Fix the workflow. Then automate it.
Most of the time an SME loses each week sits in how work moves between people, spreadsheets and systems. FPA maps it, simplifies it and automates the parts that should never have been manual.
In short: redesign the process first, automate the stable rule-based steps second, and add AI only where the work involves unstructured information or judgement that can be reviewed.
Signs a workflow needs redesign
- The same data is typed into more than one place
- Reports take days to assemble and are out of date when they arrive
- Work stalls waiting for one person's approval or knowledge
- Nobody can say exactly how many steps a process has
- Leads, requests or follow-ups fall through the gaps
How we redesign a workflow
- Map it as it really runs. We record every step, hand-off, wait and workaround, not the version in the SOP.
- Find the waste. Duplicate entry, unnecessary approvals, waiting time and manual checks.
- Design the simpler version. Fewer steps, clear ownership, one source of truth for the data.
- Decide what to automate. Rule-based steps go to automation, information-heavy steps may use AI, and decisions stay with people.
- Build, test and measure. We pilot with the team and track hours saved and errors avoided.
What we automate
From FPA's working list of SME use cases:
| Workflow | Typical approach | Where people stay involved |
|---|---|---|
| Recurring reports & dashboards | Scheduled data pulls, automatic refresh, AI-drafted commentary | Review and sign-off of the numbers |
| Data formatting & cleaning | Rules for structure, AI for messy text fields | Spot checks and exception handling |
| Lead capture & follow-up | Capture, enrich and route leads, then schedule follow-ups | The sales conversation |
| Rostering & scheduling | Constraint-based draft rosters | Final roster approval |
| Forms & applications | Guided assistants that check completeness | Approval decisions |
| Scheduled research | Web research pipelines that brief you weekly | Deciding what to act on |
Automation, AI assistant or AI agent?
Use automation when the rules are clear. Use an AI assistant when a person needs help with reading, drafting or finding information. Consider an AI agent only when the work is multi-step, spans several tools and needs some judgement, and even then with review points. We explain the difference in more detail on our AI agents page.
Related
Common questions
Why redesign a workflow before automating it?
Automating a broken process just makes the same problems happen faster. Redesigning first removes unnecessary steps, approvals and re-keying, so there is less to automate and what's left is stable enough to automate reliably. In many SMEs the redesign alone saves a large share of the time.
What is the difference between workflow automation and AI?
Workflow automation follows fixed rules: when X happens, do Y. It's predictable and cheap, and it suits structured, repeatable work. AI handles unstructured information, such as reading an email, drafting a reply or classifying a document, where the rules are hard to write down. Most good solutions use automation for the backbone and AI only for the steps that need it.
Which workflows are best to automate first in an SME?
Look for work that is frequent, rule-based, time-consuming and low-risk. Typical first candidates are recurring reports, moving data between systems, cleaning and formatting data, reminders and follow-ups, and routing requests to the right person.
What tools does FPA use?
We are tool-agnostic and prefer what you already pay for. Depending on the work that might be your existing Microsoft 365 or Google Workspace, an automation platform, a database and dashboard, or a custom AI workflow. We choose the lightest option that will hold up.
Which workflow is costing you the most?
The Snapshot helps you find it in two minutes.