A McKinsey study found that only 16% of enterprise automation initiatives deliver their expected ROI within the first year. The problem is rarely the technology. It is almost always target selection — teams automate the wrong workflows and wonder why the results disappoint.
The companies that get automation right do not start with the flashiest process or the one the CEO mentions in a meeting. They start with a disciplined audit of where time, money, and mistakes actually accumulate. This post walks you through exactly how to do that.
Why target selection matters more than tooling
There is no shortage of automation tools. Between AI agents, RPA platforms, workflow orchestrators, and plain old scripts, you can automate almost anything. The constraint is not capability — it is focus.
Every automation project carries setup cost: scoping, building, testing, training, and maintaining. If you spend that budget on a workflow that runs ten times a month, you will wait years to break even. If you spend it on a workflow that runs 500 times a day with a 4% error rate, you might recoup the investment in weeks.
The skill is not building the automation. The skill is choosing where to point it.
The five signs of a good automation candidate
Before you run a formal audit, learn to recognise the patterns. Strong automation candidates almost always share these traits:
- High volume — The task happens frequently. Daily is good. Hourly is better. If it only occurs once a quarter, it is unlikely to justify the investment.
- Low judgment — The decision logic is mostly rule-based. "If X, then Y" workflows automate cleanly. "Use your best judgment" workflows do not.
- Structured inputs and outputs — The data involved is predictable. Forms, spreadsheets, database records, and API payloads are friendly to automation. Free-text emails and ad hoc Slack threads are hostile.
- High error rate or rework — Humans make mistakes on repetitive tasks. If a process has a known error rate or frequently needs corrections, automation removes the variance.
- Painful bottlenecks — The task creates a queue. Other people or systems are waiting on it. Delays compound downstream.
If a workflow hits three or more of these, it belongs on your shortlist. If it hits all five, stop reading and go automate it.
How to run a workflow audit
A workflow audit is not a six-month consulting engagement. It is a focused sprint — one to two weeks — that produces a ranked list of automation candidates. Here is how to run one.
Step 1: Observe
Spend the first few days collecting workflows. You are not evaluating yet, just cataloguing. Good sources:
- Shadow sessions — Sit with the people who do the work. Watch them click, copy-paste, switch tabs, and wait. The most automatable workflows are often so routine that the people doing them forget to mention them.
- Ticket and request queues — Pull data from your help desk, ops inbox, or project management tool. What requests show up repeatedly with nearly identical responses?
- Handoff points — Map where work passes between people, teams, or systems. Every handoff is a potential bottleneck, a potential error source, and a potential automation point.
- Time logs — If your team tracks time (even loosely), look for tasks that consume disproportionate hours relative to their complexity.
Aim to identify 15-30 candidate workflows. Do not filter yet.
Step 2: Score
Now evaluate each workflow against a consistent rubric. We use a simple five-factor model, each scored 1-5:
| Factor | 1 (Low) | 3 (Medium) | 5 (High) |
|---|---|---|---|
| Volume | Monthly or less | Weekly | Multiple times daily |
| Rule-based logic | Requires deep expertise and nuance | Mix of rules and judgment | Pure if/then logic |
| Error rate | Rarely goes wrong | Occasional mistakes | Frequent errors or rework |
| Cost of delay | No one notices if it waits | Mild inconvenience | Blocks revenue or customers |
| Data structure | Unstructured, ambiguous inputs | Semi-structured | Clean, structured data |
Each workflow gets a score from 5 to 25. Simple addition. No weighted averages, no elaborate models. You want something your ops team can fill out in a spreadsheet during a single working session.
Step 3: Rank and cut
Sort your list by total score. Draw a line at the top five. These are your candidates for deeper investigation.
Before committing, apply two sanity checks:
- Integration feasibility — Can you actually connect to the systems involved? A workflow scored 24/25 is worthless if the core system has no API and no export function.
- Stakeholder willingness — Will the people who own this process actually adopt an automated version? The best-scoring workflow in the audit means nothing if the team resists the change.
Drop any candidate that fails either check. Move the next one up.
A real-world example
One of our clients — a mid-size logistics company — ran this audit across their operations team. Here is what their top five looked like:
| Workflow | Volume | Rules | Errors | Delay cost | Structure | Total |
|---|---|---|---|---|---|---|
| Invoice data entry | 5 | 5 | 4 | 3 | 5 | 22 |
| Carrier rate lookups | 5 | 4 | 3 | 5 | 4 | 21 |
| Customer onboarding forms | 3 | 4 | 4 | 4 | 5 | 20 |
| Weekly KPI report assembly | 2 | 5 | 3 | 2 | 5 | 17 |
| Exception escalation routing | 4 | 2 | 3 | 4 | 2 | 15 |
The team initially wanted to automate exception escalation routing because it felt like their biggest pain point. But the audit showed it was high-judgment and low-structure — a poor automation candidate. Invoice data entry scored highest and had the clearest path to implementation. They started there, saved roughly 30 hours per week within the first month, and used that momentum (and budget headroom) to tackle carrier rate lookups next.
The audit took four days. The first automation was live within three weeks.
Common traps to avoid
Even with a scoring framework, teams fall into predictable traps:
Automating for prestige, not impact. The executive team gets excited about an AI-powered customer insights dashboard. Meanwhile, someone in accounts is manually copying 200 invoice line items a day. The dashboard is interesting. The data entry is where the ROI lives.
Automating broken processes. If a workflow is inefficient, automating it just makes it inefficiently fast. Fix the process first, then automate the fixed version. Automation should encode good practice, not cement bad ones.
Over-scoping the first project. Your first automation should be boringly simple. A single-step data extraction. A form-to-database pipeline. A notification trigger. Win fast, build trust, then expand scope.
Ignoring maintenance cost. Every automation needs ongoing care — API changes, edge cases, schema updates. Factor maintenance into your ROI calculation. A complex automation with high maintenance overhead can easily turn negative.
What to do after you have your shortlist
Once you have your ranked list and your top candidate selected, the next steps are straightforward:
- Document the current workflow — Map every step, input, output, and exception. You cannot automate what you have not specified.
- Define success metrics — Before you build anything, decide what "working" looks like. Time saved per week? Error rate reduction? Throughput increase? Pick one or two numbers and commit to measuring them.
- Build a minimum viable automation — Automate the happy path first. Handle the 80% case. Route exceptions to humans. You can expand coverage later.
- Measure and iterate — Compare your success metrics after 30 days. If the numbers are good, expand. If they are not, diagnose whether the issue is the automation itself or the workflow selection.
Key takeaways
- Audit before you build. Spend a few days scoring workflows on volume, judgment, error rate, delay cost, and data structure. The highest-scoring candidates are almost never the ones leadership assumes.
- Start boring. The best first automation is the one nobody talks about at the all-hands meeting — high-volume, rule-based, and quietly consuming hours every week.
- Measure from day one. Define your success metrics before implementation. Automation without measurement is just hope with extra steps.
If you are thinking about automation but not sure where to start, we can help you run this audit. Get in touch — we will find the workflows worth your investment.