Portfolio · AI Product Build

otlo

AI-Native Community intelligence platform for the people who build communities - distribution, reach and signal in one surface.

The thesis

A published positioning line, and a stated category.

otlo carries a published positioning line and a stated category — community intelligence — both venture-strategy artefacts. The validation work is deciding whether the thesis survives contact with the people it names.

Community managers and creators currently piece together reach and engagement signal from five disconnected platform dashboards, none of which talk to each other and none of which say what actually drove growth. The bet is that the aggregated signal itself — not another posting tool — is the product.

The challenge

Proving signal, not just aggregation.

Dashboard aggregation alone is a commodity — half a dozen tools already pull platform metrics into one screen. The hard problem is turning that data into a specific, actionable read: which post, which cohort, which channel actually moved the number that mattered, stated plainly enough to act on.

A second challenge is trust: community managers are protective of their audience data and skeptical of another tool asking for platform access. Adoption depended on being useful within the first session, not after a lengthy onboarding.

The approach

One conclusion per week, not a dashboard to interpret.

01Signal over surface areaFewer connected platforms at launch, in exchange for a sharper read on the ones that matter most to the first cohort.
02A stated conclusion, not a chartEach weekly digest leads with one sentence — what worked, what did not — with the supporting data underneath rather than in front.
03Design partners named in publicEarly access limited to a small group of named community builders whose feedback shaped the read before wider release.
The build

Aggregation, generation and evaluation.

Underneath the positioning sits a production AI system: signal aggregated across the platforms design partners actually use, a digest generated agentically against a one-conclusion-first format, and an evaluation harness that scores every digest against a design-partner-verified answer before it ships. Human review stays in the loop until the eval scores justify full automation.

Where it stands

In validation.

01PositioningPublished and held for three months without a rewrite.
02Design partnersA named cohort, not anonymous beta users.
03Open questionWhether the weekly read alone is worth paying for, or whether it needs a second surface.
Delivery so far

What exists today.

01
Positioning and category definition
The "community intelligence" line, tested against design partner language before publishing.
Weeks 1–3
02
Aggregation core
Signal pulled from the first three platforms design partners actually use.
Weeks 4–8
03
Weekly digest format
The one-conclusion-first format, iterated against what partners actually read versus skipped.
Weeks 9–11
04
Evaluation harness
A scoring set built from design-partner-verified conclusions, run against every digest before send.
Weeks 12–13
05
Validation cohort
A named group of community builders using the digest weekly, feedback logged in public.
Ongoing
From the product

A look at the surfaces.

The weekly digest
One conclusion, stated first — the format the validation is testing.
Signal aggregation
Cross-platform reach and engagement pulled into a single read.
Design partner cohort
A named group of community builders shaping the product in public.
More from the build

Additional surfaces and detail.

Evaluation harness
Every digest scored against a design-partner-verified answer before it ships.
Digest generation
Agentic summarisation producing the one-conclusion-first weekly format.
Observability view
Cost, latency and accuracy visible per digest, for the team running it.

Aggregating a fourth or fifth platform is easy; proving the first three produce a conclusion worth paying for is the actual risk. Validation targets the risk, not the easy work.

Either the design partner cohort converts to paid at a rate that supports the business, or it does not — the honest outcome is sometimes that the thesis does not survive, and we say so in public.

Venture log

Validation is a decision, not a phase.

The honest outcome is sometimes that the thesis does not survive.

Read the log