CRM Lexinton · Inquiry intakeWhat arrives in writing is never retyped.More than 3,500 inquiries came in this way, with no manual entry.6 channels · inbox read every 15 minEach inquiry arrives withcontactpropertyagentmanual entryOctober 28, 20267 min readSix channels, zero manual entry: how a real estate agency's inquiries come in on their ownHow I got inquiries from listing portals, Facebook and the website into a real estate agency's CRM on their own: no duplicates, with their property and agent.IntegrationsAutomationBest practicesRead the article
Classification with AI3.2 s0.5 sper case, in the ones Jev handles on its own, without losing a single correct answerAccuracy: 105 of 105, with and without JevPicks from a closed listno answercall another dayopt-out97%wrong numberOctober 2, 20266 min readAn AI that chooses instead of writing: from 3 seconds to half a second with JevI added Jev, a model that classifies in half a second, in front of an OpenAI model: it handled 33% to 41% of the cases on its own, cheaper and with no loss in accuracy.AIAgentsEvaluationRead the article
TomaNota · TasadorThe AI doesn't write the price.The model picks products; the system calculates the total.8 tools · none accepts a priceWhat the model can passcatalog productquantitypriceSeptember 24, 20267 min readAI agents that don't make up numbers: where the model ends and the code beginsHow I design AI agents that don't make up prices: the model understands and writes, and totals and money go through tested code. Two real cases.AIAgentsBest practicesRead the article
CRM Lexinton · Official WhatsAppThe client replies. The automation goes quiet.7 flows · only 8 AM to 7 PMThe client repliesScheduled follow-upcanceledScheduled remindercanceledSeptember 22, 20265 min readThe official WhatsApp API inside your system: a shared inbox and follow-ups that stop on their ownHow I integrated the official WhatsApp API into a real estate agency's CRM: an inbox for the whole team, Meta templates and follow-ups that stop on their own.WhatsAppIntegrationsAutomationRead the article
Tasador · 113 queriesI built 4 pieces. I switched on 1.The simplest, because it won the measurement.nDCG@25: 0.747 → 0.821LexicalDenseFusionRerankerSeptember 18, 20266 min readBefore you switch on the AI, measure itI built a full RAG to value properties, measured it on 113 queries and the simplest option is the one switched on. What to measure before paying for AI.AIRAGEvaluationRead the article