Effort scales with volume
Every order, every supplier, every reference adds hours of data entry, reconciliation and chasing. Business grows, the team does not keep up, and nobody can say when it will break.
HomeTransformation & AI
Modernise operations
Assessment, roadmap and industrialisation — with deep expertise in supply chain and applied AI.
THE STATE OF OPERATIONS
Organisations have always adapted their operations to the tools available.
It has become possible to do the opposite.
What changed is not the power of the models. It is the cost of putting one into production on a narrow scope. What follows describes what we observe on engagements, and what we conclude from it.
Three failures
Cost exceeds value returned
The tool is paid for everyone and fits no one. You fund modules never opened, work around those that do not fit, and the workaround ends up in a spreadsheet.
The tool does not exist
The need is too specific to the trade for a vendor to address. The market is too narrow, or the practice too new. No budget will buy what nobody publishes.
Different sectors, different sizes, the same three failures.
How everyone coped
Four workarounds, found everywhere. None of them addresses the failure: they move the burden.
The spreadsheet that runs the company
It works, until the day its author goes on holiday.
The half-fitting SaaS
You pay for all of it, use a third of it, and maintain the rest by hand.
Custom development
Quoted, debated, postponed. The estimate arrives before the scope is settled.
Going without
The decision stays with judgement, the load stays with the team, and the subject comes back every year.
Why the classic route stayed out of reach
Cost
A firm quote is demanded on a scope that is not yet defined. The estimate protects the supplier, not the project.
Timeline
Twelve to eighteen months between the decision and first real use. The problem carries on in the meantime.
Arbitration
Nobody internally is positioned to choose between a business rule, a statistical model and a generative model — three answers whose maintenance costs bear no comparison.
The choice came down to funding a large programme or going without. For years, going without usually won.
WHAT CHANGED
You no longer commit to a programme. You instrument a decision.
The useful scope has shrunk to the point of fitting in a few weeks. A decision that recurs often, data already available to inform it, a gain you know how to measure: that is the smallest unit of transformation that still produces a real effect. Below it, nothing happens. Above it, you are back in the eighteen-month programme.
On the flow side
Demand forecasting, stock allocation, supplier quality control, document reconciliation. What decides what ships and what stays.
On the desk side
Re-keying, checking, chasing, looking things up. What fills the day without appearing in any indicator.
THE EVIDENCE
What separates a production system from a demonstration
You do not wire a model into order release just to see what happens.
It runs on your real data
Not on a curated extract. The messy cases are in scope, or the system does not survive its first month in operation.
It is integrated, not parked alongside
Outputs come back into the ERP or the tool the team already has open. One more screen is one more screen nobody opens.
It has a stop criterion, written before starting
The threshold below which you stop is fixed at scoping. Without it, a disappointing system carries on out of habit and embarrassment.
Someone on your side can fix it
Documentation, skills transfer, a trained point of contact. If we are still indispensable after a year, the engagement failed.
What it unlocks
The gain does not always take the form of an hour saved. It comes in four shapes, and they are not equivalent.
A capability that did not exist
The system replaces nothing: it makes possible a decision you could not take, for lack of reading the data in time.
A subscription you cancel
The half-fitting tool no longer has a reason to be paid for, or worked around.
A manual process you retire
The hours recovered are not reassigned to the same task elsewhere. That is the condition for the gain to be real.
A dependency that falls away
The team evolves the system without going back through a supplier, us included.
DIGITAL TRANSFORMATION & AI
AI in service of operations,
not demonstrations
We work where data is most complex and most valuable: the supply chain. Forecasting, supplier quality, document flows — we start from a measurable use case, then industrialise it.
Assessment
Mapping flows, data and on-the-ground pain points.
Use case
A costed scope, an expected gain and a stop criterion.
Industrialisation
Production rollout, IT integration, security and compliance.
Adoption
Team training, run and continuous improvement.
−23 %
Forecast error
Demand models trained on customer history, seasonality and supplier variability.
×4
Quality-control speed
Automated non-conformity analysis and supplier-audit prioritisation.
−70 %
Document data entry
Extraction of delivery notes, invoices and customs documents, fed back into the ERP.
BUSINESS APPLICATIONS
Low-code applications
Digitise business processes without waiting for a full IT cycle: focused, integrated, maintainable applications.
CHANGE MANAGEMENT
Adoption and training
A tool only exists if it is used: team training, run support and continuous improvement.
How an engagement runs
Assessment
1 to 2 weeksMapping of processes, data flows and friction points. We measure where time is lost before proposing anything.
Roadmap
1 weekUse cases ranked by expected value and real feasibility, accounting for the quality of the data you actually hold.
Prototype
2 to 4 weeksA single use case, taken through to real use by your teams. A prototype that never leaves the demo proves nothing.
Industrialisation
variableDeployment, training, documentation and skills transfer, so the setup belongs to you.
What you receive
- Mapping of processes and data flows
- Costed use cases: expected gain, effort, dependencies
- Working prototype, proven on your real data
- Technical documentation and skills transfer
- Change management plan and training materials
Frequently asked questions
Where do we start if our data is messy?
The assessment says so plainly. In most cases the first project is about data, not artificial intelligence: a model fed inconsistent data produces inconsistent answers, faster.
How long before we see a result?
A usable prototype on a real use case takes four to six weeks from kick-off. That is deliberately short: a setup you cannot try for six months never gets corrected in time.
Do we have to replace our existing tools?
Rarely. We integrate with what already works. A project that starts by replacing the ERP fails for reasons that have nothing to do with AI.
What makes a use case viable?
Three things: a decision that recurs often, data already available to inform it, and a gain you know how to measure. If one is missing, the use case waits.
Is generative AI always the right answer?
No, and saying so is part of the job. A business rule or a simple statistical model solves many problems more reliably, at a cost and maintenance burden that bear no comparison.