How I Built Business-Automating Workflows with AI Agents
How I make money helping businesses boost their productivity and cut costs by automating supply chain analytics workflows using n8n. The post How I Built Business-Automating Workflows with AI Agents appeared first on Towards Data Science.
In a previous article, I shared several case studies of AI Agents supporting the sustainability roadmaps of small, medium and large companies.
This is part of a 4-month exploration of how agentic AI can drive business digital transformations.
It all began when I discovered no-code tools like n8n, which allow you to design and automate powerful workflows with just a few clicks.
My co-founder summed up the opportunity perfectly:
“We need a method to deploy simple and efficient solutions to scale our offer of analytics and automation.”
So we got to work.
We built case studies and prototypes to show how these AI agents can help customers accelerate digital transformation.
Today, AI agents for workflow Automation represent over 60% of our revenue.
In this article, I want to share the inside story of our explorative journey and how we turned this opportunity into a business so you can learn to replicate our approach.
Far from buzzwords and clickbait articles, we will take a pragmatic approach to explaining how these tools can bring added value to businesses without selling dreams.
In short, we will cover
- What types of examples and case studies we shared to attract customers
- The problems prospects brought to us (and how we listen to their needs)
- How do we sell these solutions (ROI arguments and business impact)
If you’ve been wondering how to turn automation and AI agents into a profitable service, this article is a good inspiration to start investigating.
How Case Studies Brought Customers to Our Door
Learning by doing
When I want to learn a new tool, I create case studies.
When OpenAI released its first GPT API, I started exploring LangChain with a case study shared in this blog.
It’s my favourite way to turn curiosity into hands-on expertise.
Therefore, my blog contains over 70 case studies covering analytics for supply chain optimisation, logistics cost reduction, and sustainability.
So when I discovered AI-powered workflow automation with n8n, I replicated this proven process:
- Learn the basics of the tool
- Pick an example of an operational or business problem from the multiple projects I conducted in the past.
- Build a prototype and measure its impact
- Publish a case study
My n8n creator profile became a playground to connect business and operational problems with real-world solutions.
This approach sharpened my skills and became a boost to attract customers looking for practical solutions to their problems.
How to select your first case study?
I started by focusing on the most significant pain points I had seen during my 9+ years in logistics — across 3PLs, FMCG companies, and luxury brands.
What manual processes were draining productivity and limiting capacity?
One example that came to mind was from the distribution operations of a fashion company in China.
The pallets, shipped from Europe, were delivered by a truck in the morning.
Unfortunately, we were not connected to the supplier via Electronic Data Interchange (EDI).
We had to rely on email exchanges to know what we would receive.
This meant warehouse admins had to:
- Review the emails coming from the customer’s logistics team.
- Manually input order details into the system (PO number, delivery date, SKU ID, quantities)
What are the pain points here?
This is a bottleneck for the reception process.
Indeed, if the orders are not keyed into the system, inbound teams cannot receive the pallets.
But more importantly, this manual process impacts the productivity of admin teams that spend time on tasks that do not provide added value.
Therefore, I used this example to build my first n8n automation.
It is a simple workflow connected to Gmail that automatically analyses all the emails related to inbound orders.
A GPT-powered AI node analyses the content of the email to extract the order information.
The output is saved in a spreadsheet ready to be loaded into the system by a script.
For more information about this workflow, check the complete tutorial here
What happened after I published this workflow?
The Problems Clients Brought — and How We Solved Them
After publishing, prospects from different industries and countries began seeking help with similar problems.
This is answering a common problem: the lack of interconnectivity between systems.
Before, companies would rely on manual work to bridge the gaps.
Now, we can deploy AI agents to automate those bridges.
A furniture company in the Netherlands contacted us.
A part of their value chain (small suppliers) wasn’t connected via EDI.
Therefore, they relied on manual processes to send and receive orders.
After multiple interviews with their IT team, we learned:
- The systems could generate EDI messages
- But they couldn’t receive or process incoming messages
Therefore, we have proposed a workaround
- The sender would generate the EDI message and send it via email
- An n8n workflow would collect the email, parse the content and save the outputs.
We have now removed a significant pain point for the customer by “partially” automating the process.
In this example, AI agents provide a great added value: adapting the tool to the format of the EDI message.
This makes the tool more robust and adaptable to the different systems their suppliers are using.
For more information on implementing this solution (with the template available), check this tutorial.
This solution has multiple advantages:
- It is cheaper and easier to deploy than properly connecting the furniture company to each supplier.
- This improves the data quality and allows the company to allocate these data input teams to customer service.
This was our first victory and proof that we can support businesses with simple solutions.
For the following projects, we developed a process to qualify customers and standardised the solutions design.
- Define the problem faced by the customer
→ We need to manually input orders received from suppliers. - Find the root cause
→ Absence of Electronic Data Interchange (EDI) connections - Design a solution
→ AI-powered workflow automation to parse EDI messages.
But, how do we sell these solutions?
I’ll share our method for selling these automation solutions in the next section.
How We Sell These Solutions
After building and testing these automation prototypes, the next challenge was clear: How do we position and sell these solutions to clients?
Considering that most of our leads come from my content (Towards Data Science, YouTube, n8n Creator Page), we developed three sales approaches adapted to the request of the prospect.
The idea is to adapt the sales pitch to clear business outcomes and return on investment (ROI).
Approach 1: Automate to Improve Productivity
This is the primary concern of our customers.
They are pressured to reduce costs by improving the productivity of their operators and admin teams.
Prospect: “A customer service admin can process 6 orders per hour with our manual process. How many orders can you process with AI?”
When we have a customer with a case related to productivity, we focus our questions on:
- Defining the current workflow, including every step with a detailed description (data input, outputs, systems used)
- Estimating the current productivity (number of actions per hour) and the workforce cost ($/hour)
For instance, an agency focusing on ESG consulting approached us to help them automate their sustainability regulation monitoring process.
They ask their highly skilled (high hourly cost) ESG analysts to review the European Parliament’s activity reports and select those related to sustainability.
Prospect: “We want these analysts to focus on tasks with higher added value.”
We developed a workflow to completely automate this process with an AI agent, providing a high return on investment.
ROI = Time Saved × Hourly Rate of Analysts
This simple calculation is key to convincing the decision maker that it is worth investing resources and money in implementing your automation.
To our prospect: “This solution will save you XXX euros per month, you will recover your investment by YYY months.”
For more information about this workflow, check the complete tutorial
For some prospects, the priority is bringing more business; we can also use AI.
Approach 2: Automate for Marketing & Promotion
In my startup, LogiGreen, we were the first users of this kind of workflow.
As a small structure, we don’t have the financial means to recruit a complete Marketing and commercial.
How can we build an automated machine of content distribution to increase our reach?
Therefore, we rely on automation to promote our business — a great way to create or curate engaging content.
The screenshot above comes from an email generated by our fully automated newsletter.
- You can find links to our website in the header and footer.
- The distribution list includes all our customers and prospects.
We use n8n with GPT-4o to automate this solution.
This simple workflow uses a classification agent to curate a list of articles related to sustainability coming from multiple sources (EU news, specialised journals, social media).
The AI agent handles the editorial choices based on the instructions in its system prompt.
It is an excellent way for us to create an online community, stay top of mind with our customers, and share engaging content.
To our prospects: “With this content creation machine, you can occupy the space, provide value and create an online community centered around your business.”
We are now selling this “content creation machine” to our prospects!
For more information about this “automated newsletter”, check the video below
Some prospects that contacted us said their top priority is not just productivity or marketing.
Approach 3: Automate for Quality and Error Reduction
They must ensure quality and reduce errors, especially in critical reporting and compliance workflows.
For instance, a growing number of companies need to comply with the Corporate Sustainability Reporting Directive (CSRD)
This requires publishing ESG disclosures in machine-readable XHTML format, embedded with XBRL tags.
These reports must meet strict formatting and compliance standards to be accepted by regulators.
Prospect: “We have issues with formatting our reports correctly.”
To support this prospect, we developed an n8n workflow that helps sustainability teams and ESG consultants automatically audit these reports.
Their analysts have to email a draft of the report to a specific address.
The attachment is downloaded, and its content is analysed to find missing values and format errors.
The output of this audit is sent to an AI agent that will explain, in plain English, the errors found in the report.
We position this product as reducing errors to save time and money.
Before having this tool, they relied on a third-party company to check their reports.
By automating the audit process, companies can:
- Avoid costly reporting mistakes
- Reduce the risk of regulatory non-compliance
- Accelerate turnaround time on reports
- Free up expert resources to focus on analysis, not formatting.
The Return On Investment here comes from minimising rework costs and delays while increasing the reliability of reported data.
For more details about this automation, check this tutorial (including the n8n template)
You will discover that the core block of this audit tool still uses deterministic code (for reliability).
However, AI is improving the user experience by providing explanations in plain English.
To our prospects: “You don’t need to be an XHTML expert to correct the report.”
We are selling this to our customers: AI provides advanced analytics-as-a-service capabilities to any company.
Conclusion
Over the past few months, we have transformed automation from a personal exploration into a promising business offer.
By focusing on three key approaches — improving productivity, boosting marketing reach, and ensuring quality — we helped companies across industries solve real problems, save time and increase their impact.
We don’t sell dreams with AGI replacing their entire supply chain departments.
The opportunity here is enormous.
Businesses seek people who can bridge the gap between operational challenges and automation solutions.
If you’re curious about starting your automation journey, I will provide frequent updates in this blog.
If you want to implement these solutions for your business, feel free to connect with me on LinkedIn.
About Me
Let’s connect on Linkedin and Twitter; I am a Supply Chain Engineer using data analytics to improve logistics operations and reduce costs.
For consulting or advice on analytics and sustainable supply chain transformation, feel free to contact me via Logigreen Consulting.
Samir Saci | Data Science & Productivity
A technical blog focusing on Data Science, Personal Productivity, Automation, Operations Research and Sustainable…samirsaci.com
The post How I Built Business-Automating Workflows with AI Agents appeared first on Towards Data Science.
