Chapter One
One-page summary
What this is, who it’s for, and where we stand — in 60 seconds.
Elimentary is building agentic AI for financial institutions — specifically, an AI workflow platform that automates the repetitive 80% of a private-credit analyst’s day. Imagine ChatGPT, but trained for the people who give out loans.
Vision & mission
| Vision | Revolutionise the credit industry using embedded AI. |
|---|
| Mission | Create AI agents for private credit and other credit firms to enable faster, better applications. |
|---|
| Goal | Automate ~80% of the manual, repetitive work — while keeping humans in the loop for override and discretion. |
|---|
| Tagline | “Agentic AI for Financial Institutions.” |
|---|
Who uses it
| User type | What they do | Real-world parallel |
|---|
| Risk Analysts | Watch loan portfolios, predict default risk, monitor covenants | A doctor monitoring patient vitals — for loans |
| Finance Analysts (KPOs) | Generate & reconcile servicer reports, manage fund/treasury | The person making huge Excel sheets every month |
| CXOs (CRO, CFO) | View one big dashboard summarising everything | The boss who only wants the summary |
| Compliance / Audit | Need transparent decision logs | Like CCTV footage — for decisions |
USP
Most competitors ship general-purpose AI agents that finance teams must configure themselves. Elimentary is purpose-built for private credit — agents designed specifically for this team and these workflows, out of the box.
Resolved direction
Direction (locked)
Linear / Cal.com territory. Light surfaces, mint-green primary, neutral table-friendly chrome. Calm, clinical, confident. Influences explicitly named in the client’s own design_tokens.json.
Chapter Four
Source · Public Website
elimentary.com — positioning and brand voice.
Tagline
“Agentic AI for Financial Institutions.”
“Move beyond manual reporting and fragmented core data. Elimentary provides the autonomous infrastructure to unlock liquidity, accelerate lending, and ensure institutional-grade data integrity.”
Five product pillars
- Intelligent Orchestration — post-origination asset lifecycle automation
- Private Credit — loan tape analysis and collateral deep-dives
- Early Warning Signals — delinquency prediction and prevention
- Credit Underwriting — rapid origination with self-learning rules
- Data Integrity — unified source of truth across institutions
Quantified outcomes (web)
| 90% | reduction in underwriting turnaround |
| 25% | reduction in charge-offs |
| 1000× | audit coverage expansion |
| 60× | higher conversion efficiency |
Audiences & voice
Audiences: Credit unions, Private credit firms, Fintechs. Teams: Credit Risk & Strategy, Capital Markets, Risk & Compliance.
Voice: “Professional, institutional, confidence-driven; emphasises operational scale, regulatory compliance, and risk management with measurable outcomes.”
Chapter Five
Source · Client’s AI Prototype
Two videos showing Dharmin building an HTML prototype of Elimentary inside Claude.
What it is
Dharmin used Claude with the canvas/files feature to generate a complete first-pass design system as HTML files: agent_catalog.html, app_shell.html, component_library.html, Color_System.html, dashboard.html, home_empty_state.html, Typography.html, plus design_tokens.json.
The actual videos — click to play
The two screen recordings the client shared. Click to open in a lightbox; press F for fullscreen.
The big reveal — the design_tokens.json
The token file’s description string says: “Comprehensive design system for Elimentary — purpose-built AI for private credit. Finance-grade, dense, professional. Influences: Linear, Cal.com…”
The design north star isn’t a guess — it’s literally written into the client’s tokens.
Two open product decisions in the prototype
- Home model — chat-first vs dashboard-first. Both built. Default not chosen.
- App shell — light shell with mint accents vs darker navy sidebar. Both built. Default not chosen.
Chapter Seven
Module status — build map
Every module mapped to where it lives in the prototype today, and what’s missing.
Status legend: Built = fully functional; Partial = something there but doesn’t match brief; Scaffold = placeholder; Missing = nothing yet; Archived = deferred.
Sheet 1 · Product modules
| Module | In prototype | Status | Missing |
|---|
| 1 · Auth & Onboarding | /login | Partial | Org wizard, tour, sample data |
| 2 · Chat Interface | /chat | Partial | 11 sub-modules; biggest gap |
| 3.1 · Agents Marketplace | /agents | Partial | Marketplace polish, live badges, category chips |
| 3.2 · Agent Detail | /agents/[id] | Partial | Full spec page (capabilities, params, examples) |
| 3.3 · Agent Execution Flow | inside agent run screen | Partial | Step indicator, animated steps, real-time logs |
| 3.x · Three flows (Ingestion / Loan Tape / Service Report) | none | Missing | Parent + 2 specialised variants |
| 4 · Agent Creator | /workflows | Built | Polish; align labels with Make.com reference |
| 5 · Dashboards | /dashboards + /dashboard | Partial | 4–5 layout templates (Cash Flow, Risk Heatmap, Insights, Deep Dive) |
| 6 · Excel Plugin (IMP) | /excel-plugin | Scaffold | Whole module |
| 7 · ML Pipeline | /ml-pipeline | Scaffold | Whole module — brief is empty |
Sheet 2 · Settings & Profile
| Sub-module | In prototype | Status | Missing |
|---|
| 6.1 · Team Members List | /settings/team | Built | — |
| 6.2 · Invite Users | /settings/team | Built | — |
| 6.3 · Roles & Permissions | /settings/team | Built | Capability matrix polish |
| 6.4 · User Profile & Preferences | /settings/profile | Built | — |
| User Profile Page (org details, access scope) | /settings/profile | Built | — |
| Prompt Library | /settings/prompt-library | Built | Bulk actions, version history |
| Skills Management | /settings/skills | Built | Per-agent overrides |
| Workflow Monitor | /settings/workflow-monitor | Built | Live tail, retry-with-changes |
| Tools Management | /settings/tools | Built | Per-tool config (e.g. SQL warehouse) |
Sheet 3 · Archived (out of scope per client)
| Module | In prototype | Status | Note |
|---|
| 9.1–9.7 · Plugins & Connectors (Marketplace, details, install, connectors, setup wizard, webhooks) | none in IA — partial files in /settings/integrations + /settings/data-sources | Archived | Confirm: phase-2 or genuinely cut? |
| 10.1–10.3 · Session History (Recent sessions, search/filter, tagging) | /my-work (in “Quick access”) | Partial | Already partly built — client marked it archived but it’s useful enough to keep |
Surplus in prototype (not in client’s brief)
| Module | Where | Disposition |
|---|
| Predictions | /predictions | Hidden from sidebar; file kept. Closest in brief: “Default Risk Predictor” agent. |
| Portfolio (companies) | /portfolio, /portfolio/[id] | Kept under “Quick access.” Useful entity even if not a named module. |
| Reports | /reports | Kept — mentioned in client’s prototype IA even though not in XLSX module list. |
| Alerts | /alerts | Kept under “Quick access.” Maps to Module 5.3 Breach Alerts. |
| Audit Log, Billing | /settings/audit-log, /settings/billing | Kept under “Account & data.” Not in client’s brief but useful. |
Chapter Eight
Open gaps & questions for client
Everything to flag back to Manoj & Dharmin.
Audience & positioning
Blocker PDF says “credit unions and community banks”; task analysis & website lean into “private credit / KPOs / fintechs.” Which is the design north star?
Important Within “private credit,” who’s the daily primary user — analyst or CXO?
The existing product
Blocker Credentials for credit-intelligence.elimentary.com — need to audit before tagging screens.
Brand & visual
Important Do brand guidelines exist? PDF asked; no answer.
Important Reference apps “shared separately over WhatsApp” — can we re-share?
Important Mint-green direction in Dharmin’s prototype — locked, starting point, or one of several?
Module scope
Blocker Module 7 (ML Pipeline) — need description, flow, reference. Brief is empty.
Important Module 6 (Excel Plugin) — sidebar pane only or also ribbon? Web/Mac/Windows? Same login as web?
Important Settings sheet — most rows are headings only.
Important Sheet 3 (“Archived”) — genuinely cut, or phase 2?
Important Automations — new sub-module? Your note in the XLSX (next to Agent Execution Flow) says: “Automations as a section can be thought of for use-cases such as 'Generate monthly Service Report for this week and mail to xyz'.” Is this an additional module to design now (scheduled / triggered agent runs — trigger config, recipient list, retry policies, like Zapier Zaps), or a thought to park for later?
Blocker XLSX columns Screen Components / Entry Point / States are blank for almost every row.
Two open product decisions in prototype
Important Home model: chat-first vs dashboard-first?
Important App shell theme: light vs darker navy?
Chapter Ten
References & inspirations
Everything the client mentioned, in one place — brand inspirations, competitors, reference URLs, source files.
Brand & visual inspirations
From the kick-off PDF and the prototype tokens.
PROTOTYPE TOKEN
Linear
Named in the client’s design_tokens.json. Calm, dense, keyboard-first. Slate-tinted neutrals.
linear.app
PROTOTYPE TOKEN
Cal.com
Named in the client’s design_tokens.json. Mint-green primary, light surfaces, generous whitespace.
cal.com
PDF · BRANDING ADMIRATION
Wisprflow
Brand the client admires — modern, fluid, AI-native voice product.
wisprflow.ai
PDF · BRANDING ADMIRATION
Slack
Brand the client admires — warm, friendly, productivity-tool feel.
slack.com
Personality / voice references
From the PDF — “Who would be the perfect face of your brand?”
PDF · BRAND FACE
Jamie Dimon
CEO of JP Morgan Chase. Institutional weight, gravitas, financial authority. The “steady hand at the wheel” figure.
PDF · BRAND FACE
Boman Irani
“Reliant, steady, expert but yet super approachable.” The personality lever the brand should pull — warmth without losing authority.
Additional inspiration apps
Additional product references for feel and craft bar.
INSPIRATION
11x
AI workforce / agents marketing site. Bold typography, confident voice, dark editorial aesthetic.
11x.ai
INSPIRATION
Mymind
Editorial-grade product site. Calm, considered typography — rare craft bar for AI product UI.
mymind.com
INSPIRATION
Superpower
Health intelligence app — institutional + personal in one. Useful precedent for tone.
superpower.com
AI AGENT PLATFORM
Artisan
AI sales reps / SDR agents. Bold positioning, clear “agent as employee” framing — useful precedent for how to talk about agents.
artisan.co
AI AGENT PLATFORM
Sierra
Conversational AI for customer experience. Calm, enterprise-grade UI — close in tone to where Elimentary should land.
sierra.ai
AI AGENT PLATFORM
Decagon
AI customer service agents. Strong agent-detail and execution-flow patterns worth studying.
decagon.ai
AI AGENT PLATFORM
Ema
“Universal AI employee.” Onboarding flow + skill marketplace patterns are worth a close look.
ema.ai
AI AGENT PLATFORM
Cresta
AI for contact centers. Real-time AI assistance + agent monitoring patterns — mirror of Workflow Monitor.
cresta.com
AI AGENT PLATFORM
Relevance AI
AI workforce platform — agent builder + tools mgmt patterns directly comparable to Modules 4 & Settings.
relevanceai.com
Competitor landscape
From PDF Section 5. Primary = direct head-to-head; Secondary = adjacent or aspirational.
Primary competitors
PRIMARY COMPETITOR
ModelML
Direct competitor in agentic AI for finance.
modelml.com
PRIMARY COMPETITOR
Rogo
AI for financial analysts — widely benchmarked in the space.
rogo.ai
PRIMARY COMPETITOR
Essential AI
Foundation models with finance-specific applications.
essential.ai
Secondary competitors
SC
SECONDARY · LIKED
Shortcut AI
“Agent-based workflow automation — AI feels operational rather than theoretical.”
tryshortcut.ai
SECONDARY · LIKED
Lyzr AI
“Agent-based workflow automation — AI feels operational rather than theoretical.”
lyzr.ai
Kd
SECONDARY · LIKED
Kdense
“Focus on financial data processing and model output clarity.”
Nx
SECONDARY · LIKED
Nexttoken Co
“Focus on financial data processing and model output clarity.”
Other named in XLSX
XLSX · DASHBOARDS REF
Spindle AI
“Take Spindle AI inspiration for this section” — named for Insights, Breach Alerts, Drill Down (Module 5).
XLSX · AGENT CATALOG REF
Google Agents · Claude Skill Library
“Imagine something similar to Google Agents and/or Skill library of Claude” — for the Agent Catalog (Module 3.1).
XLSX · CHAT EMPTY REF
Claude / ChatGPT
“Empty chat state of Claude/ChatGPT” — named pattern reference for Module 2.1.
Module-level reference URLs
From the XLSX module sheet — specific links the client gave for specific modules.
EXISTING APP · TO AUDIT
Credit Intelligence (existing app)
Referenced for dashboard layouts (Module 3) and Insights (Module 5.2). Credentials needed.
credit-intelligence.elimentary.com
XLSX · MODULE 4 REF
Make.com
Reference for the Agent Creator (no-code visual canvas, node palette, test & publish).
make.com
PUBLIC SITE
Elimentary website
Public positioning, taglines, and the five product pillars.
elimentary.com
Source files we’ve received
References still pending
Not yet shared
The kick-off PDF says: “Reference apps — shared separately over WhatsApp.” These haven’t been catalogued. Re-share request is in the open-questions list (Ch. 8).
Other influences worth keeping on the shelf
DESIGN PARTNER REF
Stripe
Calm institutional aesthetic, dense data presentation done with restraint.
stripe.com
DESIGN PARTNER REF
Hebbia
Considered for “Direction B”: warmer cream canvas, signature blue numbers, denser tables.
hebbia.ai
n8n
DESIGN PARTNER REF
n8n / Zapier
Comparable visual canvas patterns for Agent Builder — alongside the client’s explicit make.com reference.
