← Intelligence Hub
Product Intelligence • Volume 6

What We've Built & Why

Every feature maps to a data point. Every decision is traceable to research. This is the product-to-research pipeline — proof that we build what matters.

The Product Promise

"Upload your documents once. The AI handles the rest — alerts, estimates, savings, filing prep — and gets smarter every time you talk to it."

Lesser is built for one person: a 28-42 year old South Asian tech professional with RSUs, foreign accounts, and a spouse who files jointly. We know this because we studied 1,885 of them. → See composite client

What's Live in Production

Core Platform

FeatureWhat It DoesResearch Backing
AI Document ExtractionUpload PDF → structured data in seconds. W-2, 1099s, K-1, 1098, prior 1040, paystubs.Prior 1040 is #1 requested doc (133 calls, 20.9%) → Finding
AI Tax AdvisorStreaming AI chat with 60+ financial fields loaded. Dollar-anchored insights. Proactive on first message.49% of calls are AI-automatable → AI opportunity
Psychographic Emails40 copy variations (5 archetypes × 8 milestones). Zero AI calls — static lookup.5 archetypes from 705 calls → Psychographic
Archetype EvolutionClient archetype evolves at 3 triggers: onboarding, extraction, every 5 chat messages.Emotions stack and shift → Evolution system
Alert System18 alert types (RSU cost basis risk, over-401k, wash sale + IRA, FBAR required, etc.)IRS fear is #1 (586 mentions) → Fear patterns
Federal Tax EnginePure TypeScript. All 2025 brackets, LTCG, AMT, NIIT, CTC/EITC. Includes OBBB changes.Clients need instant estimates → Tech details
State Tax Engine~20 states: CA (9-bracket), NY+NYC, WA cap gains, TX, NJ, MA, IL + more.34.5% of consults involve multi-state → Finding
Expert Pipeline22-status workflow. Admin kanban. Client assignment. Meeting scheduling via Cal.com.Meeting no-shows = #1 service failure → No-show data
BillingStripe checkout with on-demand session creation. 4 pricing tiers validated.2.2:1 accept ratio when value shown → PMF signals
Referral SystemReferral tracking with reward amounts.Referral is #1 acquisition → Acquisition data

Decisions Backed by Data

Every significant product decision traces to a specific finding from the intelligence library.

01

"NRI Tax Expert" not "H1B Specialist"

77% of foreign-asset clients are NOT on H1B. Broadened positioning to capture full market. → Demographics

02

Dollar-Anchor Rule in AI responses

Clients respond to specific dollar amounts, not vague advice. 150 calls showed pricing acceptance after one concrete value demo. → Sales mechanics

03

Warmth-first, features-second design

Relationship signals (1,879) beat expertise signals (1,200) beat price signals (518). Product tone follows. → Soul truth

04

Archetype-aware email system

68% are trust-seeking. Generic "Action required:" emails erode trust. Rewrote all 28 status emails per archetype. → Shipped system

05

AI handles first-line support

49% of consultation calls involve questions AI can answer — pricing (43.6%), timelines (20.7%), filing status (18.7%). → AI opportunity

06

Weekend + evening expert coverage

34.4% of calls happen on weekends. Peak hour is 11 PM local (9:30 AM IST). Staffing matches demand. → Temporal data

07

Prior year 1040 upload in onboarding

Most-requested document (133 calls, 20.9%) — outpaces even W-2 requests. Clients want to compare. → Recommendation

08

Proactive insights on first message

Competition answers questions. Lesser anticipates them. Ranked dollar-impact findings computed before user types. → Atlas principles

Product Roadmap (10 PRDs Written)

Every planned feature has a full Product Requirements Document with user stories, acceptance criteria, and data backing.

PRDFeatureResearch LinkStatus
PRD-01Auto-Draft GenerationExtraction pipeline → return intelligenceIn Progress
PRD-02Smart Document Requests81% of doc requests are for missing types → GapPlanned
PRD-03Morning Briefing (Admin)Expert efficiency — route by specializationPlanned
PRD-04Draft Ready NotificationArchetype-aware email when draft is completeShipped
PRD-05Invoice Follow-up AutomationPayment delays → automated sequencesPlanned
PRD-06Weekly Pipeline DigestAdmin visibility into 22-status pipelinePlanned
PRD-07Meeting Prep SummaryAI generates pre-call brief from extracted dataPlanned
PRD-08Quick ActionsReduce expert clicks for common operationsPlanned
PRD-09Ask Lesser AI (Public)298 real Intercom questions → training dataResearch
PRD-10Time-Sensitive Smart Alerts176 "deadline" fears → proactive remindersIn Progress

Known Gaps (Prioritized)

From analyzing 519 document requests across 705 consultation calls — the documents clients need that we don't yet collect automatically.

Document GapRequests% of AllAction
Prior year 104013320.9%Added to onboarding
Business expenses12018.8%PRD-02
Foreign bank statements6610.4%PRD-02
E-file authorization (8879)6410.0%Automated
State tax forms375.8%Planned
Pay stubs345.3%Extraction live
Estimated tax payment records253.9%Planned
The takeaway

81% of all document requests are for types not in the original upload flow. Each gap we close reduces expert call load and improves time-to-completion. This is the difference between building features from assumptions vs. building from 705 real conversations.

← Previous
Growth Intelligence
All Chapters →
Intelligence Hub

Product Intelligence • Volume 6 • 10 PRDs, 18 alert types, every decision data-backed

Part of the Intelligence Hub