As new research from IBM reveals 8 in 10 UK businesses are struggling to keep pace with AI governance, Bobby Brown, Founder and CEO of specialist data consultancy Nucleo comments on the findings. 

1. How do legacy systems – particularly ERP – tend to limit the AI ambitions of mid-market firms?

Legacy ERP systems were built to bring control, consistency and structure to core business processes – and they still do this very well. The challenge is that AI introduces a different requirement: real time access to connected, high-quality data across the organisation.

Where firms run into difficulty is when that controlled data sits in silos or lacks the context AI needs to generate reliable outputs. It’s not a failure of ERP, it’s a gap between how systems were originally designed and what modern intelligence demands.

Upgrading technology without fixing broken processes is just putting a shiny coat of paint on legacy problems.

True transformation happens when you combine the control of an ERP with the predictive power of AI. When you clean up data pipelines, strengthen governance and surround your core systems with the proper context, ERP serves as a reliable and secure launchpad for AI to scale safely.

2. How do manual workarounds, disconnected data and ageing core systems impact efforts to embed AI across finance and supply chain functions?

Manual workarounds are often seen as a temporary fix, but in reality, they more often than not become the operating model. Spreadsheets, side systems and human interventions fill the gaps left by disconnected core systems and that creates inconsistencies at scale. For example, if you’re feeding in data that’s been manually tweaked, duplicated or interpreted differently across different teams, there will never be one single version of truth.

In finance and supply chain, where timing and accuracy are hypercritical, that lack of consistency undermines everything.

It also creates hidden time and resource drains. If teams are spending more time validating outputs than using them, then AI burnout is your next big issue. People should be benefitting from AI output, not having to effectively mark them. And this all comes at a financial cost.

3. If the goal is smarter, AI-driven operations – rather than simply upgrading an ageing system – how might it change the way mid-market firms approach ERP modernisation?

If the goal is AI-driven operations, then ERP modernisation needs to start with outcomes, not systems. All too many firms focus on replacing technology rather than redesigning how the business runs.

The question that needs to be asked and answered is “what decision do we want to improve, and what data do those decisions depend on?”. With this, you can start identifying the critical workflows, what data is required and where AI can genuinely add value. This will give you structural clues for the architectural system you’ll need to support your business’ development.

This approach will also force you to simplify, and separate the wheat from the chaff. AI doesn’t work well in overly complicated environments, so streamlining processes will deliver far more value than upgrading platforms.

But, perhaps most importantly, this process will bring accountability back to the business. The people responsible for outcomes need to be involved in shaping the system, not just handling it after implementation.