Leo Postovoit

AI for Demand Forecasting

I recently worked on an confidential AI project in the F&B space that I’m very proud of.

I took on this awesome #martech project that blended all of my skills. I created a AI-powered predictive forecasting tool (powered by PowerBI, ChatGPT, Pinecone, Nodejs and Python, developed as microservice) that better enriched their data to predict demand and bring insights to the whole team.

The challenge

The organization had strong online sales but struggled to predict demand, leading to missed revenue opportunities and product loss.

The people at this org needed clarity from the rest of the business. Tgere were many unconnected systems, including:

  • Production and supply chain tool (Manufacturing/ordering)
  • POS (Sellable inventory)
  • employee scheduling (SaaS)
  • Web Analytics (GA4)
  • SEM/paid social campaign management tools

Each had valuable data that other teams could use, but it was locked up among. And maybe you’d see other decision-makers, a watercooler conversation would occasionally bring collaboration—we had to find a better way.

Technical details: Here’s more specifics on how I’m actively working with AI projects to leverage advanced usage of ChatGPT and similar tools on the API level.

I’m finding new uses for these LLMs in creative applications focusing on ways to implement it responsibly and ethically alongside other efforts, such as data analysis, automation, and content creation.

Project approach

  1. I analyzed a host of data sources:
    • An existing PowerBI project
    • E-commerce sales – Shopify Plus data
    • Google Analytics (GA4) data
    • Production schedules Supply chain data (including waste reports)
    • Paid Social and SEM campaign data
  2. I integrated the client’s main live data into a MySQL database, and then built out powerful automations and reporting dashboards in PowerBI.
  3. I transformed this data stream into a AI-powered predictive forecasting model for employee, production, and marketing management.
    • I fed these BI data streams into my custom application (nodejs, express)
    • Leveraged ChatGPT4 to generate insights and predictions based on parsed data (NumPy, pandas)
    • Leveraged a custom Generative Adversarial Network (GAN) for better results: The GAN reviews queries, generated synthetic data, and improves the robustness of the model
    • Outputs of this data went back into the PowerBI interface, stored snapshots in a vector database (Originally MongoDB, now Pinecone)

Outcomes from the project

The results were almost immediate: We radically enabled the whole team (not just the C-suite) with access to this powerful business data in contextual places where they make these decisions.

And, yes we did! As a result, we optimized production, staff, and retail planning. After just 8 weeks and some fine tuning, we saw a 15% gross sales increase and 45% reduction in waste.

📈

Because this org actively decides its work while operating software on a semi-siloed approach, I wanted to make sure the forecasting data was available to staff at all levels, wherever they make key decisions.

This enhanced data helped leadership make even better production planning decisions that made for more effective supply chain management: We also saw a 23% improvement in stockout prevention, ensuring popular items remained available during peak times.

From front-of-house staffing and inventory logistics, to financial forecasting and production shortages, the whole data pipeline came together into a insightful application that gets smarter each day through the power of machine learning and AI.

AI’s impact on the world around us is profound, and I firmly believe it must be used ethically and surgically. I’m actively learning about tools in content generation, audience analysis, and more. This knowledge enables me to help organizations navigate this complex landscape while maintaining their integrity.

By understanding the business and the potential tools, we found a profitable, ethical and easy-to-understand way to apply the best of what we can do today. I designed an approach to weave advanced data analysis tools within their platforms.

Opportunities abound

The new frontier of AI, tech debt and the race to create loyal audiences

I’ve loved helping organizations break free from technological constraints, embrace audience-focused approaches, and ethically implement cutting-edge AI solutions. I can help you navigate this all.

My diverse background in journalism, product development, software development, and AI positions me as a unique asset for media companies seeking to thrive in a dynamic landscape.

My proven success across newsrooms, studios, and emerging creator platforms demonstrates my ability to drive measurable results, whether boosting traffic, optimizing production, or harnessing AI for business growth.

Are you a publishing organization struggling to unlock
your potential due to challenges with your technology?

I specialize in solving complex challenges, and want to work together to build sustainable innovation for you.

Let’s chat!