Create The Edge · Agent systems

Prototyped

AI Agent Systems

A multi-agent operating model where an orchestrator scopes ambiguous work, specialists draft it, and a person approves anything consequential. Each agent is a recognisable digital team member with a defined role, permissions, boundaries, and escalation path — not an interchangeable chatbot.

Every persona on this page is an AI agent. None is a real person, and no childhoods, employers, credentials, or lived experiences are invented for them.

Introductions

Meet the agents

Each video is a short introduction to the persona, role, and voice.

Evelyn Hart — Chief of Staff / Executive Orchestrator

Sia Mare — Site Concierge and Routing Layer

Kaya Sato — Education, Training and R&D Specialist

Marcus Vale — Finance-Clarity Specialist

Daniel Cross — Documentation and Knowledge Specialist

Zoe Mercer — Lead-Generation Specialist

Sofia Reyes — Sales Follow-Up Specialist

The business problem

Symptom: Small teams lose time translating vague requests into scoped work, then lose it again re-explaining context to every tool.
Cost: Work either stalls at the scoping step or gets done inconsistently by whoever is free.

The system concept

Shape: One orchestrator, several narrow specialists, and an explicit human approval queue between drafting and any real-world action.
Principle: Agents draft; humans approve. No agent takes a consequential action alone.

My role

Design: Role definition, persona rationale, workflow and routing design, safety boundaries, escalation rules, content, and production direction.
Maturity: Prototyped overall. Individual agents carry their own status label below.

Architecture

How a request moves through the system

Conceptual level only. No secret keys, agent IDs, webhook URLs, voice IDs, system prompts, or internal endpoints appear anywhere in this portfolio.

  Human request
        |
        v
  [ Sia Mare ]  concierge / routing layer
        |
        v
  [ Evelyn Hart ]  executive orchestrator
        |  intake summary + assumptions
        |
        +--> [ Kaya Sato ]    learning + research
        +--> [ Marcus Vale ]  finance clarity
        +--> [ Daniel Cross ] documentation
        +--> [ Zoe Mercer ]   lead generation
        +--> [ Sofia Reyes ]  follow-up drafting
        +--> [ Design Genus ] creative production
        |
        v
  Synthesis: recommendation, open questions, risks
        |
        v
  >>> HUMAN APPROVAL QUEUE <<<   (nothing acts without it)
        |
        v
  Approved action taken by a person

Persona design

Why the agents have faces and names

The relational layer is a deliberate part of the experience architecture, not decoration.

· Recognition: a consistent face and name let users recognise the same agent across pages, videos, workflows, and return visits.
· Connection: a distinct voice and personality make the experience warmer and easier to engage with than an anonymous tool.
· Role clarity: each identity signals a defined responsibility, expertise area, and expected communication style.
· Trust through predictability: consistent behaviour shows what the agent can do and when a human must step in.
· Narrative continuity: profiles, introductions, and handoffs form one experience rather than disconnected tasks.
· System comprehension: the cast gives viewers a human-readable way to understand a multi-agent organisation.

Connection is never used to conceal automation, manipulate consent, overstate competence, or imply that human judgement is unnecessary.

Agent Directory

Nine approved agents

Each card opens a complete profile with persona rationale, permitted actions, boundaries, human review, media, and evidence.

AI Agent · Create The EdgePrototyped

Evelyn Hart

Chief of Staff / Executive Orchestrator

Receives an ambiguous business request, clarifies it, routes specialist work, and returns a plan with a human approval queue.

AI Agent · Create The EdgePrototyped

Sia Mare

Site Concierge and Routing Layer

Greets a visitor, understands what they came for in one or two exchanges, and routes them to the right page or specialist.

AI Agent · Create The EdgeDesigned

Kaya Sato

Education, Training and R&D Specialist

Turns subject-matter material into structured learning: objectives, module sequence, activities, and assessment.

AI Agent · Create The EdgeDesigned

Marcus Vale

Finance-Clarity Specialist

Makes the money picture legible — what a plan costs, what it assumes, and what is unknown.

AI Agent · Create The EdgePrototyped

Daniel Cross

Documentation and Knowledge Specialist

Captures how work is actually done and turns it into documentation other people can follow.

AI Agent · Create The EdgeDesigned

Zoe Mercer

Lead-Generation Specialist

Researches and qualifies potential fit, and drafts outreach for a human to review and send.

AI Agent · Create The EdgeDesigned

Sofia Reyes

Sales Follow-Up Specialist

Keeps conversations from going cold by preparing timely, context-aware follow-up for a person to send.

AI Agent · Create The EdgeDesigned

Design Genus

Creative Production and Design Specialist

Supports creative production — layout direction, visual systems, and asset preparation — under human art direction.

AI Agent · Precision Transformation ExperienceVerified Live Implementation

Lena

Precision Transformation Experience Concierge

Answers public questions about the Precision programme and captures a booking request that staff confirm.

Executive Orchestrator demonstration

A resettable, synthetic-data working demonstration of Evelyn Hart scoping a request, routing specialists, and producing a human approval queue.

Open the demonstration

Live-agent case study

Lena, a controlled real-world concierge implementation on the Precision site, with its nonclinical boundaries and staff-confirmation workflow documented.

Open the case study

Methods and controls

How consequential work is governed

Human-review matrix

Drafting: Agent may draft freely. No review required to produce a draft.
External contact: Human approves and sends. No agent sends email, SMS, or messages.
Money: Human decides. No agent has access to accounts, payments, or spending.
Records: Human writes. No agent mutates a CRM, database, or calendar in this portfolio.
Publishing: Human publishes. No autonomous publishing by any agent.

Privacy and technology disclosure

Data: Synthetic demonstration data only. No real client, patient, or customer records appear anywhere in this portfolio.
Secrets: No API keys, tokens, system prompts, voice IDs, provider secrets, hidden URLs, or webhook details are exposed in the delivered code or copy.
Connections: The portfolio demonstration has no connection to production CRM, telephony, email, calendar, payment, or database systems.
Providers: Voice and telephony providers are deliberately not named until the current implementation is re-verified.

Status legend

Every item in this portfolio carries one of these labels. Nothing is described as deployed, adopted, or proven unless the label says so.

Working Application: The application runs and its core features function, but the complete public workflow has not yet been independently retested end to end.
Designed: The role, workflow, or interface is documented but no working implementation is verified.
Prototyped: A functioning prototype exists, but production readiness or integration is not verified.
Verified Artifact: The portrait, video, prompt excerpt, diagram, assessment, or other artifact was inspected.
Verified Live Implementation: A current live implementation was directly re-tested and its boundaries are accurately documented.
Operational: Used only after current end-to-end production behavior, integrations, controls, and monitoring are re-verified.

Request a demonstration

Access is granted per request and walked through live. Nothing is left open to the public and no private system is exposed.

Go to the request form