Fixing the process before relying on technology or automation

Use ERP, AI and automation to strengthen sound merchant processes rather than making weak work happen faster.

7 questions answered

Questions answered on this page

01Will a new ERP system actually improve the business?
02Why hasn’t our ERP implementation solved our operational problems?
03How do I know whether a process problem is really a systems problem?
04Should we fix the process before introducing new technology?
05How can AI help a merchant business without creating more complexity?
06Where should automation be used in a merchant business?
07What should we automate and what should remain human-led?

Will a new ERP system actually improve the business?

A new ERP system can improve visibility, consistency and control when the business is clear about the processes and decisions it needs to support. It will not compensate for inaccurate product data, unclear ownership, weak stock discipline or branches following different rules without good reason. Start with the commercial and operational outcomes required, such as better availability, fewer pricing errors, stronger purchasing or reliable branch reporting. Map the critical work and decide which variation is necessary. Then test whether the proposed system supports that model and whether people can adopt it. The value comes from better work, information and decisions. Replacing software without changing the conditions around it often creates an expensive version of the same problem.

Why hasn’t our ERP implementation solved our operational problems?

ERP projects disappoint when implementation is treated mainly as a technology installation. Existing problems are transferred into new workflows, data is migrated without enough cleansing and employees invent workarounds when the configured process does not fit real branch activity. Training may explain which buttons to press without clarifying standards, ownership or the reason for change. Leaders then blame the system for poor adoption or blame users for defects in the design. Review where the intended process differs from daily practice and why. Check data, decision rights, local exceptions and measures as well as configuration. The system can support control, but management still has to define the work, enforce sensible standards and correct problems that sit outside the software.

How do I know whether a process problem is really a systems problem?

Trace the failure from the customer or business impact back through the work. Ask whether people understand the expected process, have the authority and capacity to follow it, and receive accurate information at the right time. Compare branches or teams using the same system. If one performs reliably and another does not, the software may not be the main constraint. A systems problem is more likely when the required workflow cannot be completed, data cannot move accurately, controls are missing or the system creates unavoidable duplication. A process problem is more likely when steps, ownership or standards are unclear. Many cases contain both. Define the failure precisely before buying technology or redesigning work around the wrong diagnosis.

Should we fix the process before introducing new technology?

Yes, to the point where the business understands the intended outcome, essential steps, ownership, information and valid exceptions. Automating an unclear process can make mistakes faster and harder to see. That does not mean every manual step must be perfected before technology is considered. System capability may allow the process to be simplified, and design should be iterative. Begin by removing obvious duplication, resolving conflicting rules and agreeing the minimum standard across branches. Test the future process with the people who use it and the customers or functions affected. Then configure technology around a deliberate way of working. The sequence is understand, simplify, design and test, rather than copying the current process into a new platform unchanged.

How can AI help a merchant business without creating more complexity?

AI can help with focused tasks where the inputs, expected output and human responsibility are clear. Potential merchant uses include summarising information, drafting routine communication, identifying patterns in customer or stock data, supporting knowledge search and reducing repetitive administration. Begin with a contained problem and check data quality, confidentiality, accuracy and the consequence of error. Keep a named person responsible for reviewing outputs and decisions. Avoid adding separate tools that create duplicate data, unclear ownership or uncontrolled customer communication. AI should remove friction from useful work or improve the evidence available to people. It should not become another layer that staff must manage because the business adopted it before deciding what problem it was meant to solve.

Where should automation be used in a merchant business?

Use automation where work is repetitive, rules are stable, data is reliable and exceptions can be identified safely. In a merchant this may include routine data transfer, standard reports, reorder prompts, invoice matching, reminders or administrative steps around orders and delivery. Prioritise areas with meaningful volume, avoidable error or delay rather than automating a task simply because it is possible. Measure the complete effect, including maintenance, exception handling and the work moved elsewhere. Keep controls that reveal when automation fails or data changes. A successful automation reduces total effort and improves consistency without hiding risk. If employees spend more time correcting exceptions or reconciling systems, the process has been moved rather than improved.

What should we automate and what should remain human-led?

Automate stable, rules-based work and keep human judgement where context, relationships, negotiation or material risk matters. Pricing routines, stock prompts and credit workflows can support decisions, but unusual customer circumstances or significant commercial exceptions may still need experienced review. Customer communication can be drafted or triggered automatically, while sensitive complaints, performance conversations and important account decisions require personal ownership. Define the boundary through value and consequence. Ask how easily an error can be detected and reversed, and who remains accountable. Human-led does not mean informal or inefficient, and automated does not mean unsupervised. The strongest design gives people better information and removes repetitive effort while keeping clear responsibility for decisions that shape trust, margin and risk.

What you may be seeing

A major system investment has gone live, but branches still use spreadsheets, manual notes or local workarounds.
The same stock, pricing, order or reporting errors continue under a different screen and workflow.
Technology projects begin with product demonstrations before the business has agreed the problem and desired outcome.
Automation saves time in one team but creates exceptions, reconciliation or customer problems elsewhere.
AI tools are being tried without clear data rules, ownership or a way to check the accuracy of their output.

What good looks like

The business defines the required outcome and understands the current process before selecting or configuring technology. Essential standards, ownership, information and exceptions are agreed with the people who perform and manage the work. Data is cleaned and governed as part of the change. System design is tested against real branch, stock, customer and supplier situations rather than an ideal demonstration. Automation is used for stable, repetitive work where errors can be detected and controlled. Human judgement remains clear for commercial exceptions, relationships and higher-risk decisions. Leaders measure the overall result, including adoption, rework and downstream impact. Technology strengthens a deliberate operating model instead of becoming a substitute for one.

What may be happening underneath

Process clarity: The business has not agreed the essential steps, outcome, ownership and valid exceptions before configuration.
Data quality: Product, customer, pricing, stock or supplier data is inaccurate, duplicated or governed inconsistently.
Local variation: Branches follow different practices, but necessary flexibility has not been separated from avoidable inconsistency.
Implementation ownership: The project is treated as the responsibility of IT or the supplier rather than the leaders who own the work.
Adoption: Training covers system use without reinforcing the commercial reason, standard and management expectation behind it.
Automation risk: Efficiency is measured at the automated step while exceptions, controls and downstream work remain invisible.

Questions worth asking

01What customer, commercial or operating result should this technology improve?
02Where does the current process first fail, and is that failure caused by the system, the work or both?
03Which branch differences are necessary and which represent uncontrolled variation?
04Is the underlying data accurate enough for the system or automation to use safely?
05Which decisions require human judgement, and who remains accountable when technology supports them?

Where to go next

Identify whether the weakness sits in the process, information, ownership or system before investing further. The Business Control Score helps expose wider control gaps, while BGC business systems resources support clearer processes and more disciplined implementation.

Take the Business Control Score Explore Business Systems toolkits