AI Across the Entire Platform

AI is everywhere in Agend

Not bolted on. Not a chatbot in the corner. AI is woven through every one of our 20 apps — generating courses, drafting emails, matching capabilities, scoring engagement, and watching for trouble — so the people running associations like WDAA can stop juggling tools and start leading their members.

StudioPort 3064 Course CreatorPort 3073 LMSPort 3062 LoopPort 3063 DashboardPort 3060 ScopeSyncPort 3065 Atlas / QMSPort 3066 CMSPort 3067 CRMPort 3069 DirectoryPort 3061 EventsPort 3070 JobsPort 3071 Dev ToolPort 3068 ImagendVision QMS ToolsAtlas InsightsCross-app AI CORE

15 apps. One AI core. Each app uses the layer differently — same audited, tenant-scoped pipeline underneath.

Architecture

Not a feature. A pipeline.

One stack from app to model. One contract for every association. One audit trail for every query.

Layer 1 Apps (15+)
Studio Course Creator LMS Loop Dashboard ScopeSync Atlas/QMS CMS CRM Directory Events Jobs +4
Layer 2 Agent specialists (6)
Aggie Scout Beacon Atlas Lumen Sentinel
Layer 3 MCP tools (41)
Audit log RLS-scoped queries Tool registry Opt-out guards Token meter
Layer 4 Multi-provider model registry
Claude Gemini OpenAI OpenRouter Ollama (local)

One agent team, every association.

Tenant-aware via Postgres Row-Level Security — not per-customer forks, not per-customer prompts to maintain. Aggie answering Mel at WDAA never sees the data sitting in another association’s row.

Models swap with an env var.

Claude → Gemini → OpenRouter → local Ollama, all behind one interface. Pricing changes? Latency spikes? New frontier model lands? We flip a flag — the apps don’t care.

Every query audited.

The MCP tool layer logs who asked what, when, on whose tenant, with which model. Compliance officers love it. Pavan loves it more — she can show the APVB exactly what AI touched the CPD record and when.

“If we change a model tomorrow, no association even notices.”

16 capabilities

Every way Agend uses AI

Sixteen distinct AI capabilities, shipped or in flight, across the platform.

Studio · Port 3064 Shipped

AI Image Generation

Generate logos, banners, and social cards in WDAA’s exact brand — colours, type, voice — without firing up Photoshop.

Why: comms teams own the schedule, not the design backlog.

Imagen 3.0
At WDAA: Comms team reclaims 6 hrs/month.
Studio · Port 3064 Shipped

Auto Image Tagging & Library Org

A vision model captions and folders every uploaded asset so a five-year archive is searchable in seconds, not weeks.

Why: members’ faces and conference moments are findable, not lost.

Gemini 2.0 Flash
At WDAA: 5,000 conference photos catalogued in an afternoon.
Course Creator · Port 3073 In flight

AI Course Drafting

From PDF to lesson outlines, quizzes, and learning objectives — aligned to whatever framework your accreditation body insists on.

Why: the bottleneck is editorial review, not authoring from scratch.

Claude
At WDAA: Pavan ships an APVB-aligned course in 2 weeks, not 3 months.
Web Dashboard · Port 3060 Shipped

Six Specialist Agents

Aggie, Scout, Beacon, Atlas, Lumen, Sentinel. They live inside Loop channels and answer in your team’s tone.

Why: one conversation, six specialists, no tab-switching.

Claude Opus 4.7 + Haiku 4.5
At WDAA: Mel asks “who’s at risk?” once, gets the cross-system answer.
Web · Port 3060 Shipped

Membership Analyst (Lumen)

Renewal-risk plus engagement scoring across CRM, Events, LMS, and Loop — one view of every member’s temperature.

Why: silent members are the ones who churn first.

Claude Opus 4.7
At WDAA: Louise spotted 12 silent vets before they lapsed.
Web · Port 3060 Shipped

Revenue Strategist (Scout)

Sponsorship tier analysis, pricing scenarios, corporate-account targeting — with the deal history baked into every recommendation.

Why: the right call on Tuesday saves the renewal on Friday.

Claude Opus 4.7
At WDAA: $42K Western District renewal saved with a Tuesday call.
Web · Port 3060 Shipped

Compliance Sentinel

CPD posture, Code-of-Conduct watch, escalations — before the regulator even asks.

Why: audits are painful only when they’re a surprise.

Claude Opus 4.7
At WDAA: zero APVB-audit surprises this year.
ScopeSync · Port 3065 Shipped

RFP Requirement Extraction

Workshop transcripts and tender PDFs become structured requirements with confidence scores — so reviewers spend time on the 70%, not the 100%.

Why: the proposal deadline doesn’t care that you only got the brief on Friday.

GPT-4 (JSON mode)
At WDAA: corporate sponsor brief processed in 48 hrs.
ScopeSync · Port 3065 Shipped

Capability Matching

Each requirement is matched against the platform capability registry with reasoning — gaps surface before they become commitments.

Why: a clear “we already do that” vs “we’d need to build that” in minutes.

OpenAI
At WDAA: proposal mapping went from a week to an afternoon.
ScopeSync · Port 3065 Shipped

User Story + UAT Auto-Drafts

Approved requirements turn into stories with plain-language UAT steps a board member can read — no PM round-trip.

Why: the people approving the spec must understand the spec.

GPT-4o
At WDAA: board-approved spec ready for build, no PM round-trips.
Atlas / QMS · Port 3066 Shipped

Workflow Auto-Classification

New workflows are tagged Quality / Compliance / Operations on the fly — the dropdown nobody filled in is now answered for them.

Why: dropdowns are the silent killers of every QMS adoption.

Rule-based NLP
At WDAA: no admin field to fill, the system just knows.
Atlas / QMS · Port 3066 Shipped

Anomaly Detection

Catches duplicate workflows and over-allocated owners before they ship as confused processes to members.

Why: two CPD-renewal flows is one too many.

Jaccard similarity
At WDAA: spotted three duplicate CPD-renewal flows before they confused members.
Atlas / QMS · Port 3066 In flight

Workflow AI Assistant

Embedded chat with full account context plus tool calling — new staff don’t need a 40-page induction PDF.

Why: onboarding is the most expensive workflow you have.

Claude 3.5 Sonnet
At WDAA: new admin productive on day one.
CRM · Port 3069 Shipped

AI Email Drafting

Tone-controlled follow-ups grounded in the deal history — warm, professional, urgent, choose your register.

Why: the blank-page problem is half the reason follow-ups don’t happen.

Multi-provider
At WDAA: Dev’s renewal-reminder email open rate went 24% → 38%.
CRM · Port 3069 Shipped

Deal Activity Summarisation

Bullet trajectory of every deal — what happened, what’s next, where the energy is — in a glance.

Why: the cost of a bad handover is a lost deal.

Multi-provider
At WDAA: handovers between Mel and the board take minutes, not hours.
Directory · Port 3061 Shipped

Hybrid Semantic + Keyword Search

Find members by intent, not exact phrase, with graceful fallback to keyword when the embedding model is offline.

Why: nobody types the exact phrase that’s in the database.

Ollama embeddings + pgvector
At WDAA: “rural mixed-practice vets” returns the right 38 members instantly.
A day in the life

WDAA, 8:42am Tuesday.

Western District Veterinary Association. 230 members. Three-person admin team. Here’s how AI changes their morning.

  1. 8:42am

    Mel opens her laptop.

    Aggie has already drafted her morning brief: 3 deals progressing, 1 at risk (a $42K renewal — Western District corporate sponsor — 14 days quiet), platform green, 12 at-risk members surfaced, Pavan’s CPD post ready to review.

    Mel didn’t ask. The brief was waiting.

  2. 9:15am

    Louise checks the membership dashboard.

    Lumen has flagged the 12 vets on her at-risk list. Each has a “what changed” note: career shift, surgery downtime, moved practice. Three are tagged “high-value, intervene this week.”

    Louise has 30 minutes of focused work, not 3 hours of spreadsheet hunting.

  3. 10:30am

    Dev drafts the November conference comms.

    Beacon writes a personalised invite for each member tier: senior vets get the keynote angle, nurses get the CPD-credit angle, students get the bursary angle. Sentinel reviews for compliance language before the send.

    3 hours of writing becomes 20 minutes of review.

  4. 1:45pm

    Pavan ingests the new APVB practice guidelines.

    A 50-page PDF. ScopeSync extracts each new competency, maps to existing LMS modules, flags gaps. Pavan reviews the gaps, accepts the rest.

    No more reading the PDF twice and second-guessing.

  5. 4:10pm

    A sponsor RFP arrives.

    ScopeSync runs requirement extraction and capability matching while Mel makes coffee. Draft proposal lands in her inbox before she’s back.

    48-hour turnaround, no Friday-night scrambling.

Architectural reassurance

Four pillars under everything.

The hard part isn’t the prompts. It’s the plumbing that makes them safe to ship to a hundred associations at once.

Tenant-isolated by construction

RLS at Postgres, not in agent code. WDAA’s data never touches another association’s session — even by mistake.

Provider-agnostic

Claude today, Gemini tomorrow, local Ollama for embeddings. Swappable behind one interface — pricing and latency stay our problem, not yours.

Audited end-to-end

Every tool call logged with account, user, agent, inputs. Compliance officers love this — the regulator-friendly answer is always one query away.

Cost-aware

Token usage tracked per agent per tenant. Budget alerts before surprises — nobody wakes up to a five-figure inference bill.

By the numbers

The shape of the platform today.

0
Apps

AI woven through every one

0
Specialist agents

One team, shared across every association

0
MCP tools

Each tenant-scoped, audited, opt-out aware

0
Model providers

Claude, Gemini, OpenAI, OpenRouter, Ollama — flip with an env var

Want this for your association?

AI woven through every part of the operation. One platform. One conversation. Your data, your tenant, your team.