ERP’s biggest challenges haven’t changed.

Where we started.

In my first 10 years, many ERP programs began with a project charter that spelled out what the business expected to gain. Improve production by a set amount. Reduce inventory waste. Close the month in fewer days. After go-live, you were to go back and check. Was there less waste in the warehouse at the next stock count? Was the business leaner? Did you hit the goals as defined?

This was, of course, complex to measure because you can’t control everything outside the scope of a project, but the principle was there.

Over the decade that followed, that rigor took a back seat. Modernization itself became the primary driver. A system was nearly out of support, so it had to be replaced. The business needed a SaaS operating model in the cloud. All were valid reasons to invest, but the business case often came down to more “we had to” versus “we are investing in specific outcomes.”

The industry landed on something more like a snapshot of success—the date, the budget, and whether the system went live as expected—instead of counting on outcomes.

On time, on budget, and often off target.

Even by those standards, plenty of programs didn’t succeed.

The industry’s track record underscores this. By now, most of us have heard the Gartner projection that by 2027, more than 70% of recently implemented ERP initiatives will fall short of their original business goals. Part of the reason is history: many ERP programs were built as monolithic systems due to misaligned judgment calls—and then, in a bit of a vicious cycle, talent moved away after programs became notoriously problematic. All of that still boils down to the people working beside you, helping you make the right decisions at the most critical points in your project.

Birmingham City Council is an extreme example. Decision-makers there planned to implement ERP mostly out-of-the-box, then built customizations around processes teams knew were problematic (including a banking reconciliation function that couldn't carry forward). In the end, the council couldn't produce auditable accounts, had to turn off fraud auditing for more than 18 months, and booked £2 billion in transactions to the wrong year. A program first costed at around £19 million is now forecast toward £144 million.

This is where human judgment does the work. A clean core doesn't mean zero customization; that isn't realistic. It means knowing which extensions are worthwhile, which create technical debt, and where and how to implement. That discernment guides where expanded agentic capability belongs too. It’s people who spot the risk, identify the right path forward, and build a foundation that drives your next phase of growth.

The most essential factor, then and now.

When programs delivered, the difference came down to people. At the core, most ERP products could handle basics: raise a purchase order, post an invoice, close the month. Microsoft stood out because ERP, CRM, data, and now AI sit in one connected stack, a stronger foundation than one assembled from parts.

The returns bear that out: Independent Forrester studies project more than 100% three-year ROI for enterprises and a 16-month payback for midmarket organizations on Dynamics 365 ERP.

Even so, the platform alone didn’t guarantee the result.

The same Dynamics 365 can underpin a business that pulled ahead or one that struggled all the way from implementation to adoption. The difference was the team supporting you, their industry knowledge, and the solutions they brought to the table.

Successful programs had strong leadership, adoption, and change management, alongside a delivery team that built trust and credibility with their counterparts. At HSO, we’ve been intentional about hiring people from the industries we serve. They speak the customer’s language and have stood in their shoes, so when a hard decision arises, they can make the case for the right change to a process or surface the most relevant industry-specific solution.

What does AI change about ERP implementation?

It’s true: the mechanics of delivery are changing faster than I’ve seen in my career.

Agents now can capably absorb the repetitive work that once consumed much of a project plan. That starts with capturing workshops, generating functional and technical specs, configuration, development, testing, and data migration—a list that will continue to grow and evolve.

Based on our work with AI-native delivery, my estimate is that it is possible to reduce delivery timelines by 20% to 30%, if not higher, especially as the technology advances. That's dependent on a customer’s ability to keep up with the pace: their team's capacity, data readiness, and their overall comfort with AI-enabled implementation.

So it's not that speed isn't part of the emerging picture, but it's important to be precise about where that acceleration is directed.