01intelligence
Applied AI
Models put to work on the real problem: what is worth building, what to ignore, and systems your team can actually run in production.
The age of arbitrage,
every process, tax, service
is under the eye of AI.
For years companies have bought AI and left the value sitting in their own data. Amplfr Labs exists to prove it can be applied where it actually pays - deliberately, and well.
Six years of it, and counting.
Start with a diagnostic →Scroll
The problem
The hard part was never the model.
The models are a commodity. Anyone can buy them. What is scarce is the call: what is worth automating, what to leave alone, and what a decision still costs three years after the pilot.
What we believe
The value of AI is application, not technology.
The models are solved. What gets built, for whom, and who can run it after - that is the whole difference.
A call that leaves with the adviser was never yours.
The measure of an engagement is whether your team can make the next one without us in the room.
The expensive mistakes happen before a model is chosen.
Most failed initiatives solved the wrong problem precisely. We start at the problem, not the tooling.
The money is already in your data.
Six years of finding it: the duplicated work, the overpayment, the arbitrage nobody had the time to go looking for.
How we work
01
Diagnose.
Before any technology enters the conversation we take apart the problem - what is actually worth solving, what is not, and what the real constraints are. Most of the value is decided here.
How we work
02
Decide.
We set the direction: what to build, what to buy, what to ignore, and what each path costs over years rather than quarters. Recommendations come with the reasoning shown.
How we work
03
Apply.
Where it serves you we build alongside your team rather than around it - deliberately, at a pace that holds up under real conditions rather than under a demo.
How we work
04
Transfer.
We leave the call with you, along with everything built to serve it. The engagement is finished when your team can make the next one without us.
02automation
Automation & Efficiency
The duplicated work, the manual reconciliation, the process nobody owns - mapped, automated, and measured against what it used to cost.
03arbitrage
Recovery & Arbitrage
Money already sitting in your data: overpayments, mispriced positions, and cross-border packaging fees paid in the wrong country. We find it and get it back.
04build
Products
Software built to last and shipped at a pace that holds up under real conditions, not a demo that looks right on stage.
Our capabilities
Systems we build, run, and keep with the teams we build them for.
A growing set of tools built out of the calls Amplfr Labs makes in the room - the reconciliation engine, the automation layer, the recovery rulebook - and kept running long after the first engagement closes.
Contact for demo →Built in production, not in a slide, and sharpened with every engagement.
Every part of it has run against real data, in a real business, with someone accountable for the result.
The firm
Amplfr Labs
A small team, AI-native, six years into the same work: automations, efficiencies, and arbitrage found in the data a business already holds. Duplicated processes, overpaid fees, positions priced against stale information - the things that never show up on a dashboard because nobody owns them.
We are not a research lab and not a reseller. We make the call in the room, build against it beside your team, and measure the result in what it saved.
Research & notes
Notes from the work.
01.
Applied, not advised
Why the value of a model shows up in what gets built around it, and why so many pilots never survive contact with a real process.
Case · Regulatory engine
The premise checked out. One load-bearing assumption in it did not.
A new EU packaging regulation creates a right to a refund where producer-responsibility fees are paid in one Member State and the goods are first made available in another. The right is real, and stated in the Commission’s own guidance.
We read that guidance end to end before building anything. The word reimburs appears in it once, in 144,601 characters. No procedure, no deadline, no form, no named counterparty, no evidence standard. The right exists; the plumbing does not.
An advisory engagement delivers that finding and invoices for it. Here it changed the build: the engine files through national routes that already move cash, and treats the new right as the argument that opens the conversation rather than the mechanism that closes it.
- Guidance read in full
- 144,601 chars
- Occurrences of “reimburs”
- 1
- Findings needing the new right
- 0 of the top 4
02.
The money in the data
Where inefficiency actually hides in a mid-sized business - and why it is almost never where the org chart says it should be.
Case · Recovery audit
The highest-leverage dataset is one almost every producer already files.
Producers declare packaging volumes to a compliance scheme in each country they sell into. Separately, and for unrelated reasons, they file Intrastat dispatch data: every intra-EU shipment, by destination Member State. Both numbers exist. Almost nobody reconciles one against the other.
Where goods went to a business in another Member State, the recipient often became the producer there - so the fee was paid at home as well, or paid in the wrong country altogether. The gap appears the moment the two datasets are laid side by side. No new law is needed to see it.
That is one of fifteen findings in the audit taxonomy. The rest are duller and just as expensive: commercial packaging declared at household tariffs, which are a multiple of the rate; packaging licensed once by the supplier and again by the buyer; spec weights that drifted from delivered weights years ago.
- Findings in the taxonomy
- 15
- Evidence usually already held
- Intrastat, MRNs
- Typical audit
- 25-40 hours
03.
Reimbursed - but by whom?
One sentence in the Commission’s packaging guidance creates a right to a refund, and no procedure yet exists to use it. A worked example of arbitrage in the rules.
Case · Arbitrage in the rules
A right with no procedure, and a scheme that has been paying it for years.
“If fees are paid in a Member State, and afterwards a distributor makes the packaging available for the first time on the territory of another Member State, the fees must be reimbursed.”Commission Notice C/2026/3084, OJ 10 June 2026
The guidance says that once, and nothing further about it: no implementing act, no form, no deadline, and no authority named as the one who pays.
The mechanism is not hypothetical, though. Italy’s CONAI has refunded on exactly this basis for years, intra-EU dispatches included, through an export-deduction route with a hard annual deadline in February. The right the regulation states without plumbing is one a national scheme has been plumbing all along.
So the work starts opposite to where the headline points: not with the novel argument, but with the route that already runs. The regulation is why the meeting happens; the national scheme is how the money moves.
- Instrument
- Reg. (EU) 2025/40
- Applies from
- 12 Aug 2026
- Basis verified
- 4 Sep 2026