Enterprise Resource Planning (ERP) systems in manufacturing are integrated software platforms that help companies manage and streamline core operations, such as production planning, inventory control, supply chain management, and financials.
By unifying these processes in a single system, ERP boosts efficiency, reduces errors, and provides real-time insights to support smarter decision-making.
Now, with the rapid rise of artificial intelligence, many manufacturers are looking to embed AI into ERP to maximize the value of their systems.
To explore how the UK is approaching this shift, The Manufacturer sat down with Arturo Buzzalino, Chief Innovation Officer & GVP at Epicor, for a deep dive into AI adoption in ERP and the opportunities it presents.
Can you give us some background on the current landscape of AI adoption in ERP systems within the UK?
AB: Adoption is real, but early. Most UK firms aren’t ‘building AI’, they’re learning how to apply it inside ERP, where the private, structured data that runs the business lives.
Pilots are everywhere. These are typically in repetitive, decision-support areas (e.g., surfacing data for planners, accelerating routine reporting) but at-scale rollouts are rarer because foundations (cloud migration, data hygiene, process standardisation) are uneven.
The point many misses is that ERP will be the primary vehicle for meaningful AI, because it sits across finance, order-to-cash, purchase-to-pay, and shop-floor execution. These are the places where timely, accurate context matters.
The winners work backwards from a few high-ROI operational problems, embed AI where work already happens, and prove value with tight, measurable use cases before scaling. That’s the posture we see gaining traction across UK manufacturing.
Why do UK businesses need to focus on foundational technologies before AI?
AI amplifies whatever it touches, so if your ERP/data/process landscape is messy, AI will accelerate the mess.
The right order is:
- Modern, cloud ERP to standardise and connect core processes
- Data hygiene (cleansing, deduping, consistent master data)
- Process simplification (retire duplicate reports, reduce bespoke workflows), then
- Targeted AI experiments.
This order matters because AI’s value hinges on reliable, accessible context. If inventory locations disagree across systems or yield calculations differ by plant, an assistant or recommender will surface inconsistent answers and erode trust.
How important is moving ERP systems to the cloud for UK companies?
Critical. Cloud ERP is the enabling layer for speed, consistency and scale, the three things AI needs to be safe and useful in operations.
Practically, cloud consolidates versions and configurations, which makes it feasible to roll out new AI capabilities (copilots, search, recommendations) across sites without a bespoke project per plant. It also centralises identity, authorisation and data access so AI can retrieve the right facts with the right permissions, reducing shadow spreadsheets and unsanctioned data copies.
In short, if AI is the engine, then ERP-in-the-cloud is the chassis and road network. Without it, you’re limited to small, fragile, standalone pilots. With it, you can ship value safely and repeatedly.
What are the biggest data issues you see in UK manufacturing and distribution?
There are probably three recurring issues.
First, legacy estates (older on-prem ERPs plus spreadsheets) create brittle interfaces with missing context. Second, silos and duplicate reporting mean multiple ‘truths’ (e.g., demand, stock, and margin calculated differently by team), which undermines trust in any AI-surfaced answer. Third, inconsistent standards (item masters, locations, units of measure) make retrieval-augmented AI hallucinate because the inputs disagree.
These issues are solvable. Standardise masters and transactional schemas where feasible, rationalise reports to a governed set and expose ERP data via secure retrieval patterns so copilots can cite their sources.
There’s a very simple way to test your readiness: can a line supervisor, buyer, or scheduler ask a plain-English question and get a consistent, explainable answer every time? If not, fix the plumbing before you add more taps.
How can UK firms spot and fix inefficiencies before bringing in AI?
Start on the floor and always work backwards from problems. Shadow your schedulers, buyers, and warehouse leads for a day and list the tasks that are frequent, manual, interrupt-driven, or require hopping between screens/spreadsheets.
Cluster them by outcome (e.g., prevent stockouts, reduce past-due, shorten changeovers), then quantify impact (time, rework, service level).
Pick 3–5 with clear ROI and fix the basics: retire duplicate reports, simplify approval chains, tighten master data ownership, and standardise how a given decision is made.
This two-step approach avoids automating waste and gives you measurable variables when you do pilot AI.
When do you think is the right time for a UK business to start testing AI in ERP?
As soon as three conditions are true:
- You’re on cloud ERP (or have equivalent connectivity/governance),
- Data is clean enough for the use case (one truth for the entities AI will touch), and
- The workflow is stable (you know the happy path and the expected exceptions)
Then run a narrow pilot on a repetitive, high-volume decision with tight guardrails. Think something like retrieving shipment risk for open orders, guiding a cycle count, or drafting a vendor comms summary. Success criteria should be explicit (e.g., minutes saved per transaction, fewer touches, SLA uplift), and the pilot should be embedded where users already work, not in a separate tool.
From there, expand to adjacent steps. The rule of thumb is if a human can describe the task crisply and the system can cite the data behind the answer, you’re ready to try AI.
What makes an AI feature intuitive and human-like, in your view?
Three qualities: context, flow and explainability.
Context means the assistant knows the user’s role, the transaction in view, and the company’s vocabulary, so it anticipates the next question and speaks in local terms (SKUs, bin codes, routings). Flow means it’s available where work already happens (the PO, the job, the pick), with one-tap actions that reduce cognitive load rather than adding steps. Explainability means that it shows its working behind its answers, so users trust and adopt it.
For frontline teams, intuitive also means zero-to-low training: natural language, short responses, and sensible defaults. If a picker can ask “What’s the fastest way to complete wave 27?” and get a concise, data-backed plan they can act on immediately, that’s human-like.
How can ERP providers make AI tools easy for busy frontline workers to adopt?
Design for time-to-answer, not wow.
Put the assistant in the workflow (on the job/PO/pick screen), keep interactions concise, and bias for actions over essays.
Optimise for one-handed mobile use and intermittent connectivity. Let users speak or type in plain English and return answers that cite ERP facts they already trust. Use role-aware guardrails (e.g., a warehouse lead can trigger a cycle count, a picker can only request it) and log decisions for audit.
Above all, measure adoption on shift, seconds saved, touches removed, rather than just clicks.
Looking ahead, what unique opportunities do you see for AI in ERP in the UK market?
There are three that stand out to me. A major one is industry-tuned AI models that fine-tune commodity LLMs with ERP ontologies and domain data. That means the assistant will speak manufacturing/distribution, improving retrieval and recommendations without building a foundation model from scratch.
There’s also the opportunity for faster ERP onboarding where AI maps masters, flags anomalies, and scaffolds workflows, compressing months into days and reducing change-fatigue risk.
Lastly but by no means least is manufacturing productivity as a whole. The mix of legacy estates and high manual load in UK plants makes targeting repetitive tasks (expedites, confirmations, QC checks) and exception handling a huge opportunity. There is very real innovation up for grabs as human time shifts from combing spreadsheets to redesigning processes. That’s how UK firms turn AI from hype into margin, service, and resilience gains.
For more articles like this, visit our Digital Transformation channel.
