Bottomline AI

Inspiration

Before this, I spent five years running product at a half-billion-dollar SaaS. Every Monday started the same way — scanning the loudest Slack threads, skimming the support tickets flagged as "strategic," cross-checking a few competitor changelogs, then making roadmap calls off whatever stuck.

Then I'd watch deals we should have won slip in the pipeline anyway.

The honest truth: I was making seven- and eight-figure product bets off maybe a few percent of the signal. My customers were telling me exactly what to build, what to charge, and which competitor was winning their deals — across Reddit, Hacker News, community Slacks, GitHub issues. I just couldn't read all of it.

The pain wasn't a lack of data. It was: the data exists, and I can't act on it in the time I have.

Bottomline AI is the tool I wish I'd had.t it as a Product Manager.

What it does

Bottomline AI gives product leaders three surfaces over one evidence base of their market.

Ask. A conversational interface where PMs ask the questions they're already asking themselves on Monday — "What's driving churn this quarter?", "Which competitors are winning deals we're losing?", "Is my pricing stale?" — and get grounded answers with citations back to the source conversations, tickets, and competitor pages.

See. A dashboard of top signals, ranked by expected P&L impact. The pricing moves, problem clusters, and competitive threats worth acting on this week — scored so the loudest noise doesn't drown out the biggest lever.

Strategize. The ranked signals are the product strategy. For every top-ranked problem cluster, pricing threat, or competitive move, Bottomline AI drafts a candidate initiative with the evidence bundle attached — so the roadmap debate starts from "do we agree with the ranking?" instead of "what do we even know?"

Behind all three surfaces: every public signal — Reddit threads, HackerNews comments, GitHub issues, competitor pricing pages, YouTube reviews — gets extracted into typed entities (problems, features, competitors, pricing moves, metrics) and linked into a causal network. This problem causes churn. That feature solves this pain. Open the graph and you literally see the shape of your market: clusters of problems around your product, edges connecting competitor features to lost deals, pricing entities clustered around "too expensive for SMB."

Every answer grounded in evidence. Every signal ranked by what moves the bottomline.

How we built it

An ingestion layer pulls signals from the channels PMs can't keep up with. An extraction layer identifies typed entities and the causal relationships between them. A vector layer finds semantic neighbors, so related complaints cluster even when the words differ. A graph layer connects everything into a navigable causal network.

For Strategy, grounded in data as well as best product strategy frameworks used at Big Tech

Challenges we ran into

Ingesting and chunking data from heterogeneous sources
Handling ingestion errors

Accomplishments that we're proud of

A 3D knowledge graph rendering 800 causal edges across 1,296 connected nodes, with animated particles pulsing along every edge — the first time I could see a product category instead of read about it.
An end-to-end pipeline — collect → extract → embed → link → cluster → rank → render — resumable and idempotent from first run.
An interactive legend that toggles entity types in and out of the graph in real time. Click a node, the camera flies to it, the panel tells you its role in the market.

What we learned

The hard part isn't the model, it's the pipeline.
A knowledge graph only earns its keep when you can see it.
PMs don't need more data. They need bottomline AI.

What's next for Bottomline AI

Expand to other strategic personas in company - Marketing Strategy, Pricing Strategy, Sales Strategy
Enrolling early adopters for bottomline.ai for Product Manager personas. Email [email protected] if interested