Panos.AI is meant to become the world's 1st digital advisor to immediately help companies and their employees with automating business processes more efficiently, because it already knows processes and solutions – thanks to the integrated artificial intelligence!


During the hackathon, we wanted to tackle the challenge of providing first strategic insights to companies for kickstarting automation initiatives based on publicly available data. We wanted to prove, that it is technically possible using NLP to recommend potential and quantified candidates for automation using job postings in English language with salary information, which typically also include listings of manual tasks. The output is an overview of ranked departments within companies which have been classified based on the job postings according to their automation potential quantified by employee cost per year to show potential savings as well as benchmarking. This would allow customers to quickly decide in which departments they should look first to improve efficiencies/cost for meeting company goals/competitive situation.

How does it work?

This prototype classifies text between two classes using embeddings: 1-Automation unlikely / 2- Automation likely

The Cohere API & SDK allows us to test different pre-trained models from different transformers without minimal changes to our code. Our dataset contains 4345 job postings classified via rule-based logic & expert review. But we will use just 500 for the prototype.


  • Get the dataset
  • Get the embeddings of the reviews (for both the training set and the test set)
  • Train a classifier using the training set
  • Evaluate the performance of the classifier on the testing set

Why is this unique? 

Classifying business automation needs from publicly available data is unheard of, until now, companies have primarily access to generic content on where automation potential could reside. This is typically followed by lengthy consulting, workshop and software implementation activities limited to certain departments missing the overall picture for informed strategic decisions.

Statistics show that therefore 40-70% of transformation projects are not as successful as initially planned or even fail completely.

Technology stack used

  • Dataset:
  • Microsoft Excel & Power Query (visual data cleaning)
  • Google Colab & Drive
  • Python
  • Pandas
  • Cohere API & SDK (large BERT pre-trained model) 
  • scikit-learn (support vector machine)
  • Microsoft Power BI (quick visualization of insights) 

Target Profile

Customers in general:

  • Companies of almost any type, industry, size, location, language

Customer Size segments, by priority:

  1. Large companies with 250 or more employees/EUR 50 million turnover
  2. Medium-sized companies with 50 or more employees/10 million euros or more in sales
  3. Small companies with 10 or more employees/2 million euros in sales

Typical customers, by priority:

  1. Companies that need or plan to "optimize" processes but do not yet know where to start/what to do & need help
  2. Companies that have already automated processes but were unsuccessful want to check and improve their status
  3. Companies that have already automated processes and have been successful in doing so, but would like to review, improve, and expand their status throughout the company

Typical contact person for customers:

  • Executive level/management (CEO, CFO, CMO, COO, CTO, CIO etc.)
  • Department/group manager (procurement, accounting, human resources, marketing, sales, organization/operations, production, IT, etc.)

Special customer segments by industry, which e.g., have an increased need for automation & require an industry solution:

  • Public sector

Specific customer segments by rollout, languages/regions/country by priority (example):

  1. German, DACH: Germany, Austria, Switzerland
  2. English, UK or USA, Australia
  3. Other languages/countries according to customer needs (Spanish, French, Portuguese etc.)

Partners in general:

  • Manufacturers of software solutions
  • Service provider for additional, individual consultation
  • Service provider for implementation of software solutions
  • Service provider for additional, individual training
  • Providers of process, industry, solution-specific content (analysts, consultants, freelancers, etc.)