A roadmap for CPH Chemicals. Three layers. One hundred and twenty days to the first measurable win. Microsoft Dynamics stays — it gets surrounded by judgment.
The mistake is treating them as competitors. The opportunity is using each for what it is best at — and drawing a clean line between them. Below is that line.
Microsoft Dynamics & the legacy stack
“The customer’s credit limit is $50k because that's what we typed in last year.”
The AI-native layer that surrounds it
“The customer's risk profile rose this week on local-currency volatility — suggest a $35k limit.”
Dynamics forces the world into a rigid spreadsheet.
The AI-native layer thrives in the chaos of global chemical trading.
CPH needs both.
Every employee uses Claude / ChatGPT / Gemini as a personal copilot. Zero IT. Tomorrow.
Open Layer 1 →Off-the-shelf AI tools surround Dynamics, one process at a time. This year.
Open Layer 2 →Claude Code & agentic IDEs let anyone spin up the app they need. The horizon.
Open Layer 3 →The deepest reason to go AI-native is not productivity. It is the shape of the curve when CPH's order book doubles.
Manual data entry scales linearly. Phone calls scale linearly. Open-book reporting and reconciliations scale linearly. Every additional deal is another email someone has to translate, another PDF someone has to retype, another margin someone has to recalculate by hand.
Growth becomes a cost centre. The bigger CPH gets, the slower it moves.
The unstructured mess of the world — emails, delays, risk shifts, foreign documents — is absorbed by the AI layer. Humans focus only on the high-stakes, high-judgment moments. The cost of one more deal approaches the cost of a few more API calls.
Growth becomes a margin. The bigger CPH gets, the faster it compounds.
Twenty dollars per person, per month. Five minutes to set up. No integration with Dynamics, no risk to records of truth. The point of Layer 1 is not productivity — it is literacy. You cannot fund Layer 2 with a workforce that has never used the tools.
Paste a Polish or Mandarin RFQ email. Get a clean English deal summary with quantities, Incoterms, and red flags.
“Summarise this RFQ. Pull qty, grade, Incoterms, payment terms. Flag anything ambiguous.”
Drop your last 5 caprolactam deals → ask for the margin pattern and a recommended quote for this buyer.
“Given these last 5 deals and today's spot, what should I quote, and at what margin?”
Drop a 30-page Bill of Lading PDF. Extract HS codes, weights, and any discrepancies versus the contract.
“Compare this BoL to the attached contract. List every mismatch.”
Draft a customer email explaining a 4-day delay at Rotterdam — professional tone, in their language.
“Email to customer X: 4-day delay, professional, in German, offer a make-good.”
Summarise the last six months of email with this principal, list open commitments, suggest three talking points.
“Brief me on this account before our 3 pm call. Open items + risks.”
Turn messy meeting notes into a structured account update, owners, and next steps.
“Convert these notes into: decisions, action items (with owner + date), open questions.”
Compare a Safety Data Sheet against current REACH requirements. Flag what is missing, cite the article.
“What does this SDS lack to be REACH-2024 compliant?”
Drop a commission statement & the underlying contract. Get a line-by-line variance.
“Reconcile this commission statement against the contract. Show every variance with reason.”
Rewrite a job spec for a Singapore-based trader, in CPH voice, optimised for LinkedIn.
“Rewrite for LinkedIn. Keep our tone. Lead with mission, not benefits.”
Translate, summarise, and propose a reply to the eight-email chain you just got copied on.
“Summarise this thread. Who said what. What does the next reply need to say?”
Funding ask : ~ €20 / employee / month. Risk : zero — nothing writes back to Dynamics. Reward : a workforce that knows what these tools can and cannot do, before we wire them into anything that matters.
We go to theresanaiforthat.com, we pick the best-of-breed tools for each core process, and we glue them on top of Dynamics. Dynamics keeps the truth. The AI tools do the translation, judgment, and drafting. Below — three pillars, three honest comparisons of old way vs. AI-native way.
Twenty-plus warehouses, hazmat documentation, customs across continents. The job is no longer to know where the truck is. The job is to know, before the customer does, that it will be late.
Autonomous inventory reconciliation across the 20+ warehouse network — no more month-end stock-count drift.
Reads carrier ETAs, weather, port congestion. Decides which container moves where and when.
Computer vision over warehouse cameras — hazmat handling, PPE compliance, near-miss detection.
Sourced via theresanaiforthat.com — the working catalogue of every AI tool on the market.
The bridge between buyer and seller is built on paperwork. CPH's edge — for thirty-five years — has been navigating that paperwork. The AI layer does not replace that edge. It scales it.
Reads any invoice in any language. Pushes structured data into Dynamics. Days of typing → seconds.
Re-prices customer risk continuously using public + private signals — currency, sector, behaviour.
Reads news, prints, and forums for early signals on caprolactam, polyolefins, and solvent prices.
Sourced via theresanaiforthat.com.
The agency model lives on knowing who is buying what, where. Today, CPH's sales team learns this when an inquiry arrives. Tomorrow, CPH knows it before the inquiry is sent.
Builds a 360° dossier on any company / sector — competitors, suppliers, growth signals — in minutes.
Personalised, multilingual outbound at scale. Books meetings on the calendar without a human SDR.
“Open-book” reporting in plain English. Sales leadership talks to the data; charts answer back.
Sourced via theresanaiforthat.com.
At Layer 3, every CPH employee has Claude Code (or Cursor, or Replit Agent, or Lovable) on their machine and the literacy to use it. They stop asking, “Is there a tool for this?” and start asking, “What is the app I need for the next thirty minutes?” Then they build it, use it, and discard it. This is the moat that competitors of any size cannot out-spend.
A logistics coordinator builds a one-off live dashboard for a single big shipment week — pulls ETAs from four carriers, the customs clearance status, and a checklist of pending documents. Throws it away after the containers land.
build time · ~ 40 minA senior trader builds a private margin tool that reflects her hedging style — not the corporate one. She describes the rules in English; Claude Code writes the spreadsheet logic. She uses it for a quarter, retires it when the strategy changes.
build time · ~ 1 afternoonThe QHSE lead spins up an agent that watches Indonesian, Polish, and EU chemical regulators each morning, cross-references active CPH shipments, and emails her only when something matters.
build time · ~ 2 hoursA finance analyst writes a one-time agent that reads draft board packs, flags inconsistencies in numbers across pages, and verifies every chart reconciles to the source data. Used once. Discarded.
build time · ~ 90 minLayer 1 is “AI helps me think.” Layer 2 is “AI helps the company run.” Layer 3 is “AI lets me build the exact tool I need.” That is the structural shift.
Score each statement from 0 (not at all) to 3 (we live here already). At the bottom you get a layer score and the recommended next move.
“Dynamics tries to force the world into a rigid spreadsheet.
An AI-native stack thrives in the chaos of global chemical trading.”