Portfolio · AI Product Build

otlo

An AI-native community management and intelligence platform — scoped as a product build engagement, distinct from the venture story behind it.

The scope

The build layer, isolated.

otlo is one of our own ventures, and its venture log covers the positioning and validation work end to end. This page scopes narrower: the AI-native platform itself — the signal aggregation core, the weekly-digest generation and the evaluation harness behind both — delivered the way an AI Product Build engagement runs for anyone.

Isolating it this way shows what a team buying only the build scope, with the strategy already settled, actually receives.

The brief

Signal, not another dashboard.

The build had to turn cross-platform engagement data into one stated conclusion per week, not a chart to interpret — which meant the hard problem was evaluation (is the conclusion actually right) as much as aggregation (is the data actually complete).

The approach

Evaluation before scale.

01Aggregation core, three platforms firstFewer connected sources at launch, in exchange for a verifiable signal on the ones design partners actually use.
02An eval set for "the conclusion is right"Weekly digests scored against what design partners independently identified as the actual driver, before the feature shipped wider.
03Agentic generation, human review by defaultThe digest draft is machine-generated and reviewed before send until the eval scores justify full automation.
Delivery

What the engagement produced.

01
Aggregation architecture
Cross-platform signal pulled into one data model, scoped to the first three platforms.
Weeks 1–4
02
Digest generation core
Agentic summarisation producing the one-conclusion-first weekly format.
Weeks 5–9
03
Evaluation harness
A scoring set built from design-partner-verified conclusions, run against every digest before send.
Weeks 10–12
04
Observability and handover
Cost, latency and accuracy visible per digest, documented for the team running it.
Week 13
Signal aggregation core
Cross-platform reach and engagement pulled into a single data model.
Digest generation
Agentic summarisation producing the one-conclusion-first weekly format.
Evaluation harness
Every digest scored against a design-partner-verified answer before it ships.

No — the venture log covers positioning and validation. This page scopes just the AI-native build: aggregation, digest generation and the evaluation harness behind it.

Yes. A production AI system with evaluation and observability, built where the architecture is justified, is the AI Product Build engagement — run here on our own venture first.

Scoped to requirement

We operate, we do not observe.

We build alongside the founding team and hand over something they can run without us.

Talk to the studio