Two weeks to a go or no-go, backed by numbers

Project Overview

Who is Graphius?

Graphius is a printing group from Ghent, founded in 1928. The company prints and finishes books, magazines, comics, packaging and labels for customers across Europe, the UK and the US. Today the group has more than 620 employees, spread over ten sites in Belgium, the Netherlands, France and the UK.

That fast growth also threw sand in the gears. The order process got stuck in email traffic, and customer service felt it every single day.

The challenge

Graphius processes around 40,000 orders a year. Each order generates 38 emails on average. Across the whole group, that's 60,000 to 100,000 incoming emails a month.

All that communication sits scattered across personal inboxes, folders and Outlook threads. There's no single source of truth. Mistakes in specs, missing delivery addresses and wrong page counts only surface too late. That leads to reprints, credit notes, and even more emails.

"How do we take the load off customer service so we respond faster, make fewer mistakes and work more professionally, with an investment that pays for itself within the year?"

Project Delivery

The approach: a Business Value workshop

We work in three steps: first find the value with an A.I. Opportunity Mapping, then prove the value with a Business Value workshop, and only then build. Graphius already had a clear picture of the problem, so we skipped the mapping and went straight to step two.

On April 29, 2026, we spent a full day together in Ghent. Around the table: customer service, planning, IT and Graphius management. No sales pitch — technical people and decision-makers in the same room. We named the project OrderPulse.

Over that day we did four things:

1. Sharpened the goal and the KPIs. Calm, quality and grip on the day-to-day work around orders. Measured through four KPIs that need to go down: emails per order, time to request missing info, print errors caused by wrong specs, and the mental load on staff.
2. Mapped the as-is process. From quote confirmation to production and delivery, including every pain point: incomplete information, files arriving late or wrong, extra pages nobody ordered, and a lack of overview.
3. Designed the to-be process. Together with the people who work with it every day.
4. Calculated the business case. Live, based on their own numbers on order volume, email volume and complaint costs.

What we delivered

Two weeks later, on May 11, 2026, the full package was on the table.

1 · A working mockup of OrderPulse

Not a slide with a sketch, but a clickable mockup the team could try out themselves. It included:

• An order list with filters by status, client and team.
• A detail page per order with all documents, specs, dates and communication in one screen.
• A checklist per order, with tasks and deadlines generated automatically.
• All emails per order bundled into one timeline, with an A.I. summary of the latest status.
• An A.I. check that cross-references specs from the ERP, the emails and the preflight file, and flags differences.
• Draft emails A.I. prepares when info is missing, which the employee finishes and sends from Outlook.

2 · A scope in three phases

Phase 1 is an internal dashboard and overview. Phase 2 adds the A.I. layer and the link with CERM, the ERP system for the printing industry that stays the source of truth. Phase 3 is an external client portal. Phases 1 and 2 together form the MVP, so the project delivers value before everything is finished.

3 · 26 prioritized user stories

Written out per role: customer service, planner, production, system and client. Each with a complexity estimate and a MoSCoW priority, so everyone knows what's in the MVP and what comes later.

4 · A technical architecture

Hosted in Azure Western Europe, so no data travels to the US. The mailboxes connect via webhooks and Azure Service Bus, so emails automatically attach to the right order number. Login through the existing Microsoft 365 account. The A.I. models don't train on Graphius's data.

5 · A well-founded business case

No back-of-envelope guess — a calculation based on their own numbers.

Results

The Business Value workshop mapped out the expected annual return of OrderPulse in full.

That return comes from two places. The biggest share comes from time saved: 10 percent less time per email, across 38 emails per order and 40,000 orders a year. The second share comes from fewer errors: a quarter fewer mistakes around order handling, thanks to better overview and automatic A.I. detection.

Together that adds up to an ROI of roughly 7 times the investment, with a payback time of under two months. On top of that sits a lot of value we deliberately left out of the numbers: less stress for staff, higher customer satisfaction, better client retention and a more professional image.

Maybe the most important result is this: Graphius could make a go or no-go call based on numbers, not on a gut feeling. Two weeks after the workshop, the decision was on the table.

Flowkify keeps supporting Graphius as its A.I. partner, with a focus on measurable impact and a phased approach.

Curious what A.I. could deliver in your production or order process? Get in touch with Flowkify for a tailored Business Value workshop.