Why New Systems Fail After Go-Live

Go-live day usually goes fine. The system turns on, the vendor’s team is on call, and leaders send a note thanking everyone for their hard work.

The trouble shows up later. Three months in, people are keeping side spreadsheets. Reports nobody trusts are being rebuilt by hand. Supervisors use a fraction of what the system can do. Some employees have quietly gone back to the old way. On paper, the project succeeded. In practice, the organization is paying for a system it is not really using.

Why does this happen so often? In my experience, and in the research, one cause stands out: people were trained on how the technology works, but not on how their organization would work with it.

What I saw firsthand

My doctoral research studied the managers of a university call center after the launch of a new reporting system. The managers valued the system. It gave them real-time data, better feedback tools, and more time for coaching.

But they were clear about one weakness. The vendor’s training was highly technical. It explained the software’s features but did not show them how to build and run the reports they needed for daily operations. There was also no clear plan for teaching managers who joined after launch. The system delivered results, but the learning curve was longer and harder than it needed to be.

Two very different kinds of training

Most technology rollouts include training. The question is what kind.

• Training on the technology explains what the system can do: the screens, the menus, the features. Vendors are good at this, because it is the same for every customer.
• Training on the work explains how this organization will use the system: which steps to follow for our processes, which reports answer our questions, and what changes in each person’s daily routine.

Vendors usually deliver the first kind. The second kind is the organization’s job, and it is the one most often skipped. Yet it is the one that determines whether the investment pays off.

What the research has said for decades

This is not a new insight. Researchers spelled it out in the 1990s.

In 1995, Dale Goodhue and Ronald Thompson published a study in MIS Quarterly introducing the idea of task-technology fit. Studying more than 600 people in two companies, they argued that technology improves performance only when two things are true: people use it, and it fits the tasks they actually perform. Use alone is not enough. A system can be switched on and logged into every day without making anyone more effective.

In 2000, Wanda Orlikowski of MIT added another piece in Organization Science. She showed that how a technology gets used is not fixed by its design. People shape it through their everyday work practices, so the same system can be used in very different ways from one organization, or even one department, to the next.

Put those two ideas together and the lesson is clear. Vendor training teaches the system as it was designed. Success depends on how it fits the work as it is actually done. Closing that gap is the organization’s responsibility.

Industry data points the same way. In a 2014 survey of organizations that had implemented ERP systems, Panorama Consulting found that every respondent experienced some disruption at go-live, and nearly half said it lasted more than a month. Resistance to change was the most common challenge, inadequate training was among the contributing problems, and low employee buy-in was linked to workarounds, lower system use, and fewer benefits.

Warning signs that a rollout is failing quietly

• Employees keep spreadsheets or paper notes alongside the new system.
• Only a few “power users” can produce the reports leaders need.
• New hires learn the system from whoever sits next to them.
• People describe the system as something “IT bought” rather than a tool they own.
• Leaders still ask for the same data the old way.

How to train for the work, not just the tool

1. Map the real workflows first. Before training, document how each role will actually use the system day to day.
2. Build training around roles and scenarios. A supervisor and an agent need different lessons. Use real examples from your own operation.
3. Pair vendor experts with internal experts. The vendor knows the software; your people know the work. Training needs both.
4. Develop internal champions. Trained super-users on each team give colleagues help when questions come up after the vendor leaves.
5. Plan for new hires from day one. Decide how people who join after go-live will learn the system.
6. Keep training after go-live. Schedule follow-up sessions at 30, 60, and 90 days, when real questions have surfaced.
7. Measure adoption, not just installation. Track whether people use the system for the tasks it was meant to support.

Why this matters even more in the age of AI

Artificial intelligence is changing what new systems ask of people. AI tools increasingly handle the technical side of the work: pulling data, drafting responses, spotting patterns, and even suggesting next steps. As more of the technical work is automated, the human contribution shifts toward judgment. People decide which output to trust, how it applies to a specific customer or situation, and when something does not look right.

That makes the lessons from Goodhue, Thompson, and Orlikowski more important, not less. An AI tool that does not fit the way a team actually works will be ignored, worked around, or misused, no matter how impressive it looks in a demonstration. And because people shape how technology gets used in practice, the same AI tool can strengthen one team and confuse another.

In the past, organizations could sometimes get by with training people on which buttons to press. With AI, there are fewer buttons and far more decisions. The human touch becomes the deciding factor in whether the technology delivers value.

The bottom line

New systems rarely fail on go-live day. They fail in the weeks and months after, when people try to fit a tool built for everyone into work that is specific to them.

Vendors can teach how the technology works. Only the organization can teach how the work will change. Leaders who invest in that second kind of training, and who keep investing after go-live, are the ones who see the return they were promised. In the age of AI, that investment is no longer optional.

Further reading

• Dale L. Goodhue and Ronald L. Thompson, “Task-Technology Fit and Individual Performance,” MIS Quarterly (1995)
• Wanda J. Orlikowski, “Using Technology and Constituting Structures: A Practice Lens for Studying Technology in Organizations,” Organization Science (2000)
• Panorama Consulting Solutions, 2014 ERP Report: Organizational Change Management
• Elpidio J. Garcia, A Qualitative Study of the Effects of Using a New Reporting System in a University Call Center, Nova Southeastern University (2016). See the Research page for a summary.