ProfitabilityCompleted work lost money
Open the work orders and review recorded cost evidence.
Fraction ERP Pro plan feature
AI in ERP Without the Guesswork
The Fraction ERP Intelligence Centre turns live manufacturing data into a ranked, evidence-led decision queue. It is intelligent, but it is deliberately not a generative AI agent.
Defined rules find the exceptions that need attention, explain why each one was raised and link straight back to the underlying ERP record. No prompts, no copying business data into a free external AI tool and no different answer each time you ask.
Rules-based
Evidence linked
Human reviewed
Intelligence Centre
Rules-based
Things to look at first
Productivity & planningLabour exceeded plan
Investigate the highest operation variances first.
PurchasingPurchase prices moved
Compare weighted prices with the preceding period.
Analysis confidenceMissing inputs are shown, not silently guessed.
Illustrative view
What “AI in ERP” means
Intelligence is not the same thing as generative AI
ERP vendors increasingly use AI copilots and autonomous agents to answer questions, generate text and take actions. Those tools can be useful where the task is open-ended or unstructured. But a large language model is not automatically the best way to calculate a known cost variance, find a missing price or decide whether a threshold has been crossed.
For repeatable operational control, Fraction ERP uses a deterministic rules engine. The same ERP records, date range and product rules produce the same finding. That makes the result easier to verify, explain and act on.
Why rules-based intelligence
Use probability for open questions. Use rules for known controls.
This is not an argument against AI. It is a decision to use the right mechanism for the job.
Generative AI and ERP agents
Strongest when the task is open-ended
- Working with natural language and unstructured documents
- Drafting, summarising and exploring possible explanations
- Responding flexibly when there is no single fixed route
The trade-off: outputs are probabilistic, need grounding and can be factually wrong or inconsistent.
Fraction ERP Intelligence Centre
Strongest when the control is definable
- Testing planned cost against recorded actual cost
- Finding missing data and threshold-based exceptions
- Ranking repeatable evidence for management review
The benefit: the rule, input, calculation and linked record can all be checked.
Why the distinction matters: generative AI can confidently present false content, while privacy and human over-reliance are further risks that need managing. For consequential operational decisions, evidence and verification still matter.
How it works
From live ERP data to a decision queue
The Intelligence Centre removes the friction between seeing a warning and finding the evidence behind it.
Read connected records
Use work orders, labour, materials, shipments, purchasing, capacity and delivery data already held in Fraction ERP.
Apply defined rules
Compare actuals with plans, equivalent periods, required fields and known operational thresholds.
Rank what matters
Order findings by severity and measured impact so managers know where to start.
Open the evidence
Move directly from the finding to the affected work order, operation, employee, BOM or purchase order.
Manufacturing intelligence
What the Intelligence Centre finds
It brings financial, production, purchasing, people and data-quality exceptions into one management view.
A ranked decision queue
Surfaces the most important profitability, productivity, delivery, purchasing and capacity findings first, with the measured impact and a route to investigate.
Margin and cost trends
Tracks gross margin, cost coverage, labour-cost accuracy, purchase-cost accuracy and productive utilisation over the selected period.
Work-order and operation variance
Compares planned and actual material and labour cost, flags missing baselines and ranks completed operations whose recorded time needs review.
Purchasing evidence
Finds missing prices, material costs outside plan and weighted purchase-price movements against the preceding equivalent period.
People and capacity signals
Separates productive, non-productive and unaccounted capacity, then flags utilisation or completed-work variances without labelling unaccounted time as idle time.
Data-confidence checks
Shows where missing cost reports, selling prices, material prices, purchase prices or capacity setup may limit the conclusions.
Built in, not bolted on
Why not just ask a free AI agent?
A general AI tool starts with whatever context a user remembers to provide. The Intelligence Centre starts with the operational records and relationships already inside the ERP.
| Operational requirement | External general-purpose AI agent | Fraction ERP Intelligence Centre |
|---|---|---|
| Starting context | Needs a prompt, upload, connector or copied data. | Uses connected ERP records in their existing workflow. |
| Repeatability | Wording and model behaviour can change the answer. | The same records and rules produce the same finding. |
| Evidence | Sources must be supplied, grounded and checked. | Findings link back to the affected operational record. |
| Friction | Users must leave the workflow or maintain an integration. | The decision queue is part of Fraction ERP. |
| Known limits | May return a persuasive answer even when context is incomplete. | Missing data is surfaced and the rule does not invent a cause. |
Honest intelligence
What rules-based ERP intelligence cannot do
Consistency is valuable, but rules are not magic. The Intelligence Centre only knows what Fraction ERP records and what its defined tests can establish.
That is why it presents a finding as evidence for investigation—not as an autonomous verdict.
It does not infer an unproven cause. A late order may involve materials, capacity, planning or another constraint; the rule identifies the affected work.
It does not learn a new rule by itself. Its strength is applying defined logic consistently, not discovering every hidden pattern.
It cannot repair poor source data. Instead, it identifies gaps that could make the analysis incomplete.
It does not replace management judgement. A person verifies the evidence, understands the context and decides what to do.
Frequently asked questions
AI, ERP and the Intelligence Centre
Is the Fraction ERP Intelligence Centre AI?
No. It is a deterministic, rules-based decision-support system. It applies defined tests to ERP data, ranks exceptions and links users to the evidence. It does not use a large language model or act as an autonomous AI agent.
Why call it an Intelligence Centre if it is not AI?
Intelligence is the useful interpretation of information. The centre goes beyond reporting by assessing known conditions, prioritising exceptions, explaining why they matter and directing users to the next record to review.
Why can rules-based ERP intelligence be more consistent than generative AI?
Rules are deterministic: the same inputs and rule produce the same result. Generative AI is probabilistic, so its response can vary with the prompt, context or model. For defined controls such as missing prices and cost thresholds, repeatability is often more useful than open-ended generation.
Does this mean AI has no place in ERP?
No. Generative AI can be valuable for language, documents, summaries and open-ended assistance. Fraction ERP uses rules where the operational test is known and the outcome needs to be transparent, repeatable and easy to verify.
Does the Intelligence Centre send our ERP data to an external AI service?
No external generative AI service is required to produce Intelligence Centre findings. The analysis is built into Fraction ERP and works from the records already held in the system.
Can the Intelligence Centre make decisions automatically?
No. It reduces the time needed to find and investigate exceptions, but the user remains responsible for checking the evidence and making the business decision.
Which Fraction ERP plan includes the Intelligence Centre?
The Intelligence Centre is included with the Fraction ERP Pro plan and inherited by the Moulding Edition.
Evidence before answers
Spend less time searching. Start with what needs attention.
See how the rules-based Intelligence Centre turns connected manufacturing data into a practical management decision queue.