Kenda reconciles measured LLM usage against what providers actually bill (OpenAI, Anthropic, Google), producing a single reconciled score (the Kenda Index) that tells a CEO whether their AI spend is actually working
Kenda reconciles measured LLM usage against what providers actually bill (OpenAI, Anthropic, Google), producing a single reconciled score (the Kenda Index) that tells a CEO whether their AI spend is actually working. Not a dashboard, a verdict. Most people treat AI spend as an observability problem (instrument calls, multiply tokens by price sheet, dashboard). It's an accounting problem. Token math structurally can't match the bill: cached tokens bill differently, batch discounts land after the fact, fine-tuned models don't match registry prices, each provider does it differently. The gap between estimate and invoice isn't a bug, it's a product of how providers bill, and it widens with every pricing lever they add. Also from six years in billing: systems lie about themselves, and the truth lives in the counterparty's records. Your telemetry says one thing, the invoice says another, and the money is in the gap. Built reconciliation tools before for healthcare billing. At Clarity, shipped exception queues and delivery-record reconciliation workflows for operations teams. Knows what the person chasing exceptions needs on screen because she built their screens and sat with them. Own first customer: ships LLM products across providers, built Kenda from hackathon to live on her own. Instinct runs against her own interest (notices datacenter cities inflate user numbers, claims 100+ instead of 260). Also organized React BA (8k members), got 42 signups in 48h for a tech event with zero budget, Luma featured it. Proven demand generation muscle. AI spend didn't exist as a line item three years ago. Now it's a COGS line at every AI-native company, agents are multiplying the spenders, and the person who has to explain the invoice is finance. Finance doesn't run on estimates, it runs on reconciliation. The observability tools (Langfuse, Braintrust, gateways) compute cost from token counts and call the result an estimate, but the buyer has changed from an engineer to a finance owner. Nobody is building the system of record for that person.
The challenge of accurately identifying all the known wearers of a product, especially on social media, due to platforms like Instagram limiting access to accounts with a large following.
Kenda facilitates data analysis for brand producers and agencies by categorizing customer databases and identifying potential new wearers across various platforms using AI.