AI and analytics projects in five steps In five steps to measurable value, with success and stop criteria before every implementation

An iterative approach: process by process, with a deliberate decision after every step.

The data sits in ERP, DMS and Excel. Decisions are still made by hand.

Many AI pilots fail in one of two ways: the problem is vague, or nobody stays after the demo. That is why our forward-deployed engineers work on site with you, on your data. The approach has five steps and only ends once the agreed metric improves measurably in day-to-day operations.

Step 01Understand the challengeCapture processes, data and bottlenecks. Quantify sources of loss and the biggest levers.
Step 02Set target and success criteriaPrioritize starting points by value. Target, KPI and stop criterion for each process.
Step 03Implement step by stepRedesign and digitize the process, add AI where it counts. First result in operations.
Step 04Measure and optimize impactMeasure impact against the baseline, including adoption. Adjust until the target is reached.
Step 05Embed and scale the solutionMove into day-to-day operations and build capability. Extend to other areas.

Every step delivers a concrete result. Together we review what works, what needs adjusting and how to proceed. A well-founded stop is a result too.

Step 01: Understand the challenge

Everyone usually knows where things get stuck. How often it happens and what it costs is rarely quantified. We capture that on site with the people who make the decisions today. Along the way we check which data is available.

What you walk away with

A data map and prioritized use cases, assessed by value and effort.

What we do: On-site interviews and a review of ERP, line-of-business systems and file storage. We count which decisions come up how often and how long they take.
What we agree on together: Which areas we look at and who works with us on your side.

In the warehouse of a mid-sized company two men with folders stand between high-bay racks, a forklift in the background.

Where the data sits and where it flows

Result after step 01Example · Back office · Accounts payable

approx. 1,200invoices per week

4 in 5clarified by hand today

DecisionPost the invoice or clarify it

Most frequent reason to clarifyMissing purchase order number, on 18% of all invoices

Data sits inERP, DMS and mailbox, not linked

ProposalAutomate the match against purchase order and goods receipt

Values for illustration

Examples from manufacturing and construction

Quality inspectionIndustry
TodayEvery inspection lot goes through by hand, data in three systems.
ResultMost lots are clearly within tolerance, only borderline cases need an inspector.
Change order reviewConstruction
TodayChange orders arrive by email, the measurements sit in the project file.
ResultThe time goes into finding documents. Proposal: match against the bill of quantities.

Values for illustration

Step 02: Set target and success criteria

No pilot without a target. No acceptance without measurement.

Many AI projects end with a demo and no answer to what they delivered. So before implementation starts, we agree in writing how success is measured, what may run without approval and when we stop.

What you walk away with

A signed agreement with target, measurement plan and business case. Plus a clear recommendation on whether implementation pays off.

You can also commission steps 01 and 02 on their own, as the AI Value Scan.

What we do: We measure the starting point and propose a target and limits. Then we work out the business case.
What we agree on together: Target, stop criterion and the threshold above which a person decides. And on that basis, whether we go ahead with implementation.

In a production hall two colleagues discuss a plan on a control cabinet, one pointing at it with a pen.

Baseline and target

Your target, set in writing at the startBaselineTargetbetter

Target agreementExample · Back office · Accounts payable

22%Baseline

60%Target

0.5%Stop threshold (error rate)

MetricInvoices posted without manual touch

Baseline, measured22%

Target60%, over the agreed measurement period

Approval limitAutomatic up to €5,000, above that to accounting

Runs automatically ifPurchase order, goods receipt and price match

Stop ifError rate persistently above 0.5%

Values for illustration

Examples from manufacturing and construction

Quality inspectionIndustry
TodayInspection effort per shift is high, escapes are low.
AgreedHalve inspection effort per shift. Stop as soon as escapes rise.
Change order reviewConstruction
TodayA change order sits for a long time before approval.
AgreedHalve the approval time per change order. Anything beyond the approved budget goes to site management.

Values for illustration

In a production hall a woman and a man look at a tablet together at a production line.

Step 03: Implement step by step

The legitimate concern: a system decides something nobody approved. That is why only what you approved in writing beforehand runs automatically. Everything else goes to the people responsible, with all the data.

What you walk away with

A solution that runs on your systems in daily operations. Plus a team that works with it confidently.

What we do: We connect your systems, set up the check rules and build an interface for exceptions. Every decision is logged, and a manual fallback is in place.
What we agree on together: The check rules, and when they go live for the first cases.

Schematic example

Rules handle the standard case. People handle the exceptions.

How far a workflow runs on its own is something you set upfront. Here the principle is: what the rule clearly covers runs automatically. Everything else goes to the people responsible, with the data. Until then the system waits.

Step 04: Measure and optimize impact

Whether a project paid off is often judged by gut feeling. Here we measure against the target you set at the start, over a period agreed in advance. Where rules are too strict or too loose, we adjust them.

What you walk away with

A measurement report against the agreed target, every rule change in the log.

What we do: Measurement, analysis of the exceptions, proposals for rule changes.
What we agree on together: What happens if the target is missed, up to and including a stop.

In the office an employee leans in, smiling, towards a colleague who shows a report on her laptop.

Measured against the target

Measured against the agreed targetBaselineTargetbetter

Measurement reportExample · Back office · Accounts payable

22%Baseline

60%Target

54%Measured, interim

StatusInterim measurement within the agreed period

Measured54%, target 60%

Error rate0.2%, stop criterion 0.5%

Rule changes3, each with a reason

Values for illustration

Examples from manufacturing and construction

Quality inspectionIndustry
MeasuredLess inspection effort, escapes unchanged.
DecidedThe rule stays, the band for borderline cases gets narrower.
Change order reviewConstruction
MeasuredLead time down, two assignments were wrong.
DecidedAssignment rule corrected with a reason, measurement continues.

Values for illustration

Step 05: Embed and scale the solution

Many solutions stay dependent on the vendor after the project. Here operation, maintenance and responsibility pass to your team. Whether the next area follows is a separate decision, with its own target.

What you walk away with

Your team runs the solution on its own, with documentation and clear responsibilities.

What we do: Training, documentation, handover of the rules and the monitoring. On request a proposal for the next area.
What we agree on together: Acceptance of the handover. Separately, whether and where to continue.

In a hall with injection moulding machines a woman holds a tablet while a colleague points at the screen.

Your team takes over

Handover planExample · Back office · Accounts payable

2 rolestake over operation

Own approvalfor the next area

OperationAccounts payable team lead

TechnologyYour IT

Rule changesBy your team, each with a reason

Next areaTravel expenses, after your approval

Examples from manufacturing and construction

Quality inspectionIndustry
Handed overQuality management runs the rules itself.
NextIncoming goods inspection, with its own approval.
Change order reviewConstruction
Handed overSite management monitors the rules.
NextProcurement, with its own target.

Values for illustration