Lighthouse / Independent product
Making customer signals easier to act on.
In Customer Success, knowing an account needs attention is only the beginning. The harder question is which signals matter, what they mean together, and where to focus next.
I built Lighthouse to explore that problem in a working account-intelligence application. It brings account context, portfolio health, renewal exposure, and AI-assisted briefings into one place, without pretending a score can manage a relationship.
Independent working demo · Fictional account data · No production CRM integration
The problem
A portfolio is more than a list of accounts.
Customer information can sit across usage trends, support activity, renewal dates, satisfaction measures, and notes from individual conversations. Looking at one number at a time makes it easy to miss the bigger picture.
The product question was straightforward: could I create a workspace that helps someone identify where to look first, understand why, and prepare a better conversation?
Product decisions
Three choices that shaped Lighthouse.
Start with the portfolio, not the AI.
A Customer Success manager needs to see where attention may be warranted before asking for an analysis. The Portfolio view brings account health, ARR, renewal dates, and usage changes into a comparable list, with sorting and clear overdue states.
Make the signal inspectable.
A health indicator can help prioritize, but it is not a prediction of churn or a substitute for knowing the customer. The Accounts view keeps the underlying usage, support, and satisfaction context close to the score so the person managing the relationship can investigate.
Let AI prepare, not decide.
Lighthouse can generate account analyses, possible expansion plays, executive briefings, and draft outreach. These are starting points for human review, not verified forecasts or automated account decisions. When live AI is unavailable, the interface identifies its demo fallback rather than presenting it as a real response.
Tradeoffs
Useful signals, explicit limits.
The dataset is fictional and small. The health score is a rule-based demonstration, not a calibrated risk model. Lighthouse is not connected to an employer's CRM, and it has not established measured adoption, retention improvements, or commercial outcomes.
AI-generated recommendations may contain uncertain claims. The interface identifies generated versus fallback output, but every suggested action and future outcome still needs a person to check the evidence and apply account context.
Current status
A working demonstration, not a finished platform.
Lighthouse has working Accounts and Portfolio views, responsive light and dark themes, and optional AI-assisted analysis. The next meaningful test is not adding more features. It is learning whether these views help real Customer Success professionals identify priorities and prepare better conversations.