The other day, I was scrolling through LinkedIn and stumbled upon an interesting job opening. **Donau Soja Organisation** is looking for a Product Owner/Project Manager to join their team. And you know what I thought? A classic story. A great company, an important mission, but their approach to product management is straight out of 2015. Let’s break down why hiring a person for this role is like buying a horse when the automobile has already been invented.
**Who are Donau Soja?**
In short, it’s a non-profit organization from Vienna that is building a new reality in Europe – one with sustainable, non-GMO protein (mainly soy). They bring everyone together: from the farmer with a tractor to the retailer with a supermarket shelf. Their goal is to make food supply chains more transparent, sustainable, and reliable. A mission worthy of respect. To achieve this, they have some kind of IT product, obviously a platform for tracking, certification, and interaction among participants in this very chain.
**What Pain Are They Trying to Solve?**
Let’s take an honest look at the job responsibilities. The company is looking for a “one-person orchestra.” They must:
* Gather requirements from the business and the market.
* Translate them into clear User Stories for developers.
* Prioritize tasks using a clever “value/effort/risk” formula.
* **”Lead and guide external IT vendors”**. Here’s the key point. They don’t have their own development team; they work with a contractor.
* Coordinate testing, releases, document everything, conduct training, and report to management.
Essentially, Donau Soja isn’t just looking for a manager, but a **human API** between their business and an external development team. This is a classic pain point for non-IT companies that have an IT product. Communication with a contractor always means losses in translation, delayed timelines, budget disputes, and mutual misunderstanding. And they want to plug this gap with one expensive specialist who will perform 80% of routine, mechanical work.
**What if, instead of a person, you hire a system?**
Imagine not a single employee, but a digital system powered by AI, which we’ll call the “AI Product Hub.” This system doesn’t get sick, doesn’t go on vacation, doesn’t burn out, and works 24/7. It takes on all the operational routine, leaving people with only strategy and final decisions. Sounds like science fiction? Let’s see how this can be implemented today.
—
### A Dialogue in the Smoking Room Between Two IT Directors
**Mikhail:** Sergey, did you see that Donau Soja job posting? They’re looking for another “universal soldier” to kick contractors into gear.
**Sergey:** Well, how else? Business speaks its own language, techies speak theirs. You need a translator. A live one, with experience. Someone who can formulate requirements and slap a vendor’s hand if they propose nonsense.
**Mikhail:** Why do you need a whole person for that? 80% of their work is copy-pasting and rephrasing. Look. “Gather requirements and write user stories.” We can feed a custom GPT all stakeholder call recordings (transcribed via Whisper, of course), email correspondence, and support tickets. It will identify key requests, group them, and draft User Stories with Acceptance Criteria. The human only needs to review and approve.
**Sergey:** Okay, granted. But what about prioritization? That requires intuition, market understanding.
**Mikhail:** Intuition is good, but it doesn’t scale. AI can be more objective. It will analyze how often a particular feature is mentioned in requests, correlate it with the company’s strategic goals (which we’ll also “feed” it), and calculate an objective RICE scoring. No “I think this is more important.” Just data.
**Sergey:** Alright, you’ve convinced me on that. But the hardest part is working with the vendor. Controlling deadlines, quality, documentation. Is a neural network going to call them on Zoom?
**Mikhail:** No, it will be their system assistant. From an approved User Story, the AI generates a formal technical specification for the contractor. When the vendor sends technical documentation, the AI checks it for completeness and adherence to the template. It can also monitor the repository and flag if code style or test coverage drops below the agreed-upon level. This is cold, impartial control. And for routine tasks like “when will it be ready?”, there are automated status requests and response parsing.
**Sergey:** Hmm… What about testing? UAT, regression? People have to click through everything, right?
**Mikhail:** Not always, not anymore. There are AI tools that generate test cases directly from User Stories. There are visual regression systems that detect the slightest changes in layout. AI can analyze logs after a release and automatically create bug reports with the most precise problem descriptions. A human is only needed for non-trivial scenarios. Documentation and scripts for training videos? The AI will write them in 15 minutes, not three days. Reports for management? It will gather data from Jira, Git, and AWS and compile it into a beautiful dashboard.
**Sergey:** Sounds good. But people won’t trust it. How will you make the business believe that a machine understood them correctly?
**Mikhail:** You don’t have to force it. Start small. Initially, the AI acts as a “co-pilot” for a human manager. It doesn’t make decisions but offers options: “Here are 5 User Story variations, choose the best one,” “I believe this bug is critical because 3 Category A clients reported it. Do you agree?”. Gradually, as the team sees that 9 out of 10 AI suggestions are adequate and save time, trust grows. We introduce the **Human-in-the-loop** principle: AI does, human verifies. Over time, verification becomes a formality.
### How to Verify that this “AI Product Hub” Works?
Validation is key to everything. Here, you don’t need to take our word for it; you need to measure.
1. **Artifact Quality:** Conduct blind A/B testing. Give developers two task descriptions: one written by a human, the other generated by AI. Ask which one is clearer and more complete. The results will surprise you.
2. **Speed:** Measure key metrics. Time-to-market for a new feature, time to fix a bug, time to write documentation. Compare “before” and “after” the implementation of AI assistants.
3. **Cost:** Calculate not only the development cost from the vendor but also the Total Cost of Ownership (TCO). How many hours do internal experts spend on task definition and acceptance? AI reduces this time significantly.
4. **Satisfaction:** Assess stakeholder satisfaction. Are they getting the features they need faster? Has the number of errors in releases decreased?
Instead of looking for one expensive specialist who will become a bottleneck and a single point of failure, Donau Soja could invest in creating a system. A system that automates routine tasks, makes the development process transparent and manageable, and allows their own agribusiness and sustainable development experts to focus on what matters most – product vision, rather than micromanaging contractors.
Because in 2024, it’s no longer people who compete, but management systems. And those who realize this sooner will gain a decisive advantage.
Источник: https://www.linkedin.com/jobs/view/4404076493/
