still trust in month six.
The reframe
Two decisions shaped everything. First, voice before automation: the founder’s email agent was built on a reference set of roughly one hundred sent emails before it drafted a single reply. That groundwork is why the operations team rated draft quality ten out of ten in the first month, instead of spending that month correcting a generic-sounding robot.
Second, review queues instead of autopilot. Every consequential output (an email draft, a reconciled charge, a meeting follow-up) lands in a queue a human approves. In financial workflows this was a hard rule: nothing posts to the accounting system without the bookkeeper’s sign-off. Trust in automation is built by keeping a human hand on the last step.
What we built
An email agent connected to the founder’s Google, Slack, and iMessage that drafts replies in his voice, grounded in the reference set.
A Slack knowledge agent that answers live questions across the firm’s Airtable, Drive, meeting notes, and Notion: who invested what, where a deal stands, what was decided and why.
A meeting dispatcher that turns every recorded meeting into contextual email drafts and action items with named owners, posted where the team already works.
An entity-aware expense reconciliation engine that took 224 raw card charges spanning four months and produced an evidence-ranked review queue for the bookkeeper: 65 charges fully evidence-backed, the rest sorted by what was known and what needed a question, with the questions already drafted.
The strategic implication
The engagement also included a day-long working session that trained an internal builder, who now ships workflows alongside us in Ultranative’s own stack. That is deliberate. A stack the client’s own team can extend survives the engagement; a stack only the consultant understands does not.
For investment firms, the pattern is repeatable: voice-referenced drafting for the principal, a knowledge agent over the systems of record, and review-queue automation for anything touching money. Each piece is useful alone. Together they behave like a chief of staff that never loses context.
Where it goes from here
The reconciliation engine extends naturally deeper into finance operations, ingesting bills and intercompany flows under the same review-queue rule. Multi-entity finance ops proved to be a repeatable pattern of its own: the guardrails that made a skeptical bookkeeper trust the first version are exactly what a CFO needs to see before letting AI near the books.