---
title: "AI Masterclass: Recognizing automation potential - which steps in process X really eat up time"
description: How do you recognize which steps in a process really eat up time? This practical guide shows you how to use Google Gemini, ChatGPT and Claude to identify, prioritize and cleanly implement automation potential.
---

[Digital Change ](https://blog.bloola.com/digital-change)

# [AI Masterclass: Recognizing automation potential - which steps in process X really eat up time](https://blog.bloola.com/digital-change/recognizing-automation-potential-which-steps-in-process-x-really-eat-up-time)

 Written by [Lars-Thorsten Sudmann](https://blog.bloola.com/digital-change/author/lars-thorsten-sudmann) | Mar 16, 2026 11:36:45 AM

Your team says: "The process is running."

And yet everything takes too long.

Requests are left pending, handovers drag on, approvals are double-checked, data is copied from one tool to the next. No one sees the total effort because the time wasters are rarely in a single big problem - but in many small, recurring steps.

This is exactly where AI comes in handy: not to immediately replace the entire process, but to systematically visualize **which steps in process X eat up time, why they eat up time and what is worth automating first**.

In this module, I will show you how you can use **Google Gemini, ChatGPT and Anthropic Claude** for this - not abstractly, but as a concrete work process for practical use. Google describes Gemini as an AI assistant for writing, planning, researching and learning; OpenAI describes ChatGPT as a dialog-oriented system with contextual responses and - depending on the tariff/setting - additional tools such as web search; Anthropic presents Claude as a working environment for helpful conversations and professional knowledge work.

## **1. the real problem**

Let's take a typical process X:

Lead processing, quote generation, invoice approval, recruiting or content approvals, for example.

On paper, the process looks clean:

Receipt → Check → Query → Processing → Release → Dispatch → Documentation

In reality, it looks more like this:

- Information is missing
- Data is transferred manually
- Queries are sent by email, chat and telephone
- Approvals wait for individual people
- Documentation is "quickly" added at the end
- Special cases interrupt the standard process

The result: the process appears active, but is actually full of micro-delays.

## **2 Why the problem remains**

Many companies approach the issue incorrectly.

They immediately look for "the best AI tool" before they have even properly broken down the process. Or they automate the loudest pain point, even though most of the time is lost elsewhere.

Typical mistakes are:

- Tool overload instead of process clarity
- Process descriptions that are too rough
- Lack of separation between processing time and waiting time
- No prioritization according to effort, repetition and standardizability
- AI is used for text production, but not for process diagnostics

The result: You get lots of ideas, but no reliable basis for decision-making.

## **3. the concrete solution: the 4-step automation potential check**

Instead of automating immediately, proceed in four clear steps.

### **Step 1: Break down process X into individual steps**

Not "Create offer", but:

1. View request
2. Check information in the CRM
3. Request missing data from the customer
4. Internal consultation with specialist department
5. Search for quotation template
6. Adjust figures
7. Obtain approval
8. Export PDF
9. Formulate mail
10. Document dispatch

Only at this level does it become visible where time is lost.

### **Step 2: Record four values per step**

Document each step:

- **Processing time**: How long is someone actively working on it?
- **Waiting time**: How long does the step take to move on?
- **Degree of repetition**: How often does this happen per week/month?
- **Standardizability**: Does the step usually follow fixed rules?

### **Step 3: Use AI for time wasters instead of "overall automation"**

The better question to ask AI is not:

"What can be automated?"

Rather:

"Which individual steps cause the greatest loss of time - and why?"

This is exactly what Gemini, ChatGPT and Claude are very well suited for. All three can be used as structured sparring partners, but with slightly different strengths in practical application. For Gemini, Google documents chat usage on the web, response verification and connected apps, among other things; for ChatGPT, OpenAI describes dialogic usage via the input box as well as additional tools depending on access; Anthropic provides learning and work guides for Claude to structure prompts, analysis and workflows.

### **Step 4: Prioritize results**

Then prioritize not by "cool", but by:

**Time loss × Frequency × Automatability**

This is the basis for real quick wins.

## **4 How to use Google Gemini, ChatGPT and Claude specifically in comparison**

Now it gets practical.

You should**not** let  the three tools compete blindly against each other, but use them consciously for different perspectives.

### **A. Google Gemini: good for initial structure, clustering and visualizing patterns**

Gemini is useful if you want to quickly create an initial structure from a messy, real process input. Official help pages describe how to use it in web chat, response checking and functions such as Connected Apps and Gems, which enable additional contexts and workflows depending on the account.

#### **This is how you proceed**

You don't give Gemini a "big strategy task" first, but a raw process dump:

- List of all process steps
- Roles involved
- tools
- typical queries
- media breaks
- Estimated times

Then you prompt like this, for example:

 

**Prompt to Gemini:**

"Analyze the following process. For each step, mark whether the loss of time is mainly due to manual data entry, waiting for approvals, unclear information, media breaks or coordination. Summarize similar bottlenecks and name the three biggest time wasters."

#### **What Gemini is often good at in this use case**

- Group similar problems together
- make initial patterns visible
- Organize scattered process information into categories
- Provide an understandable initial overview

#### **What I would use Gemini for here**

As an **initial analysis and structuring run**.

Not for the final decision, but for quick summarization:

"Where is there likely to be the most friction?"

### **B. ChatGPT: strong for prioritization, implementation logic and next automation steps**

ChatGPT is particularly suitable if you want to derive concrete measures from a process analysis. OpenAI describes ChatGPT as a dialog-based system; depending on the tariff and settings, additional tools such as web search and in-depth research are also mentioned in the Help Center.

#### **This is how you proceed**

Ideally, you should already give ChatGPT a structured table, for example:

| **Step** | **Role** | **Processing time** | **Waiting time** | **Frequency** | **Error rate** | **Can be standardized?** |
| --- | --- | --- | --- | --- | --- | --- |
| View request | Sales and distribution | 8 min | 0 | 40/week | low | Yes |
| Ask for data | Distribution | 6 min | 24 h | 25/week | medium | partial |
| Obtain approval | Sales Lead | 4 min | 18 h | 18/week | low | Yes |
| CRM documentation | Sales | 7 min | 0 | 40/week | medium | Yes |

Then you prompt like this, for example:

**Prompt to ChatGPT:**

"Evaluate these process steps according to time loss, frequency and automatability. Create a priority list with quick wins, medium-term automations and steps that should remain manual for the time being. Justify each recommendation."

#### **What ChatGPT is often good at in this use case**

- Build a prioritization logic from raw data
- Separate quick wins from more complex projects
- Formulate implementation options
- go through several iterations with you

#### **What I would use ChatGPT for here**

As a **decision and action workbench**.

So for the question:

"What do we automate first - and how?"

### **C. Anthropic Claude: strong for depth, clarity and working out hidden process risks**

Claude is particularly helpful if you not only want to find superficial time wasters, but also want to understand the **thinking errors in process design** . Anthropic provides learning materials and practical guidelines for using Claude for professional workflows.

#### **This is how you proceed**

You give Claude the process plus context:

- What goals the process has
- which special cases occur
- which rules or approvals are non-negotiable
- where errors become expensive
- what information is often missing

Then you prompt like this, for example:

**Prompt to Claude:**

"Examine this process not just for time wasters, but for structural causes of delays. Show which steps generate unnecessary queries, where roles are unclear, which approvals are only seemingly necessary and which automations would create risks."

#### **What Claude is often good at in this use case**

- Better differentiation between symptom and cause
- Cleaner formulations for process logic
- Indications of exceptions, uncertainties and governance
- Clearer analysis of dependencies

#### **What I would use Claude for here**

As a **quality and depth check**.

So for the question:

"Where would we accelerate the wrong step with hasty automation?"

## **5. the best practice process: not either/or, but in sequence**

The most sensible use is often this sequence

### **1. Gemini for the first pattern scan**

You pre-sort the unstructured process and cluster bottlenecks.

### **2. ChatGPT for prioritization and implementation plan**

You transform these findings into a robust roadmap.

### **3. Claude for in-depth check and risks**

You check whether the planned automations are really viable from a technical and organizational perspective.

This way, you don't get a pretty AI answer, but a **better basis for decision-making**.

## **6th practical example: medium-sized company in the quotation process**

A B2B company wants to speed up its quotation process.

### **Before**

- Inquiries come in by email
- Data is manually transferred to the CRM
- Product information is collected from several documents
- Queries to technology take time
- Approvals are made by e-mail
- Dispatch and documentation are done manually by Sales

The official perception:

"It just takes time to prepare an offer."

The real cause:

Not one step, but six small friction losses.

### **Analysis with AI**

**With Gemini** becomes visible:

The biggest patterns are missing input data, duplicate data entry and waiting times for approvals.

**With ChatGPT** , this results in prioritization:

1. standardized request form
2. Automatic CRM pre-filling
3. Quotation template with variable modules
4. Automate approval rules according to threshold value
5. Link dispatch and CRM documentation

**With Claude** it then becomes clear:

The technical queries are not just a time problem, but a standardization problem because product knowledge is not available in a cleanly structured form. In addition, some approvals have grown historically and are no longer technically necessary.

### **Afterwards**

- Structured incoming requests
- fewer queries
- less copy-paste
- Faster approvals
- Clear separation between standard cases and special cases
- initial partial automation instead of full automation fantasy

The result is usually not "100% automatic", but **30 to 50% less friction in the most relevant steps**. Whether this is achievable naturally depends heavily on data quality, process discipline and system landscape; however, the prioritization method increases the chance of starting in the right places first. This categorization is a practical derivation from the product and usage options of the three systems described above.

## **7 Immediately implementable steps**

If you want to test this in your company tomorrow, proceed as follows:

### **1. choose a real process X**

Not the biggest, but a frequent process with clear handovers.

### **2. break it down into 10 to 20 individual steps**

No generic terms. Only concrete activities.

### **3. add five fields per step**

Role, processing time, waiting time, frequency, standardizability.

### **4. start with a triple review**

- Gemini for patterns and clusters
- ChatGPT for prioritization
- Claude for causes and risks

### **5. mark only three candidates**

Not ten. Only the three steps with the greatest leverage.

### **6. decide between three types of measures**

- eliminate
- standardize
- automate

### **7. test a pilot in four weeks**

Not a mammoth project. A clearly defined sub-process is enough.

## **8 Strategic classification**

Those who do not examine processes for time wasters often automate the wrong things.

Instead of removing the bottleneck, only a single work step is made faster, while the actual loss continues to be waiting times, queries and media disruptions.

This is precisely why this type of AI use is strategically relevant:

- You use AI not just as a text aid, but as an analytical tool.
- You recognize where standardization must come before automation.
- You reduce blind investments in tools.
- You build up internal process expertise instead of tool dependency.

The competitive advantage does not lie in the fact that you "also use AI".

It lies in the fact that you **decide better what needs to be changed first**.

## **9 My conclusion**

If you want to find out which steps in process X really eat up time, you don't need a perfect transformation from day one.

You need transparency first.

Google Gemini, ChatGPT and Anthropic Claude can each help you in different ways:

- **Gemini** for a quick sample view
- **ChatGPT** for prioritization and implementation logic
- **Claude** for in-depth focus, root cause analysis and risk assessment

However, the real strength comes from the interaction:

first cluster, then prioritize, then secure.

This turns AI into a tool for better processes rather than a toy.

## 🚀 Next step

If you not only want to understand AI, but also use it in a structured way in your company, then:

👉 Find out more about our AI training:  
[https://bloo.school](https://bloo.school/en/)

👉 Find out about our Smart Market Fit offers:  
[https://bloola.com/smf](https://bloola.com/smf-system) - The Smart Market Fit course  
[https://bloola.com/smf-system](https://bloola.com/smf-system) - The Smart Market Fit system for companies

👉 Or find out more about our consulting and automation solutions:  
[https://bloola.com](https://bloola.com/en/)

[View full post](https://blog.bloola.com/digital-change/recognizing-automation-potential-which-steps-in-process-x-really-eat-up-time)

```json
{
  "@context" : "http://schema.org",
  "@type" : "BlogPosting",
  "author" : {
    "@type" : "Person",
    "name" : "Lars-Thorsten Sudmann"
  },
  "dateModified" : "2026-03-16T11:36:45.369Z",
  "datePublished" : "2026-03-16T11:36:45Z",
  "headline" : "AI Masterclass: Identifying Opportunities for Automation – Which Steps in Process X Are Really Time-Consuming?",
  "image" : {
    "@type" : "ImageObject",
    "height" : 512,
    "url" : "https://1693371.fs1.hubspotusercontent-na1.net/hubfs/1693371/Blog/digitale-change/AI-Process-Automate/AI-Process-Automate-en-hd.png",
    "width" : 768
  },
  "mainEntityOfPage" : "https://blog.bloola.com/digital-change/recognizing-automation-potential-which-steps-in-process-x-really-eat-up-time",
  "publisher" : {
    "@type" : "Organization",
    "logo" : {
      "@type" : "ImageObject",
      "height" : 0.0,
      "url" : "https://www.bloola.com/hubfs/General%20Material/Logos/bloola-logos.svg",
      "width" : 0.0
    },
    "name" : "Digitalisation inside Business"
  }
}
```