July 8, 2026
01 — The Thesis

Two kinds of software now sit inside every modern company.
One records what is true. One reasons about what to do.

Eldan's words : “We're in the middle. And when you're in the middle, the best thing the market can do is try to get you out.” The escape is not a faster ERP. It's a non-deterministic leg — judgment, on top of the deterministic stack we already run — that keeps Micromedia two steps ahead.

Mode A · Co-pilot — you ask, AI helps. → Mode B · Auto-pilot — AI does, you supervise. Both modes belong to the same non-deterministic leg; the deck below shows when each one fits.

DETERMINISTIC

System of Record

Business Central & the legacy stack

Logic
Rules written by hand, months ago, via the PID queue.
Input
Structured fields — SKU, qty, tier, rebate, lithium tax, chemical tax.
Output
Same input → same output, every time.
Question
“What was posted? What is true right now?”
Strength
Audit. Posting. Money in, money out.
Failure
Silent — the unreconciled pulsings row hides until month-end.
To add a case
PID. IT queue. Six weeks — or six months.

“Razer has 18 tiers — by SKU, by customer, by promo. BI reconciles them in Excel every single week.”

NON-DETERMINISTIC

System of Judgment

The AI-native leg that keeps Micromedia indispensable

Logic
Patterns inferred from data + context.
Input
Tender PDFs · retailer pulsings · supplier feeds · buyer emails · marketplace prices.
Output
Probably right — with a confidence score and a citation.
Question
“What does this mean, and what should we do?”
Strength
Judgment. Language. The exceptions Business Central can't see.
Failure
Loud — tells you when it's unsure, with the reason.
To add a case
Give it an example. An afternoon.

“JBL just showed 8,000 of a ‘delisted’ SKU in their warehouse — pull them before any other distributor sees the listing move.”

Business Central forces the deal into a rigid posting.
The AI-native leg thrives in the chaos in the middle: 18-tier rebates, hidden warehouse stock, marketplace pricing, custom B2B portals.
Micromedia needs both.

Comparison : Deterministic vs. AI-Native at Micromedia

Business Challenge
Deterministic (Today)
AI-Native (The non-deterministic leg)
1. Forecasting Bol's JBL order
One analyst's brain + the legacy “VI” system handles JBL × Bol. Every other brand falls back to static safety stock.
Every brand × every retailer gets JBL-grade forecasting — continuous demand prediction, tied to live closeout signals.
2. Catching JBL's hidden warehouse closeouts
A trader checks Harman's portal every morning. Catches some; misses many. Drains a senior person daily.
Watch supplier feeds 24/7. Alert the team the same hour a “delisted” SKU reappears with 8,000 units.
3. Rebate, tier & margin reconciliation
BI hand-reconciles 18-tier rebates × lithium tax × chemical tax × customer rebates in Excel — every single week.
Accruals post continuously to Business Central. The analyst reviews only the exceptions the system flags.
4. Retailer pulsings — Eldan's “spaghetti”
BI checks every pulsings row by hand. Brand claims slip when timing or SKU codes don't line up.
Match pulsings to our shipments in seconds. Only the unmatched 5% reach BI's desk, each with a probable cause.
5. Marketplace pricing — bridging the islands
Leap-to-See prices on Bol Plaza / Amazon with AI. The rest of Micromedia sets prices by hand. The islands don't talk.
One pricing engine. Marketplace AND traditional channels share competitor moves, tier math, Buy-Box position.
6. Vodafone / KPN special-deal builds
One B2B Integrator owns Vodafone, KPN, Albert Heijn, Postcode Loterij. Hand-rolls the IT spec for every new deal.
AI reads the 200-page tender in an hour, drafts the integration spec from past wins. The integrator reviews and signs.
7. Brand content for retail PDPs
Content team rewrites and translates every PDP, in NL / FR / EN, per retailer template.
AI generates on-brand PDPs in the right language, ready to push. (Already in motion — external partner build.)
Why this matters

Two scaling curves.
Only one of them is affordable.

The deepest reason to go AI-native isn't productivity. It's the shape of the curve when Micromedia's order book doubles.

DETERMINISTIC SCALING

2× orders  →  ≈1.8× headcount

Manual entry, phone calls, reconciliations — all linear. Every extra deal is another email to translate, PDF to retype, margin to recompute by hand.

Growth becomes a cost centre. The bigger Micromedia gets, the slower it moves.

NON-DETERMINISTIC SCALING

2× orders  →  ↑ API spend

The unstructured mess — emails, delays, risk shifts, foreign documents — is absorbed by the AI layer. Humans focus on the high-stakes, high-judgment moments. One more deal ≈ a few more API calls.

Growth becomes a margin. The bigger Micromedia gets, the faster it compounds.

Stylised. The point is the divergence.
See the three layers  →