How AI Is Changing Legacy ERP Modernisation (And What It Still Cannot Do)
AI can automate data migration mapping, generate test cases, and assist with documentation — but the hard parts of ERP modernisation are still fundamentally human problems.
ERP modernisation projects are legendary for their difficulty. The CHAOS Report's statistic — 75% of ERP projects fail to meet their objectives — has not improved much in two decades. AI is genuinely changing some of the underlying economics, but not in the ways most vendors will tell you.
Where AI Actually Helps
- Data migration mapping: LLMs can read legacy field descriptions and propose target-system mappings with 70–80% accuracy, then flag exceptions for human review. This used to take weeks. Now it takes days.
- Legacy code archaeology: AI-assisted analysis can build a dependency graph of 20-year-old ABAP or COBOL modules in hours. A team of humans doing this manually takes months.
- Test case generation: Given source system documentation (often incomplete), LLMs generate comprehensive edge-case test scenarios that human testers frequently miss.
- Documentation generation: AI writes first-draft documentation from code comments and schema. Usually 60% accurate and 100% faster than starting from nothing.
Where AI Fails in ERP Projects
ERP systems contain decades of business logic encoded informally — in stored procedures, in Excel sheets that feed data imports, in manual processes everyone follows but nobody documented. AI cannot discover this logic because it exists outside the system. This is the hardest part of modernisation and it requires experienced humans conducting structured interviews with people who have been at the company for 15 years.
The Organisational Change Problem
The number one reason ERP projects fail is not technical. It is political. Finance does not want to move to the new chart of accounts. The warehouse team lead will not use the new scanning workflow because it feels slower. AI can do nothing about any of this. It is a human and leadership problem that requires investment in change management, not technology.
Our Practical Approach
At AnaravTech, we use AI to compress the technical archaeology phase — going from 'we have this legacy system' to 'we understand what it does' faster than ever before. We then spend the saved time on the human-critical parts: stakeholder interviews, process mapping workshops, and change management planning. Technology accelerates the parts AI is good at. People handle the parts that matter most.
More Articles