Senior Product Manager

These 7 AI startups have a surprising investor: Telekom. 18:37, 10 Aug. 2026 From the telephone booth to unicorn shepherd. Telekom has invested in some of the most renowned German technology startups. When you think of Telekom, you usually think first of mobile phone masts and the company's own magenta. Less well known: The DAX group has been acting as a startup investor for years. Some stakes are tiny, others strategically interesting; together they paint a picture of where the journey could go. T.Capital, as the investment vehicle is called, pursues a two-pronged approach. The venture arm typically invests in Series A to Series C financing rounds and takes stakes of up to five million euros in companies. In addition, there is a tech fund with significantly larger tickets of up to 300 million euros. This allows Telekom to both accompany young providers early on and play a role in larger technology bets. The stakes show that the group has long been interested in more than just classic telecommunications. Many investments revolve around software, artificial intelligence, cloud infrastructure, or data-driven business models. Isn't everything somehow AI these days in the end? N8n. The Berlin startup N8n develops a platform for workflow automation. Companies can use it to connect different software services and automate recurring processes. The topic is interesting for Telekom because large organizations have to coordinate countless IT systems. According to the commercial register, Telekom holds a 0.08 percent stake in N8n. Quantum Systems. Quantum Systems from Bavaria develops drones and airborne reconnaissance systems for civilian and - now also - military applications. The systems generate large amounts of data that need to be transmitted, processed, and evaluated. That is exactly where Telekom's classic strengths lie. According to the commercial register, Telekom's stake amounts to 1.27 percent. Interestingly enough, until a few years ago many investors wanted nothing to do with weapons of war. That is why Quantum initially focused on civilian use. That has changed. Dash0. Dash0 develops software for monitoring and analyzing complex cloud infrastructures. Such observability platforms help companies quickly detect errors in their applications and run digital services reliably. For Telekom, which operates large network and IT landscapes worldwide, the strategic proximity is obvious. However, the exact stake size is not publicly visible because the company is organized via a US parent company in the form of a so-called Delaware flip. Lovable. The Swedish startup Lovable is one of the best-known European providers of platforms for vibe coding. Lovable enables users to create software applications with the help of artificial intelligence. For Telekom, this offers a glimpse into the next generation of software development and AI tools. The ownership structure is also not directly traceable in the German commercial register here, as Lovable is organized via a US parent company. Black Forest Labs. Black Forest Labs from Freiburg develops AI models for image generation and is one of the best-known European AI startups of recent years. Artificial intelligence is also becoming increasingly relevant for telecommunications providers, for example in customer service, software development, or network management. The specific size of Telekom's stake is not publicly visible, as Black Forest Labs has also used a Delaware flip and is organized as a US corporation. Equativ. Equativ is a French adtech company that develops technologies for digital advertising. The business is based on processing large amounts of data and efficiently operating digital platforms. The company thus operates in an environment closely connected to the digital ecosystems of large telecommunications providers. The participation runs through the company's French parent company. Kinexon. Kinexon develops tracking and positioning technologies for industry, sports, and logistics. The systems capture movements and positions in real time and make processes measurable. Such applications benefit from powerful networks and reliable data transmission, a core area of Telekom. According to the commercial register, the group holds around 0.77 percent here. In the end, the portfolio shows above all one thing: Telekom is investing not only in the next mobile communications generation, but also in the software and AI world that builds on it. A mix of German industrial expertise, international cloud software, and European AI.

Build a missed-call text-back system this weekend. You're under a sink, on a roof, or halfway through a job when the phone rings. By the time you dig it out of your pocket, it's gone to voicemail - and most callers who hit voicemail don't leave one. They just call the next name on their list. A missed-call text-back system closes that gap automatically: the instant a call goes unanswered, the caller gets a text from you instead of dead air. It's a weekend build, not a software project, and you don't need a call center to pull it off. What you're actually building. The idea is simple: your business line forwards to a service that can detect a missed call, that event fires a webhook, and the webhook triggers an automation that sends a text back within seconds. The reply can be a fixed template ("Sorry we missed you - text back and we'll get you booked") or something an AI step drafts on the fly based on the time of day or notes from a prior job. Either version beats silence. * A business number that can trigger on missed calls - Twilio is the common choice because its Voice and Messaging APIs both expose webhooks. * A workflow tool to catch that webhook and decide what happens next - n8n, Make, or Zapier all work. * Optionally, an AI step that writes the reply instead of you hardcoding one message for every situation. Build it in an afternoon. The n8n community has already published a working template for this exact pattern - worth studying even if you build your own version in a different tool (n8n Community). The shape of it: * Set up call forwarding. Point your business line, or a new Twilio number, so unanswered calls hit a Voice webhook after a set number of rings. * Catch the missed-call event. Your workflow tool listens on that webhook and receives the caller's number, the time, and the call status. * Draft the reply. A simple version sends a fixed template. A better version passes the caller's number and the time of day to an AI step and asks for a short, natural reply - "sorry we missed you, we're on a job right now, text back and we'll call as soon as we're free" reads very differently at 2pm than at 9pm. * Send the text and log it. Fire the SMS through your provider's API, then write the caller's number, timestamp, and message to a spreadsheet or CRM so nothing slips through a second time. Keep the first version boring. A single well-written template that goes out reliably beats a clever AI-personalized version that occasionally sends something odd. Add the AI step once the plumbing is solid. The one step that trips people up. Before you send a single automated text in the US, you need to register your number and business under A2P 10DLC - the carrier framework for application-to-person text messaging. Skip it and carriers will filter or block your messages outright. Twilio's compliance guide walks through what's required: your business's legal name and EIN (or your own name and SSN if you're a sole proprietor), a description of what you're texting people about, and - critically - proof that the people you're texting actually opted in. If your customers already gave you their number when they booked a job, you're usually fine. If you're planning to text people who never asked to hear from you, don't. Registration isn't instant, so handle it first, before you build the rest of the workflow. Finishing a clean automation on a Sunday night only to discover your texts won't go through until your carrier campaign clears is a bad way to end the weekend. Where to take it from here. Once the basic loop works - miss a call, send a text, log it - a few small additions make it feel like a real system instead of a demo: * Route different messages for business hours versus after-hours or weekends. * Forward the reply thread into whatever you already check daily - email, Slack, a shared inbox - so a real person picks up the conversation. * Add a second automation that nudges you if a caller texts back and doesn't get a human reply within a set window. None of this requires a phone system overhaul or a subscription to an enterprise answering service. It's a webhook, a template, and - if you want it - a short AI-drafted line that makes the first reply sound like a person instead of a bot. Block out an afternoon, start with the boring version, and let it run for a couple of weeks before you tune it further. weekend-projects automation sms twilio n8n missed-calls small-business

How to automate pricing updates on Shopify. Automating pricing updates means automating the workflow around the decision, not the decision itself: exporting sales data on a schedule, feeding it to a pricing model, pushing approved prices back into Shopify through the Admin API, and notifying your team when it's done. The analysis itself takes minutes. Most of the time merchants lose is in the manual steps around it, exporting CSVs, cross-referencing cost sheets, editing product pages one by one. Key Takeaways * Pricing is really three separate jobs: data collection, analysis, and execution. Most merchants automate the middle one and do the other two by hand. * Shopify Flow and third-party connectors can automate the data export step so sales history lands in a consistent format without manual clicking. * Pushing recommended prices back into Shopify has three options, the bulk CSV editor, Flow plus tags, or the Admin API, with the API being the only one that scales cleanly to large catalogs. * Confidence scores matter for automation specifically: only push high-confidence recommendations automatically, route low-confidence ones to a manual review queue. * Automate the plumbing before you trust the model, not before. Run recommendations manually a few times first so you know what a good one looks like. The pricing workflow has three parts. Think of pricing as three distinct jobs: data collection, pulling sales history, cost data, and competitor prices into one place; analysis, running that data through an elasticity model to find the profit-maximizing price for each SKU; and execution, pushing the new prices back into your store, updating internal reports, and notifying your team. Most merchants automate the second part, or let a tool handle it, and do the first and third by hand. That's where the time disappears. | Step | What it involves | Typically automated? | | 1. Data collection | Sales history, cost data, competitor prices | Rarely | | 2. Analysis | Elasticity model, confidence scoring, price recommendation | Usually, by the pricing tool | | 3. Execution | Updating product prices, reporting, team notification | Rarely | Step 1: Automate the data export. Shopify lets you export order history as a CSV, but doing it manually every week gets old fast. Shopify Flow can trigger an export on a schedule, and third-party connectors (Mesa, Alloy, Zapier) can route that file to Google Sheets, email, or a cloud folder automatically. The goal is simple: your sales data lands in a consistent format, in a consistent place, without you clicking "Export." If you also track competitor prices, that's a second data stream worth automating. Competitor sites change layouts, block scrapers, and rotate pricing tiers, so scheduled browser automation is generally more reliable than a one-off script. WebRun's guide to integrating browser automation with n8n walks through setting up scheduled competitor price checks, including how to handle timeouts and poll for results when a task takes longer than expected. Step 2: feed the data into your pricing model. Once your data export is automated, the next step is connecting it to whatever runs your pricing analysis. With Zorin, you upload a CSV of past transactions and the demand model fits automatically, returning a recommended price, expected profit lift, and confidence score for each product. The whole process takes about five minutes from upload to recommendation, the same elasticity coefficient that would otherwise take a spreadsheet and a statistics background to calculate by hand. The key detail is that your export format needs to match what your pricing tool expects. Zorin parses quantities, prices, and dates from your CSV. If your automated export includes extra columns or different headers, add a transformation step, a simple Google Sheets formula or a Zapier formatter, to clean the file before it hits the model. Step 3: push price changes back into Shopify. This is where most workflows fall apart. You have a list of recommended prices. Now what? * Option A, Shopify's bulk editor: download your recommendations, format them as a Shopify product CSV, and upload via the admin. Semi-manual, but faster than editing products one by one. * Option B, Shopify Flow plus tags: tag products that need price changes, then use a Flow to apply specific price rules based on tags. Works for simple adjustments but gets messy with per-SKU recommendations. * Option C, the Shopify Admin API: with a no-code automation tool like Make or n8n, connect your pricing output directly to the Shopify API. The automation reads each row of your recommendation file, calls the API to update the product variant price, and logs the change. No manual uploading. Option C is the most reliable for stores with large catalogs. Once it runs, every recommended price is live in your store within minutes, and you have a log of exactly what changed, which matters if you're pricing an entire catalog rather than a handful of products. Step 4: close the loop with notifications. Price changes affect more than just the product page. Your team needs to know what moved and why. A few things worth automating after prices update: * Slack or email alert listing every SKU that changed, the old price, and the new price. * Margin report recalculated with the new prices and your current cost of goods. * Calendar reminder to review results in 7 to 14 days, once enough sales data accumulates to measure the impact. These are simple automations in any workflow tool. The point is to avoid the scenario where prices changed three weeks ago and nobody remembers which ones or why. What this looks like end to end. A fully automated pricing workflow runs on a weekly or biweekly cycle. Monday morning, your sales data and competitor prices are automatically exported and cleaned. You upload the data to your pricing model, or it pulls automatically, and five minutes later you have recommendations with confidence scores. You review the recommendations, approve the ones above your confidence threshold, and hit go. The automation pushes approved prices to Shopify, logs every change, and pings your team. Two weeks later, you review performance against the old prices, which is also a good moment to revisit how often you're actually changing prices versus how often the model has something worth acting on. The human stays in the loop for the decision. Everything else runs without them. Where merchants get stuck. Two common mistakes show up with pricing automation. The first is automating too early: if you haven't run your pricing model manually a few times, you don't yet know what a good recommendation looks like. Automate the workflow after you trust the output, not before. The second is ignoring confidence scores. Not every recommendation is equally strong. A product with an elasticity R-squared of 0.91 is telling you something reliable. A product with sparse sales data and low confidence is a guess. Build your automation to filter on confidence, so only strong recommendations get pushed automatically and weaker ones go to a review queue. The pricing decision is the valuable part. Everything around it, the exports, the formatting, the uploads, the notifications, is plumbing. Automate the plumbing, keep your hands on the lever. Connect your sales history to see the recommendation side of this workflow running on your own catalog. Frequently asked questions. Can I fully automate Shopify pricing updates without a developer? Yes. No-code tools like Shopify Flow, Zapier, Make, and n8n can handle the data export and price-push steps without custom code, connecting directly to the Shopify Admin API. What's the best way to push bulk price changes into Shopify? For large catalogs, connecting your pricing output to the Shopify Admin API through a no-code automation tool is the most reliable option. It updates every variant directly and logs the change, unlike the bulk CSV editor or Flow-plus-tags approaches, which get harder to manage at scale. Should I automate every price change a model recommends? No. Filter on confidence score first. High-confidence recommendations are reasonable to push automatically, while low-confidence ones, usually from products with sparse sales history, should go to a manual review queue instead. How often should an automated pricing workflow run? Weekly or biweekly is typical, enough time for meaningful sales data to accumulate between cycles without letting stale recommendations sit unused. What format does my sales data need to be in before feeding it to a pricing model? It needs to match what the tool expects, typically quantities, prices, and dates per transaction. If your automated export includes extra columns or different headers, add a cleanup step before the file reaches the model. Do I need to automate competitor price tracking too? Only if your pricing process uses it. Scheduled browser automation is more reliable than manual checks for this, since competitor sites frequently change layouts and block scrapers. What's the risk of automating pricing before trusting the model? You won't yet recognize a bad recommendation when you see one. Run the model manually a few times first, get a feel for what a reasonable output looks like, and automate the surrounding workflow once you trust the recommendations themselves. Getting a pricing recommendation is one thing. Acting on it across 200 SKUs is another, and that gap is almost entirely a plumbing problem, not a modeling one. Automate the data export, the price push, and the notifications, and keep the actual pricing decision, and a healthy dose of skepticism toward low-confidence recommendations, in human hands. Dexter is part of the team at Zorin, building tools that help ecommerce merchants price with data instead of guesswork.