That's not a warning. That's what the data says. According to Nintex research, the processes most frequently targeted for automation are often the wrong ones. And McKinsey consistently finds one pattern in successful automation projects: end-to-end workflows are redesigned before any tool is selected.

So the question isn't "which tool should we use?" The question should be "which process, why, and how will we measure it?"

This post covers the five questions to clarify before starting an automation project. Each one is designed to reduce the cost of a wrong start.

Question 1: Why Is This Process Slow or Error-Prone?

The most common sentence we hear in automation conversations: "Reporting takes too long. We want to automate it."

That's a symptom, not a diagnosis.

You need to ask why reporting takes so long. Is the data scattered across different systems? Is the team manually consolidating it every time? Are formats inconsistent? Are approval steps creating a bottleneck?

Each answer points to a different solution:

• Scattered data is an integration problem

• Manual consolidation is a data infrastructure problem

• Approval bottlenecks are a workflow design problem

Automation built without understanding the real cause doesn't eliminate the surface symptom. It hides the underlying problem. Six months later, the system runs but the problem returns in a different form.

Question 2: Are You Measuring This Process?

A process that can't be measured can't be automated correctly.

Salesforce's 2026 Productivity Gap research makes a striking point: employees spend a significant portion of their time on manual tasks, but most companies have no idea exactly how much time is going where. Pipeline updates, client communications, data entry, reporting — each one consumes time, but how many hours is never tracked.

Before starting an automation project, you need to know two numbers.

• Current state: how many hours or steps does this process currently take, what is the error rate, what does it cost?

• Target state: what do you expect those numbers to be after automation?

Without these two numbers, you can't tell whether a project succeeded or failed. And you can't improve what you can't measure.

Question 3: Is This Process Consistent and Repeatable?

Automation works on processes that follow the same steps every time. It creates problems when applied to processes that require judgment, are full of exceptions, or vary significantly from one case to the next.

There's a practical test for this: can you explain this process to a new employee with a written set of instructions?

• If you can say "yes, follow these steps" — it's a strong automation candidate

• If you say "it depends, you'll figure it out with experience" — that process is not ready for automation yet

Nintex's research makes this distinction clear: high-volume, repetitive, rules-based tasks deliver the highest return from automation. Processes that require complex human judgment need complementary technology — automation alone isn't enough.

Question 4: Is Your Data Ready for This?

Gartner predicts that 60% of AI projects without AI-ready data infrastructure will be abandoned by 2026.

Data readiness is the most underestimated dimension of automation projects.

Ask yourself:

• Is data across different departments connected, or is every team working in their own spreadsheet?

• Does the sales team's CRM talk to the finance team's accounting system and the operations team's project management tool?

• Is the existing data clean, consistent, and current?

If you're answering "no" or "not sure" to these questions, you need to address your data infrastructure before starting the automation project. Skipping this step is the equivalent of building on sand instead of on solid ground.

It's not a coincidence that Informatica's 2025 research identified data quality and readiness as the number one cause of AI and automation failures at 43%. This problem typically surfaces after the project has started, and going back is far more expensive than starting right.

Question 5: Will Your Team Actually Use This System?

This is the least discussed reason for automation failure.

The system gets built. It gets tested. It goes live. Six months later, no one is using it. The team has gone back to the old way.

One of the patterns MIT NANDA consistently found in successful projects: success rates increase dramatically when the line managers who actually run the process take ownership, not a centralized AI lab. The people who use the system should have a say in how it works.

Before building, ask:

• Does this system fit into the existing workflow, or does it require a fundamental change in how people work?

• Will training be provided?

• Who will provide support during the transition period?

• How will problems be reported when they come up?

If these questions are answered before the build, most of the resistance that comes after go-live disappears.

What Do These Five Questions Add Up To?

These questions aren't a checklist. They're a summary of an approach.

Automation produces value when it's designed by someone who understands the problem. A tool selected without that understanding delivers a neutral result at best, and accelerates the breakdown of the existing process at worst.

The consistent pattern in successful projects: a concrete problem, a measurable goal, clean data, a repeatable process, and an adoption plan. All five need to be on the table from the start.

If some answers are still unclear when you ask these questions, that's not a risk signal. That's a starting point. Clarifying the ambiguity is exactly what the analysis phase is for.

Where Do You Start?

The right starting point is not a tool selection. It's an operational analysis.

Which departments spend time on what tasks? Which processes are repeatable and measurable? Is the data infrastructure ready to support automation? Is the team open to change?

The answers to these questions determine which processes get automated and what kind of system gets built. A selection made without that analysis carries the risk of joining the 80% failure group.

SOURCES

Nintex Process Automation Research (2025)  ·  Salesforce "The Productivity Gap" (2026)  ·  McKinsey & Company Operations Survey (2025)  ·  Gartner Data & Analytics Research (2025)  ·  Informatica CDO Insights (2025)  ·  MIT NANDA Initiative — State of AI in Business (2025)