← → arrow keys to navigate
Module 05 — OOH Intelligence

MX Campaign
Management Platform

A proper, modern operational media system — built from the ground up with AI at its core — that takes what MX does today and supercharges it.

Miroma Unified Media Platform March 2026 ~150h Development Effort
The Problem

MX runs on Excel, email chains,
and institutional knowledge

80%
of effort consumed before a single site is booked
15 hrs/wk
spent on invoice query resolution — backlog stretches 2+ years
64%
of team capacity consumed by Big 3 media owner work alone
30-40
simultaneous campaigns across ~100 media owners, every one a fresh start
£9M
Live Nation account threatened by competitors with modern platforms
0
systems know what you paid for the same site last quarter
The Vision

Media software imbued with AI

Not AI with a media interface. A system built for how media buying should work.

🤖 AI proposes, humans approve — nothing goes to a client or media owner without a human clicking yes
📊 Excel becomes an export format — one-click, pixel-perfect, agency-branded. Clients get schematics exactly as they expect them.
🎯 Workspace for decisions, not data entry — planners review AI-generated drafts, approve matches, and make strategic calls
📈 Commercial intelligence in real time — ratchet tracking, pricing anomalies, cross-campaign optimisation surfaced proactively
How It Works

One campaign, end to end

📧
AI Brief arrives — parsed automatically
→
📋
AI Draft campaign plan generated
→
👤
Human Planner reviews & adjusts
→
📄
AI PO matched & reconciled
→
✅
Human Approve match — one click
→
📤
AI Booking sent — API or drafted email
→
💰
AI Spend tracked against ratchets
"A brief arrives from TMP by email. The system parses it, generates a draft campaign plan with pricing pre-populated and availability checked. The planner opens a complete draft — not a blank form. She reviews, adjusts, approves. Meanwhile, the system tracks spend against Global's ratchet deal in real time."
Pricing Intelligence

Four pricing scenarios,
one intelligent system

Like the price indicators on AutoTrader — instantly see if you're paying a good price, a fair price, or overpaying.

Tier 1 — Known Grid

Auto-populated

UK media owners with published rate cards. System populates pricing, human verifies. Confidence score on every line.

Tier 2 — Known Vendor, Unknown Price

Negotiation workflow

System flags for negotiation, manages the outreach, captures the agreed price. Builds history for next time.

Tier 3 — Completely Unknown

Human process, system learns

Full manual process — but the system captures the result. Next time this format appears, it's Tier 1 or 2.

Tier 4 — US Inventory

Multi-vendor ranking

Media owners provide available inventory. System recommends which 10 units to pick from 1,000 options based on brief criteria.

Delivery Roadmap

Sequential drops,
standalone value at every stage

DropWhat the Team GetsDev Hours
Drop 1 Campaign workspace, brief parsing, PO matching with split-pane review, XLSX export. Deployed on 25% of accounts running parallel with Excel. ~35h
Drop 2 Pricing intelligence — contextual panels, historic comparison, anomaly detection. Value tracking — real-time ratchet dashboard. Buying tracker eliminated. ~21h
Drop 3 Availability integration (Pearl, media owner APIs). Booking automation. Spec caching. Dependent on API access. ~20h
Drop 4 US operations — RFP agent, contracts, three-bucket commercial model. Reporting & reconciliation. ~18h
Total Full platform — modular, each drop is standalone useful. Stop after any drop. ~150h

Calendar scheduling (how hours map to weeks) is a project management exercise. We provide the engineering hours — Vix builds the calendar.

Drop 1 — Foundation + Document Intelligence

What you get first

🖥️ Campaign workspace — card view + advanced grid for power users
📧 Brief intake — email/PDF brief parsed into draft campaign plan
📄 PO matching — split-pane review (original doc left, AI extraction right)
📊 XLSX export — pixel-perfect, agency-branded, one click
🔄 Booking confirmation reconciliation — auto-matched
🔐 SSO authentication — same login as everything else
30-60 min
2-5 min
PO matching time
1-3 hours
5-10 min
Brief to draft plan
25% pilot
1-2 agencies running parallel with Excel
The Impact

What changes

30-60 min
2-5 min
PO matching
1-3 hours
5-10 min
Brief → draft plan
15 hrs/week
3-5 hrs
Finance queries
64% of capacity
10-15%
Big 3 effort
Year-end estimate
Real-time
Ratchet tracking
Manual every time
Instant recall
Pricing intelligence

17 manual steps across the campaign lifecycle: 5 eliminated, 10 transformed by AI, 2 fully automated.

De-risking the Rollout

Built for adoption,
not just functionality

All technological change is a people problem. We've designed for that.

✓
25% pilot — deploy on 1-2 agency accounts first, running parallel with Excel
✓
Human approval at every checkpoint — nothing goes to a client or media owner without human sign-off
✓
Confidence scores on every AI output — above 95% pre-approved, below 80% flagged for review, below 60% manual handling
✓
Excel export always one click away — clients never see the difference
✓
Modular drops — each drop delivers standalone value. Stop after any one.
✓
Full audit trail — every state transition checkpointed, every change logged, every AI decision traceable
The Cost Advantage

Competitors charge 3% of media spend
for what we're building in-house

£1.5M/yr
Third-party platform @ 3% of £50M media
VS
~150h
In-house build — own the IP, deploy AI no vendor offers
"Bionic charges $199/user/month. Mediaocean is enterprise legacy pricing. CrowdSauce showed Live Nation a portal that threatened a £9M account. We build it in-house, own the IP, and deploy intelligence no off-the-shelf tool offers."
Parallel Workstreams

What we need to start

These run in parallel with development — they're not blockers.

Pricing grid data from Zoe — 2026 grids digitised for the pricing intelligence engine
Historic schematics — June 2025 onwards. More data = better AI recommendations
Pearl API access — start the conversation now. Browser automation as fallback
Media owner APIs — Big 3 procurement. Our spend gives leverage
Planner workflow observation — beyond transcripts, we need to shadow the daily work
Business Central vs Media Central — architecture is agnostic; clarity helps sequence
US process call with Aaron — understand US workflow directly

Development starts immediately. The code doesn't wait for the questions to be answered — we build in parallel.

Frequently Asked Questions

FAQ

Why not build the simple thing first, then add AI?
Phase 1 of the AI-first approach IS the simple thing — campaign workspace, PO matching, XLSX export. The difference is the foundation supports the next drop instead of requiring a rebuild. Two-stage means double the change management.
What if the AI gets it wrong?
Confidence scores on every output. Above 95% — pre-approved with notification. Below 80% — flagged for human review. Nothing goes to a client or media owner without a human approving it. Every decision is audited and traceable.
Have the planners validated this approach?
Not yet — that's why Drop 1 deploys on 25% of accounts first, running parallel with Excel. Real campaigns, real feedback, real validation before we expand. The planners try both and tell us what works.
This seems like too much scope?
~150 hours total, delivered in sequential drops. Each drop is standalone useful. Stop after any one — you still have a working system. Drop 1 alone replaces manual PO matching and brief processing.
How does this work with Nexus / Business Central / Media Central?
The architecture is system-agnostic. Our export engine produces the correct data format; the upload target is pluggable. Whether it's Nexus, Business Central, or Chase's Media Central wrapper — it's a configuration change, not a rebuild.
What about the US?
US is phased as Drop 4 — but if commercial priorities demand it earlier, the RFP agent and US data model can move up without disrupting the UK track. The architecture supports this flexibility.
Will the team actually use this?
Parallel running. Gradual rollout. Excel always one click away. On day one every AI output is human-reviewed. Trust builds over time. We've designed for adoption, not just functionality.
Competitors charge 3% of media spend — are we sure about in-house?
On £50M+ of media, that's £1.5M/year in platform costs we never pay. We build it for ~150 hours of development, own the IP, and deploy AI intelligence no off-the-shelf tool offers.
Summary

Media software
imbued with AI

A proper, modern operational system that handles the heavy lifting — so planners can focus on strategy, relationships, and commercial decisions that grow the business.

~150h
Development effort
4 drops
Standalone value at each stage
25% pilot
Parallel with Excel — prove it first
Full strategy document available on request Prototype demo ready