Workflow automation: automate the right process, not just the task
Automating a manual task is easy. Knowing which workflow to change, where the real delay sits and which technology fits is harder. SEMANU maps your processes, exceptions, systems, data and handover moments before you choose automation software or start an implementation.
- No software sales
- No vendor commissions
- Process before technology
- Advice for Belgium, the Netherlands and DACH
Do not start with the tool.
Start with the process.
Automation projects often start with a technology question. Can we do this with AI? Could we use a workflow platform for this? Should we build something custom? Those questions come too early. The first question is: what exactly happens today, why does it happen that way and where do time, information or quality get lost?
Take an employee who copies data from e-mails into the ERP. That can have several causes: the customer supplies incomplete information, two systems share no data, prices are calculated outside the ERP, exceptions live only in one person's head, or an approval is needed before the order is created. Automate only the copying and you may well be automating a bad process.
A process problem is not the same as a technology problem. That distinction decides whether automation creates value or merely adds speed to something that should have run differently.
The order that works
- Process
- Friction
- Cause
- Requirements
- Solution direction
- Technology
The order that goes wrong
- Tool
- Implement
- Hope the process fits
What is workflow automation?
Workflow automation is the automatic execution, handover or support of recurring steps within a business process.
A workflow usually consists of a series of steps that follow one another. Automation can support a single step or connect the entire chain.
A workflow in five steps
- Request
- Validation
- Approval
- Processing
- Follow-up
Quotation request
Website form → CRM → qualification → costing → quote → approval → follow-up.
Order
E-mail or webshop → validation → ERP → stock → planning → delivery → invoicing.
Document
PDF → extract data → validate → approve → register → archive.
Service
Customer request → classification → scheduling → technician → work order → ERP → invoice.
AI is a possible component of workflow automation, not its definition. Read more about AI assistants when a step has to interpret language, documents or unstructured information.
When does workflow automation become interesting?
Sales automation
Interesting when sales people lose a lot of time before or after the actual sales conversation.
- Handling leads manually and re-entering data in the CRM
- Price calculations outside the systems, quotes rebuilt from scratch each time
- Approval by e-mail, follow-up that gets forgotten
- Orders handed over to operations manually
Administrative automation
Interesting when employees mainly process information instead of making decisions.
- Copying data from e-mails and checking PDFs
- Maintaining Excel lists, registering orders, matching invoices
- Classifying documents, consolidating reports, synchronising data
Operations automation
Interesting when a lot of coordination is needed between people, locations and systems.
- Planning, dispatching, work orders and production orders
- Service interventions, stock and internal approval
- Status updates and order handover between departments
AI automation
Interesting when a process contains information that classic rules struggle to interpret.
- Free text in e-mails, documents, quotes and contracts
- Technical questions, support questions and knowledge bases
- Classification, summarisation and information extraction
Not every manual process should be automated
A good automation candidate usually has a number of recognisable properties: it recurs regularly, follows a recognisable pattern, costs meaningful time, causes errors with a cost, moves information between systems or people, has understandable exceptions, runs at sufficient volume and delivers measurable business value.
But context still matters. A process that occurs a few times a month, requires a lot of human judgement and changes constantly can be more expensive to automate than to do by hand. That is not a technical problem but an economic one.
SEMANU does not first ask whether something can be automated. The first question is: is this process worth automating?
What makes a process automatable
- Recurs regularly
- Sufficient volume
- Understandable exceptions
- Measurable business value
Research note: qualitative interview studies on RPA report workload reduction, time savings, fewer errors and room for more valuable work. The same studies stress that organisational infrastructure, responsibilities, knowledge and system landscape co-determine what is feasible. Source: Journal of Organizational Computing and Electronic Commerce.
The biggest gain often sits between systems, not inside one system
Inefficiency rarely arises inside one screen. It arises at handover moments: from website to e-mail to administration to CRM to ERP, or from sales to operations to production to finance.
Every handover raises the same questions. Does information need re-entering? Does the structure change? Is something missing? Is a check required? Does waiting time appear? What happens with an exception? And does the next system know what the previous one did?
Before
- Customer fills in a form
- An employee receives an e-mail
- Data goes into Excel
- A colleague checks
- Data goes into the ERP
- Quote in Word, PDF by e-mail
- Manual follow-up
Possible after analysis
- Request comes in
- Automatic validation
- CRM and ERP receive the same data
- Costing
- Approval where needed
- Quote
- Automated follow-up, a person steps in for exceptions
Automation does not necessarily take people out of the process. It moves human attention to the steps where judgement is needed.
When is a company ready for workflow automation?
Process
Do we know how the process actually runs?
- Do we know the exceptions?
- Do we know where waiting time appears?
- Is it clear who owns it?
People
Who does the work today and who decides?
- Who has to trust the new way of working?
- What changes in roles and responsibilities?
Data and systems
Where does the information come from and is it reliable?
- Are there APIs or integration options?
- How do authorisations work?
Business case
How much volume, how much time and where do the errors arise?
- What concretely has to improve?
- How do we measure the result?
Workflow automation is not a software category that always needs the same tool
Use existing software better
Your ERP, CRM or accounting package may already contain the functionality. That is often the fastest and cheapest route.
More about ERP softwareIntegration
Letting two existing systems talk to each other directly, so data no longer has to be transferred by hand.
Workflow platform
Low-code, no-code or orchestration software that ties steps, rules and approvals together.
AI
When documents, language or unstructured information must be interpreted before the workflow can continue.
More about AI assistantsCustom software
When the process is highly company-specific or forms a competitive difference that standard software does not cover.
More about custom softwareChange the process without new software
Sometimes a different way of working, a clearer form or a removed check is the cheapest and best solution.
The analysis does not have to prove that automation is needed. It has to determine which solution makes the most economic and organisational sense, including the option of building nothing new.
Automation is a business project, not an IT trick
Qualitative research into business automation and RPA keeps arriving at the same conclusion: success depends on human, organisational and technical factors together, with change management as the overarching factor. Research among smaller companies on AI adoption confirms the same socio-technical picture.
That is why SEMANU does not analyse only the technology. We look at the process, the people, the data, the systems, the business value and the governance together. Anyone who examines only technical feasibility discovers the organisational side after signing.
What SEMANU analyses
- Process
- People
- Data
- Systems
- Business value
- Governance
RPA success factors based on expert interviews and a literature study: ScienceDirect.
Socio-technical adoption across 27 smaller companies: Information Systems Management.
Interviews with CIOs and CTOs on AI adoption in firms: OECD.
From “this costs us a lot of time” to a well-founded automation decision
Understand the process
Triggers, activities, roles, systems, data, handover moments, exceptions and checks, as the process really runs today.
Deliverable: current-state process mapDetermine friction and value
Lost time, waiting time, repetitive work, errors, duplicated work, bottlenecks, dependencies and scaling problems, with an economic weighting.
Deliverable: automation opportunity mapDesign the future state
What disappears, what can run automatically, what stays human, what gets integrated and what becomes an exception flow.
Deliverable: future-state workflow and requirementsSelect the solution
Only here do we determine whether the direction is ERP, integration, a workflow platform, AI, custom software, a configurator or a process change. If vendors are needed, we compare them on identical scope.
Deliverable: solution direction and vendor briefingWorkflow Automation builds on the existing methodology: business analysis for software projects, independent software and vendor selection, independent project oversight and software strategy.
What does workflow automation look like in practice?
Quotation process
Today: request, sales person hunts for information, costing in Excel, quote in Word, manager approval, PDF, CRM update and follow-up mail.
- Possible: automatic data validation, costing, approval only for exceptions, quote, CRM and follow-up workflow.
- Value: faster response time, less administrative work, fewer handover errors.
Order processing
Today: order by e-mail, back office interprets the PDF, enters data, checks item codes, books it in the ERP and schedules it.
- Possible: document extraction, validation rules, human review of exceptions, ERP and planning.
- AI can become part of this once orders arrive unstructured.
Field service
Customer report, ticket, classification, planner, technician, work order, ERP and invoice: many administrative handovers between people and systems.
- Automation reduces those handovers; planning and exceptions remain human work.
Configurable products
Product configuration, technical validation, price, quote, CRM, ERP and production together form a single sales workflow.
- That can lead to CPQ or a 3D configurator, but only once the workflow is clear.
We have no stake in how much software you buy
An automation vendor earns when you use their platform. A development agency earns when something gets built. An implementation partner earns from the implementation. SEMANU sits before that decision: no platform, no licences, no development hours.
So an analysis may equally conclude that the existing software suffices, that one small integration is enough, that a process should be simplified first, that AI is needlessly complex, that custom software is not profitable or that automation has no positive business case for now.
That is not a failed analysis. That is the value of an independent analysis.
Frequently asked questions about workflow automation
What is workflow automation?
Workflow automation means automatically executing, passing on or supporting recurring steps within a business process. That ranges from data synchronisation between two systems to a complete flow of request, validation, approval and processing.
Which processes can you automate?
Processes with enough volume, repetition and predictable rules: order processing, quotation flows, document handling, administrative checks, approvals, planning and data transfer between systems. Strongly judgement-based processes call for a hybrid approach in which people keep the decision.
When is a process not suitable for automation?
When it happens rarely, changes constantly, is poorly understood or when nearly every case is an exception. Simplifying the process then usually pays off more than software.
Do we need AI for workflow automation?
Not necessarily. Many workflows can be automated with rules, integrations and existing software. AI becomes interesting mainly when a step has to interpret unstructured information: e-mails, documents, free text or images.
What is the difference between workflow automation and AI automation?
Workflow automation automates steps and handovers in a process. AI automation uses AI to interpret a step or prepare a decision. AI can be part of a workflow, but it does not have to be.
Which software do you need for workflow automation?
That depends on the process and the systems already in place. Sometimes functionality in the ERP or CRM is enough, sometimes an integration, a workflow platform, AI or custom software. SEMANU picks no tool before the requirements are clear.
What does workflow automation cost?
Cost depends on process complexity, integrations, volumes, exceptions, data quality and the chosen technology. That is why we first establish the business value and set it against implementation and running costs.
How do you get started with workflow automation?
Map one concrete process end to end: steps, roles, systems, data, exceptions and waiting times. Then determine where the biggest friction and economic value sit, and only then which technology fits.
Can workflow automation replace employees?
The aim is usually to reduce the repetitive parts of roles, not to remove a whole job. People remain needed for exceptions, decision-making, client relationships and context.
Is my company ready for workflow automation?
You need at least insight into the process, the systems involved and the intended business impact. The free Workflow Automation Readiness Check assesses that in ten questions.
You might also be interested in this
SEMANU is an independent software advisory firm for SMBs: we sell no software and receive no vendor commissions. Read how we work as an independent software advisory firm.