Social Discovery Group is looking for a Service Cost Optimization Senior Manager.
A brief about the company. It’s a giant in the social discovery world, owning dozens of dating and communication platforms like Dating.com, Dil Mil, and Cupid Media. Their mission is noble — to combat loneliness by connecting millions of people worldwide. They have a huge, distributed team, working from anywhere on the planet. In short, these guys are modern, tech-savvy, and clearly know how to count money.
Now, about the pain point. Judging by the job description, the company has encountered a classic growth dilemma. The number of partners (likely streamers, affiliates, content creators) is growing, and so are the expenses associated with them. And here arises a question familiar to any manager: how can we pay less, but in such a way that no one notices or gets offended? So that the quality of service for partners doesn’t drop, and they don’t run off to competitors. For this, they are looking for an experienced person who will dive into statistics, dig through reports, optimize processes in Bitrix CRM, and negotiate with other departments. A noble, but painfully routine job.
And this is where, as an old IT guy, a question arises for me. Why do we need a person for this? After all, this entire task could be handed over to artificial intelligence, which would do it faster, more accurately, and ultimately, cheaper.
Let’s imagine a dialogue in a hypothetical SDG meeting room. On one side — the ‘Skeptic,’ a department head who believes in people and Excel. On the other — the ‘Visionary,’ who has already mentally replaced half the office with algorithms.
Skeptic: We need someone with strong analytical skills who will pore over our reports, seek out anomalies, and suggest where we can cut costs. They will personally ensure that partners are satisfied.
Visionary: What if we put a model on this task? A person will look at reports once a week. AI will monitor data 24/7 in real-time. A person will find a couple of obvious patterns. AI will analyze thousands of parameters and uncover hidden correlations we hadn’t even suspected. For example, that partners from Brazil streaming on Tuesdays bring in 15% less revenue per dollar spent than partners from Thailand streaming on Saturdays. Can a person find that? Only if they stumble upon it by chance.
Skeptic: Sounds like fantasy. Where do we start? Our data is in CRM, in billing, in Google Sheets… it’s a mess.
Visionary: By creating a single source of truth. Step one — we build a Data Lake. We dump absolutely everything into it: data from Bitrix CRM regarding partner inquiries, financial transactions, their performance metrics, even data on user engagement they bring in. Step two — we unleash an analytical AI system on this data. This could be a custom Python solution using libraries like TensorFlow and Scikit-learn, or ready-made MLOps platforms. This system will be our ‘optimization manager.’ It won’t just find anomalies; it will segment partners based on dozens of non-obvious characteristics and propose a specific strategy for each segment: for these — change the bonus program, for those — offer a new, cheaper tariff, and for the most inefficient ones, perhaps it’s worth disconnecting them.
Skeptic: But costs aren’t just direct payments. They also include our support team’s time spent resolving their issues.
Visionary: Exactly! AI will analyze all tickets in Bitrix. It will identify the most frequent and costly types of requests for us. And it will immediately propose a solution: for 80% of typical questions — create a chatbot that will resolve them instantly. And for the remaining 20% of complex cases — it will prepare a problem summary and ready-made solution options for the support operator. There you have it: reduced operational costs without sacrificing quality. We’re not just saving money; we’re improving the partner experience — they get answers faster.
Skeptic: Alright, but I don’t trust a ‘black box.’ How will I know that AI won’t mess things up?
Visionary: We won’t give it the keys to the car right away. Initially, it will operate in an advisory mode. The model will issue recommendations, and a human analyst (perhaps even hired for a more junior position) will review and approve them. After a month, we’ll compare: how much money a person saved by following their intuition, and how much could have been saved by following AI’s recommendations. Numbers are stubborn things. Gradually, when we see that 9 out of 10 AI recommendations are accurate and effective, we can automate their execution.
And how do we validate the results? It’s simple, just like with any IT product.
First — A/B tests. We take two similar groups of partners. For one, we apply the strategy developed by AI; for the other, we leave everything as is. After a month or two, we look at key metrics: partner churn, their average revenue, Net Promoter Score (NPS), and, of course, our costs.
Second — simulation on historical data. ‘What if we had implemented this algorithm a year ago?’ The model can calculate this with high accuracy and show how many millions we missed out on while manually digging through Excel spreadsheets.
Ultimately, instead of searching for one expensive senior manager who will become a ‘bottleneck’ and burn out under the weight of reports, the company could invest in creating a system. A system that won’t go on vacation, won’t get sick, will only get smarter with every new gigabyte of data, and ultimately bring far more benefit. And people? People will be left with the most interesting work — setting tasks for this system and interpreting its unexpected conclusions.
Источник: https://www.linkedin.com/jobs/view/4410183334/
