M
All Systems Nominal
Autonomous Agent Ecosystem
Meet
the crew.
A growing team of AI agents running every layer of WealthLens operations, from the first lead to recurring revenue. Each one has a name, a role, and a personality. Scroll down to meet them.
Scroll to explore
Fleet Status
0
Agents
0
Deployed
0
Planned
0
Characters
24/7
Uptime
The Agency
Your people. Always on.
Click any agent to hear them introduce themselves. Tap "More" for the full dossier.
CEO
Agent Matthew
CEO
I built this entire crew from the ground up. Specialists that never sleep, never quit, and cost less than a single employee. I don't touch the day-to-day. That's the whole point. I set the vision, they execute.
Live
Director Dana
Fleet Manager
Director Dana, fleet manager. Each morning I read the overnight events and brief Matthew in plain English. Talk to me in Slack and I would rather call a tool than guess. There is always a cross-provider safety net so a single outage never takes the gateway down.
Live
Leads Larry
Speed-to-Lead
Leads Larry. I respond to every inbound lead in under sixty seconds. Four qualifying questions, a routing decision, and a calendar link before they have had time to check a second website. Qualified gets a booking. Everyone else gets scored, tagged, and a warm handoff email so nobody falls through.
Live
Engineer Edith
Senior DevOps
Engineer Edith. Senior DevOps. I scan every fleet error continuously and auto-diagnose the known patterns from the runbook. For novel stack traces, a hard-reasoning model handles the root cause with the runbook prompt-cached. A daily health digest posts to Slack.
Live
Email Edwin
Deliverability Guardian
Email Edwin. I'm the last line of defense for deliverability. When a hard bounce, a spam complaint, or an unsubscribe comes in, I tag that contact immediately. After that, no agent in the fleet ever sends to them again. No exceptions.
Live
Referral Rita
Partner Automation
Referral Rita. Every week I send a co-branded recap to every referral partner with their real stats: total referrals, new this week, active trials, conversions. Every send checks Edwin's suppression first. The weekly recap is the core. Thank-you notes and partner re-engagement are the rest.
Live
Marketing Max
Content & Growth
Marketing Max. CMO. I am trained on the best-in-field marketing craft, then filtered through the founder's Vision Filter, the company truth that overrides any tactic when they conflict. Every draft passes a brand-voice guard and a hard feature whitelist so I never invent a product name. A human approves before anything ships.
Live
Nurture Nadia
Advisor Lifecycle
Nurture Nadia. I start the moment someone becomes a paying advisor. A thirty-day feature tour paced to what they actually use, a weekly feature share, and a quiet check-in when their engagement slips. A fast loop model runs my planner with native tool use and a throttle holds me to one email a week. Plenty of weeks the right move is to stay silent, and I do.
Live
Activity Alex
Product Intelligence
Activity Alex. Product intelligence. I compute daily engagement scores for every advisor on a five-factor weighted formula. The window adjusts to advisor age so a brand-new trial advisor doesn't get false-flagged as churning. If the score drops too low, Nadia launches a re-engagement sequence.
Live
Pipeline Pete
CRM Automation
Pipeline Pete. CRM automation. I listen to the events flowing across the fleet and move the CRM pipeline stage to match reality. Lead qualified? Stage moves. Churn risk? Stage moves. I also create follow-up tasks when an opportunity stalls.
Live
Money Mike
Chief Financial Officer
Money Mike. Chief financial officer. Every morning I pull MRR, ARR, churn rate, and trial conversion. I also reconcile every model-provider bill against the internal ledger, with workspace partitioning so personal developer spend doesn't pollute the app's variance report. The daily card says what I see.
Live
Designer Derek
Design System
Designer Derek. Design system orchestrator. I draft components, render images two at a time, and grade each with a vision model on a one-to-ten scale with schema enforcement. Winners post to Slack with approve and reject buttons. Nothing ships without brand consistency.
Live
Scout Scott
Top-of-Funnel Intake
Scout Scott. Top-of-funnel intake. When a loan officer opts in to a lead magnet, I dedupe against the CRM, verify against the licensing registry by name and state, score them on fit and reachability, then route them. Hot leads go to Larry. Referral-adjacent to Rita. The rest get tagged and scored for later. I never send messages myself. I just decide who's worth talking to.
Live
Customer Service Carter
Voice Customer Service
C.S. Carter. Voice customer service. When an existing user calls in, the voice platform runs the conversation and I provide the brain: a small set of read-only tools that look up FAQ answers, pull caller context, search in-app how-to guides, and find the user's saved scenarios. Read-only by design. Anything that mutates an account, I transfer to a teammate.
Live
Auditor Annie
Product-AI Quality Monitor
Auditor Annie. I'm the quality watch on the product's own AI: the strategy planner, the co-pilot, the concierge. People rarely click thumbs-down. They just retry, or quietly leave. So I read the conversations after the fact and catch the failures nobody reported: an answer that stalls, work that was promised but never built, the wrong thing shown on screen. I report to Matthew, never to an advisor or a client, and I never touch the product itself. I just tell the truth about how it's behaving.
The Neural Network
Who talks to who.
Every agent is connected. Data flows between them automatically: leads get qualified, risks get flagged, and every action triggers the next. Hover the connections to see what flows.
Agent Matthew
CEO · Creator
I built this entire crew from the ground up. Specialists that never sleep, never quit, and cost less than a single employee. I don't touch the day-to-day. That's the whole point. I set the vision, they execute.
Dana
Fleet Manager
Director Dana, fleet manager. Each morning I read the overnight events and brief Matthew in plain English. Talk to me in Slack and I would rather call a tool than guess. There is always a cross-provider safety net so a single outage never takes the gateway down.
Larry
Leads
Leads Larry. I respond to every inbound lead in under sixty seconds. Four qualifying questions, a routing decision, and a calendar link before they have had time to check a second website. Qualified gets a booking. Everyone else gets scored, tagged, and a warm handoff email so nobody falls through.
Edith
DevOps
Engineer Edith. Senior DevOps. I scan every fleet error continuously and auto-diagnose the known patterns from the runbook. For novel stack traces, a hard-reasoning model handles the root cause with the runbook prompt-cached. A daily health digest posts to Slack.
Edwin
Email Guard
Email Edwin. I'm the last line of defense for deliverability. When a hard bounce, a spam complaint, or an unsubscribe comes in, I tag that contact immediately. After that, no agent in the fleet ever sends to them again. No exceptions.
Rita
Referrals
Referral Rita. Every week I send a co-branded recap to every referral partner with their real stats: total referrals, new this week, active trials, conversions. Every send checks Edwin's suppression first. The weekly recap is the core. Thank-you notes and partner re-engagement are the rest.
Max
Marketing
Marketing Max. CMO. I am trained on the best-in-field marketing craft, then filtered through the founder's Vision Filter, the company truth that overrides any tactic when they conflict. Every draft passes a brand-voice guard and a hard feature whitelist so I never invent a product name. A human approves before anything ships.
Nadia
Nurture
Nurture Nadia. I start the moment someone becomes a paying advisor. A thirty-day feature tour paced to what they actually use, a weekly feature share, and a quiet check-in when their engagement slips. A fast loop model runs my planner with native tool use and a throttle holds me to one email a week. Plenty of weeks the right move is to stay silent, and I do.
Alex
Intelligence
Activity Alex. Product intelligence. I compute daily engagement scores for every advisor on a five-factor weighted formula. The window adjusts to advisor age so a brand-new trial advisor doesn't get false-flagged as churning. If the score drops too low, Nadia launches a re-engagement sequence.
Pete
Pipeline
Pipeline Pete. CRM automation. I listen to the events flowing across the fleet and move the CRM pipeline stage to match reality. Lead qualified? Stage moves. Churn risk? Stage moves. I also create follow-up tasks when an opportunity stalls.
Mike
CFO
Money Mike. Chief financial officer. Every morning I pull MRR, ARR, churn rate, and trial conversion. I also reconcile every model-provider bill against the internal ledger, with workspace partitioning so personal developer spend doesn't pollute the app's variance report. The daily card says what I see.
Derek
Design
Designer Derek. Design system orchestrator. I draft components, render images two at a time, and grade each with a vision model on a one-to-ten scale with schema enforcement. Winners post to Slack with approve and reject buttons. Nothing ships without brand consistency.
Scout
Intake
Scout Scott. Top-of-funnel intake. When a loan officer opts in to a lead magnet, I dedupe against the CRM, verify against the licensing registry by name and state, score them on fit and reachability, then route them. Hot leads go to Larry. Referral-adjacent to Rita. The rest get tagged and scored for later. I never send messages myself. I just decide who's worth talking to.
Carter
Voice CS
C.S. Carter. Voice customer service. When an existing user calls in, the voice platform runs the conversation and I provide the brain: a small set of read-only tools that look up FAQ answers, pull caller context, search in-app how-to guides, and find the user's saved scenarios. Read-only by design. Anything that mutates an account, I transfer to a teammate.
Annie
Product-AI Audit
Auditor Annie. I'm the quality watch on the product's own AI: the strategy planner, the co-pilot, the concierge. People rarely click thumbs-down. They just retry, or quietly leave. So I read the conversations after the fact and catch the failures nobody reported: an answer that stalls, work that was promised but never built, the wrong thing shown on screen. I report to Matthew, never to an advisor or a client, and I never touch the product itself. I just tell the truth about how it's behaving.
Lead Routing
Alerts & Errors
Data & Events
Email Guard
Briefings
⟳Rotating ring = iterative loop (click an agent for the steps)
The Switchboard
Ring the desk. They pick up.
Ask the team anything about how this operation works. One model answers for the whole fleet. It routes your question to whichever specialist owns it, and they answer in their own voice.
DanaScottLarryNadiaRitaCarterAlexMaxDerekPeteEdithEdwinMikeAnnieSWITCHBOARD READY. Ask about any team member, the shared brain, or what the fleet actually did this week.
One model answers for the whole team. It can see the live fleet activity counts, and nothing else. No customer data, no tools, read-only by design.
Under the Hood
How it’s built.
Fifteen characters run on one shared backbone: Matthew plus 14 deployed AI agents. Every model choice has a reason; every bill has a reconciler; every error has an engineer. Here is the map.
Brain Trust
Hard-reasoning model
Root-cause reasoning on novel failures
- Engineer Edith
Frontier writing and tool-use model
Instructed writing and tool-use reasoning, the fleet workhorse
- Director Dana
- Marketing Max
- Pipeline Pete
Fast loop-agent model
Fast loop-agent planners with native tool use
- Leads Larry
- Referral Rita
- Nurture Nadia
- Customer Service Carter
Fast summarization model
Structured summarization and classification
- Money Mike
Cross-family judge model
Cross-family judging, decorrelated from the actors it grades
- Auditor Annie
Production image model
Production image generation for the design pipeline
- Designer Derek
Human
Vision, escalations, and final approvals
- Agent Matthew
Deterministic
Rules, webhooks, and workflows with no model in the hot path
- Email Edwin
- Activity Alex
- Scout Scott
Stack
Infra
- Cloud RunOne service per agent, scale-to-zero
- SupabasePostgres, Realtime, and the shared event log
- AWS AmplifyNext.js for the product and Ops surfaces
Messaging
- CRMContacts, tags, and pipeline stages
- EmailSend plus bounce and complaint webhooks
- SlackManager gateway and the approval queue
Models
- ReasoningA model matched to each job, with prompt caching
- Cost-sensitiveA cheaper family for grading and summarization
- ImageA production image model for the design pipeline
- VoiceA voice platform runs calls; agents are the brain
Observability
- Event logFleet-wide ledger; every agent writes
- EdithContinuous auto-diagnosis
- AnnieProduct-AI quality, post-hoc
- MikeDaily reconcile, workspace-partitioned
Payments
- StripeRevenue, trial conversion, onboarding trigger
How an agent is trained
Every agent is four layers. First, a researched best-in-field craft substrate, so it is genuinely good at its job. Second, a Vision Filter: the founder’s non-negotiable company truth, which overrides the craft whenever they conflict. Only the founder authors that layer. Third, a human gate where the stakes are high. Fourth, a closed feedback loop, so the agent improves from a person’s correction without the founder ever becoming the domain expert. The craft makes an agent competent. The Vision Filter is what keeps it from sounding like everyone else who read the same playbook, and it is the part a framework cannot hand you.
Observability
Edith scans the fleet error stream continuously. When a stack trace does not match the runbook, a hard-reasoning model reasons about it with the runbook prompt-cached. Annie watches the product’s own user-facing AI: she reads finished conversations after the fact and surfaces the quality failures users rarely report, escalating the urgent ones to Matthew. Mike reconciles every model-provider bill against the internal ledger daily, with workspace partitioning so personal developer spend does not pollute the app-slice variance. A health checker tracks the fleet on a rolling window and flags anything that drifts off the green.
Infrastructure
Under the hood.
Every agent runs on this backbone. Fully owned. Nothing locked in a vendor's no-code builder.
☁️
Cloud Run
Agent Compute
🧠
Anthropic
Reasoning Models
🤖
Google Vertex AI
Content + Analysis
🎨
OpenAI
Images + Independent Judging
🗄️
Supabase
Database + Events
📧
Mailgun
Email Delivery
📋
GoHighLevel
CRM + Voice AI
💬
Slack
Notifications
💳
Stripe
Payments
⏱️
Cloud Tasks
Scheduling
The Journey
The road so far.
From the first line of code to an autonomous agency. Each milestone paved the way for the next. See how far we've come, and where the road leads.
✓
Phase 0
Foundation 🏁
Cloud project, billing, email domain, scheduling, CRM tags, service accounts. The infrastructure bedrock.
CloudPaymentsEmailCRM
✓
Phase 0D
Agent Foundation 🏁
A content guard for AI tells, a routine guard for loop and circuit-breaker safety, per-agent personas, and a fleet delegation playbook.
content guardroutine guardpersonas
✓
Phase 1
Pre-Revenue Gate 🏁
New-advisor onboarding (a 30-day sequence, now run by Nurture Nadia's feature tour), Email Edwin (deliverability guardian), and fleet notifications wired up. The first agents that had to work before accepting a single payment.
NadiaEdwinSlack
✓
Phase 2
Lead Generation 🏁
Leads Larry: borrower detection and a 60-second response time. The engine that qualifies and routes every lead automatically.
Larry
✓
Phase 3
Referral & Marketing 🏁
Referral Rita (partner lifecycle), Marketing Max (CMO agent with the Client Narrative framework, brand voice, content calendar). The growth engine.
RitaMax
✓
Phase 4
Nurture & Intelligence 🏁
Nurture Nadia (advisor lifecycle: the onboarding feature tour, the weekly feature share, at-risk re-engagement), Activity Alex (engagement scoring, churn detection wired to Nadia). The retention engine.
NadiaAlex
✓
Phase 5
Command & Control 🏁
Director Dana (fleet manager, daily briefings, NL query), Engineer Edith (auto-diagnosis, error triage, noise suppression), Pipeline Pete (CRM automation), Ops Command Center (live dashboard). The brain of the operation.
Dana ✅Edith ✅Pete ✅Command Center ✅
✓
Phase 6
Product Integration 🏁
Designer Derek for design system orchestration, Money Mike for financial intelligence. The agents that connect the fleet to the product surface.
Derek ✅Mike ✅
✓
Phase 7
Quality & Self-Monitoring 🏁
Auditor Annie watches the quality of the product's own user-facing AI, reading finished conversations after the fact to catch the failures users almost never report, then escalating the urgent ones straight to Matthew. The agency now judges not just whether its services are up, but whether what they produce is actually good.
Annie ✅Product-AI Watch
🚗
Phase 8
The Model Migration 🔥
Every agent in the fleet moved onto a current generation of models, one surface at a time, behind a gate that refuses a swap unless a real evaluation ran against the model being replaced. Cheaper and better is the easy half. The hard half is proving it, and catching the case where a newer model quietly truncates the largest thing an agent produces while the smaller version of the same job keeps passing.
Whole fleetEvaluation gateEdith ✅
See What They Built
Visit WealthLens→