how we create an innovative business solution
Hello! The AI S&OP startup team is in touch, we are students of the largest (and also the best) master's program in Artificial Intelligence from ITMO's AI Talent Hub. Our master's program has a very cool activity – a project seminar, where you need to spend the entire semester developing some kind of applied industrial solution, as well as justify its commercial significance. To begin with, I would like to introduce our team, which was formed this semester to solve one non-trivial problem, which will be discussed further. So, AI S&OP is Sesorov Alexander “YADRO” AI Benchmarking Engineer, Kazantsev Arseniy “Vita” Machine Learning Engineer, Karpova Daria System Analyst of the BI department and Groza Artem “SimpleWine” Data Scientist.
![](https://habrastorage.org/getpro/habr/upload_files/c48/abe/a64/c48abea64e2a2b54bb71887d29235e54.png)
The last few years have been a real test for businesses all over the world. The pandemic, geopolitical conflicts, climate cataclysms – these “black swans” have turned the usual work patterns upside down. The supply chain management sector has been hit especially hard. Good old Excel tables and managers’ intuition have suddenly proven powerless in the face of new challenges. The result? Empty shelves, mountains of unclaimed goods, and a headache for business owners.
![](https://habrastorage.org/getpro/habr/upload_files/0bf/e8e/cd1/0bfe8ecd10fc3709e7afcfb42bff5eb5.png)
If you study the current market of forecasting solutions for business, you will find that not everyone uses machine learning in their products, and those who do set a high price for the implementation and support of their black-box solutions.
Taking into account all the new input data, it is immediately clear that under such conditions, small and medium businesses remain deprived of the opportunity to use the most relevant technologies to optimize their operational activities. This is where the idea came to implement AI S&OP – an innovative B2B SaaS solution for optimizing supply chain management using advanced machine learning and artificial intelligence technologies, which would allow small businesses to use a data-driven approach, intelligently analyze data, but at the same time pay adequate money for the work that is usually performed by an entire department of analytics and machine learning. In addition, one of the key concepts of the product is ease of use, so that any non-technical specialist can intuitively understand the service and choose the best tools for their tasks.
Who are we trying for?
The Russian supply chain management software market has undergone significant changes after the departure of international players. The resulting niche was quickly filled by domestic developments (Goodsforecast, Novo BI, Astro SCM), but after analyzing the capabilities they provide, we realized that most of them are aimed at large businesses and require significant resources for implementation and maintenance. At the current stage, we do not want to compete with large market players, so we decided to focus our attention on the following segments:
– Owners and managers of small and medium-sized retail enterprises
– Heads of logistics and inventory management departments in medium-sized companies
– Innovative startups seeking to optimize business processes
Key stakeholders include senior management (CEO, CFO, CIO), supply chain management, logistics, sales and marketing, IT, purchasing and inventory planning managers.
![](https://habrastorage.org/getpro/habr/upload_files/eee/ab6/2c5/eeeab62c5f77d1a573015bf5d689dbf1.png)
Among the main advantages of our solution, we highlight the following:
1. Increasing the availability of goods at points of sale
2. Optimization of inventory levels
3. Automation of stock replenishment processes
4. Increase in goods turnover
These advantages formed the basis of the hypotheses that we want to test at the time of beta testing of our service.
What is innovative?
In the AI S&OP project, we want to offer a flexible approach to implementation:
1. On-premise solution: for companies that prefer local data storage. As we found out during the audience research, this option is still quite popular among companies concerned about data privacy.
2. Cloud solution: suitable for retail networks of various sizes, does not require the purchase of additional equipment and software, rapid deployment is possible. This is where we started our development.
We included the following in the technology stack at the MVP stage:
– Advanced machine learning models for forecasting;
– Report generation systems based on artificial intelligence;
– Interactive dashboards for data visualization.
![](https://habrastorage.org/getpro/habr/upload_files/090/374/4b5/0903744b54be10d6bd82a8af8aaff8e9.png)
Thus, we mean that our solution will allow companies to:
– Optimize inventory management;
– Reduce logistics costs;
– Increase turnover and profit;
– Free up resources for strategic planning and development.
How We Researched Our Audience
Our team is still conducting cast-dev interviews (customer development) with the selected target audience, but at the current stage we have already interviewed five company representatives (special thanks to the managers who agreed).
The purpose of the interview was to identify the main problems (or, as business analysts now like to say, “pains” of customers), to find out what software small and medium-sized businesses use or use at all, why they do not want or want, but are unable to, switch to more modern solutions.
Thus, we identified three main hypotheses to test:
Hypothesis 1: Small and medium-sized companies experience significant challenges in managing their supply chains and require affordable and easy-to-use solutions.
Result: 70% confirmation
Particularly useful for this hypothesis was an interview with the commercial director of a St. Petersburg company (the company wished to remain anonymous), which deals with switchboard equipment (OKVED 27.12 Production of electrical distribution and control equipment) and is a typical representative of our target audience: the number of employees does not exceed 50 people, quite conservative management of the company, average age of employees. From this interview, we learned that many companies from their supply chains (and these are more than ten of the same partner companies) still remain at the level of supply planning “in Excel”, and that the demand for new solutions is at the level of “haven’t they thought of it yet?” nothing newer.” Everything is as usual: management wants it quickly, simply and not very expensive.
Hypothesis 2: Small and medium-sized companies will prefer cloud solutions due to their flexibility, speed of deployment, and lack of need for significant capital expenditures.
Result: 80% confirmation
Cloud solutions turned out to be preferable for the majority of respondents from companies that are not associated with scientific or intellectual activities. Of the five companies surveyed, only one (which is engaged in the development of new control systems and has more than three patents) was not ready to even simply consider the option of cloud storage of information. The other four companies noted that the lack of the need for significant capital expenditures on IT infrastructure is a significant advantage for them and that some other processes have already been transferred to cloud services. If we want to further cover the segment of “conservative companies”, then, of course, we will have to work on a local solution.
Hypothesis 3: The transition to new software may cause difficulties related to integration and user adaptation, so support from the vendor is important.
Result: 50% confirmation
Half of the respondents expressed concerns about possible difficulties associated with the integration of new software and adaptation of employees – the transition to new software for employees who have been planning deliveries in Excel for more than ten years should be gradual and calm so that current business processes are not affected.
We also got an insight: after 2022, when supply chains have become more complex, the most difficult thing is to meet delivery deadlines and plan deliveries in advance – in conditions of uncertainty, fines for failure to meet contract deadlines have become tougher.
What do we want to check after implementation?
It's simple: we want to make sure it gets better. We assume the following, judging by the already conducted alpha testing of our service:
The use of AI/ML technologies will improve the accuracy of demand forecasts by 20% compared to current methods;
Automating the replenishment process can reduce the labor intensity of manual adjustments by 30%;
Cloud solutions can reduce IT infrastructure costs by 25% compared to on-premise solutions (of course, only for those companies that need it).
We will talk specifically about the technical part of our project, the models we chose between, and the architecture of the solution in the next part.