The CXO Who Signed an ERP May Have Already Made a Mistake

You signed an 18-month, $5M contract with a system integrator to roll out your ERP. The board approved it. Legal reviewed it. You shook hands.

Here is what nobody told you before you signed.

You are not the exception. You are the rule.

It starts at requirements. And it never recovers.

Nobody tells you this at project kickoff: the requirements phase is where ERP projects are actually won or lost. Not testing. Not go-live. Requirements.

A consultant reads through your documents — process maps, old spreadsheets, operational manuals. They are good at their job. They are also on three other projects. They work sequentially through what you send them, flag what looks relevant, and still have more questions than answers. Six weeks in, they are still asking your Finance Director questions the sales team already asked in the pre-sales cycle.

The requirement buried on page 34 of your operations manual — the one describing how your intercompany recharges are calculated when a subsidiary operates across two tax jurisdictions — doesn’t look like a system requirement. It looks like an accounting footnote. It doesn’t make it into the gap analysis.

It surfaces in UAT. Eight months in.

Now it is a scope issue. A budget issue. A delay. And the first phase running late means every phase after it is already compressed before it starts. Configuration waits on requirements. Testing waits on configuration. Training waits on testing. The 18-month project is already an 18-month project in trouble.

Here is what the middle of your project actually looks like

You are six months in. Configuration is underway. Your best people — the ones who understand both the ERP and how your business actually works — are the only ones who can do it properly. So they are doing it.

Every time a configuration changes, the testing has to be repeated. Not some of it. All of it. Because a change to how your system handles multi-currency intercompany transactions could ripple into invoicing, into tax reporting, into bank reconciliation. Your SMEs know this. So they test it again. Manually. Step by step. Every time.

Take something every business running a distribution or manufacturing operation will recognize: lot traceability. Your ERP needs to track a batch of goods from goods receipt, through warehouse put away, through pick and dispatch, all the way to the customer invoice — and link back to the supplier lot number at every step. Simple in principle. But the configuration involves item setup, warehouse parameters, lot control activation, lot assignment rules at each transaction type, and links to quality inspection workflows. Change one parameter — say, you decide lot assignment should happen at pick rather than receipt — and you need to retest every single transaction in that chain.

Your warehouse manager is doing that retest. Again. For the third time this month. While your operations run without her.

And nobody has written any of it down

The decisions made in the requirements phase — what was agreed, what was deferred, what was explicitly ruled out — are scattered across email threads, workshop notes, and the memory of a consultant who has since moved to a different engagement.

When the configuration team asks why the system handles purchase order approvals above €50,000 differently depending on the legal entity, nobody can find the answer. It was decided in a workshop in month two. The person who agreed it is no longer on the project.

When you go live and something breaks in post-go-live support, the team picking up the ticket has no access to what was configured, why it was configured that way, or what business requirement drove it. Every complex issue starts from scratch. The institutional knowledge built up over 18 months has walked out the door with the consultants.

Your warehouse team on day one

Your go-live comes. Day one. Month one is fine — everyone is on high alert, the hypercare team is there, the obvious issues get caught.

Month two, a warehouse operative tries to process a return from a customer against a partially received order that was originally created against a blanket purchase agreement. It is not a day-one process. Nobody thought to test it. The system doesn’t handle it cleanly. She has never seen this screen before. She freezes.

Your shipments start to back up. Your customer calls.

This is not a technology failure. It is a training failure baked in from the start — because your SMEs were testing, not training, and your end users never got enough hands-on practice before go-live to handle anything that deviated from the core script.

Meanwhile, your competitor started six months after you
They will go live before you.

Not because they have a better team or spent more money. Because their implementation contract included AI-accelerated delivery, and yours didn’t.

Requirements.

Their requirements were gathered in weeks, not months — not because AI replaced the decision-making, but because their senior team had every document, every process map, every workshop transcript analysed and structured before the first decision meeting. The right people made faster decisions because they had complete information, not fragments. The CEO had real visibility at every stage — not hearsay from a weekly project update, but data from the system showing exactly where things stood.

Testing.

Every configuration change was regression-tested the same day. Not because testing became instant — but because AI generated the full test script library directly from their requirements, in a matter of clicks. When the configuration changed, the team ran the regression suite. By morning they had a complete report showing exactly which processes passed, which failed, and what needed attention. Their SMEs reviewed results, fixed what needed fixing, and were back in the business. The testing didn’t own them. They owned the testing.

Training.

While testing was running, training was being built in parallel — something that simply cannot happen in a traditional project where SMEs are the bottleneck for both. Training videos and assessments were generated directly from the configured system, in over 140 languages, in days. Role-specific. Always current. End users completed training and assessments before go-live, monitored digitally without anyone chasing. Only users who passed their assessments were granted production access. No guesswork about readiness. A hard gate, fully automated, fully visible. Their SMEs never ran a training session. They stayed on the project.

Post go-live.

When something breaks in production — because something always does — their support agent knows the full history. What requirement was agreed, what configuration decision was made, what was tested, how the user was trained. The answer is not buried in an email from eight months ago. It is there, immediately, in full context. The agent knows exactly why that invoice wasn’t sent — because it was deliberately turned off in a configuration change driven by a specific business requirement, documented, traceable, searchable.

Their warehouse team practiced the lot-traceability return process twelve times before day one. They knew exactly what to do.

WHAT COMPANIES RUNNING THIS MODEL SAY

"It fundamentally changed how we deliver ERP transformation — shorter deployment cycles, reduced operational risk, and a smoother transition for our teams. This was not incremental improvement."
PATRICK BROWMAN, CIO — NEW WAVE GROUP

"Infomind AI’s ability to generate training videos from automated test scripts made user onboarding seamless, reducing support needs and enhancing user adoption."
ANISE MADH, GENERAL MANAGER GLOBAL INFOR PRACTICE — WIPRO

"We onboarded our team on Infor faster than expected with multilingual video training. A partner we now see as strategic for test and release management."
JAN LARSEN, SENIOR CONSULTANT & PARTNER — CCG / PROSOURCE

You cannot fix this mid-project. The only moment you have full control is before you sign.

The methodology is in the SOW. The phases are sequential. Once you are six months in, you are on the trajectory you started on.

One question is worth asking your SI before you do:

Does this contract include AI-accelerated delivery across requirements, testing, training, and post-go-live support?

If the answer is no — and if your competitor is about to ask the same question to a different SI — you already know what happens next.