In this video Mohammed Adris recounts how, over the past year, he built several bespoke AI agents for his recruitment agency—automating job‑ad publishing, finance dashboards, candidate‑client matching, and dispute monitoring. He then observes that the very SaaS tools he pays for (Microsoft, Google, LinkedIn, his CRM) have begun shipping native AI agents that duplicate these custom workflows. Using an Amazon Marketplace analogy, he explains that platforms harvest third‑party usage data to create competing features, leading to tighter API limits, bundled AI, and higher subscription costs. He forecasts that within one to two years most generic recruiter‑focused agents will be built‑in, making unique agency processes the only sustainable competitive edge. To mitigate risk, he advises agencies to own/export their data, codify proprietary processes, prune overlapping AI subscriptions, and maintain the ability to swap tools quickly. He also highlights the looming rise in cloud‑infrastructure costs and the shift toward pricing AI as a subscription add‑on. Finally, he promotes his small coaching community, recruitment hiring/upskilling program, and interest in acquiring recruiting businesses, inviting viewers to contact him via LinkedIn.
Chapters
Over the last year, I've built my own AI agents for my own recruitment agency. One writes and publishes every job advert in our CRM so recruiters don't have to do it. One builds a weekly finance dashboard that's taken straight out of Xero. Another one tells us which clients a CV should have gone to, but it actually didn't, and it reminds the recruiter. And then another one scans through our CRM to see where we have submitted CVs and haven't had feedback, and then monitors them for potential disputes. And then one by one, the tools that I already pay for started launching their own AI agents— Microsoft, Google, LinkedIn, even my CRM. And I've been running my own recruitment agency for over a decade now, and I've watched a lot of software trends come and go. This one feels a little bit different. So in this video, I'm going to tell you why it's happening, what I think happens next, and what I'm going to do about it so it doesn't cost me my entire business. 2 years ago, if you wanted an AI agent, you paid a software company to build it for you. Today you can build one yourself in an afternoon. And that one change has every software company you use worried. So let's look at what's happened over the last 18 months. People stopped Googling things and started asking ChatGPT. So Google launched AI mode, same search bar, but now you talk to it like as if you're in a chat. Then Google Chrome got Gemini built in. Some of the things that you'd open Claude for, Google wants you doing inside the browser instead. Microsoft put together Copilot, which now works inside Outlook, Excel, Teams. And from July this year, they put Microsoft 365 Business prices up and folded Copilot Chat into the base plan. So you're paying for it whether you ask for it or not. And for us as recruiters, LinkedIn launched Hiring Assistant, an AI agent that sits inside Recruiter and does sourcing and outreach for you. So Why is all of this happening all at once? Well, it's because the brain isn't theirs anymore. GPT, Claude, Gemini, anyone with a card can use them. You plug one into your CRM, one into your inbox, one into your outreach tool, all through API and MCP connections, and you've built your own software. And I've watched founders who I've worked with, with zero coding background, build a sourcing agent in a weekend. And from the software company side, The biggest threat now is you. If you can build the agent, you can build the workflow. And if you can build the workflow, you can start asking why you're paying for a seat for the parts you've already replaced. Interesting question, because software companies aren't racing to beat each other anymore. They're actually racing to stop you from leaving. And if you really think about it, Amazon has a habit. It lets sellers prove what sells, then launches its own version. And I think your software is about to do the same thing with your AI agents. So I just want to explain that, that how it works with Amazon. First off, anyone can sell on Marketplace, and the sellers, they do the hard bit. They find the product, they test the price, they build the reviews, and they prove that there is demand for a given product. And in 2020, Wall Street Journal reported that Amazon staff had used data on individual third-party sellers when developing Amazon's own private label products. You might know them as Amazon Basics. Amazon, they said that would break its own internal policy. 2 years later, in the EU, Amazon signed commitments with European Commission not to use non-public seller data for its own retail business. Whatever you think of the story, the model is simple. Someone else proves the demand, The platform moves in with its own demand, ranks their own product a little bit higher, and then sells it a little bit cheaper. Now put that on your own software. Every time you connect an API key or an MCP connection, that's a door between your agent and their platform. And I want to be precise here because people get this wrong. They can't see your prompts inside Claude, but they can see what you pull, how often you pull it, and which apps are connected. So if 10,000 users are all pulling candidate data at 7 AM in the morning and pushing back a shortlist, they know exactly what people are building. That's free product research. And the next step is a native version built in, no API key, one click bundled into the plan that you already pay for. I'm not saying that any one company copied another person. All I'm saying is that the pattern is the same that Amazon ran. So here's my forecast. Within the next year or 2, most generic AI agents that recruiters are building right now will exist natively inside tools that they already pay for— summarizing CVs, writing job adverts, drafting outreach. If that's your edge, then it has a shelf life. So if your AI agent does something that every recruiter would want, assume that your software will ship it. So the best thing to do is build the bits that only your agency would want. And look, if you can take everything in this video and work out your own AI setup, you'll probably get there. But it takes time, and you'll build things, the platforms will give it away for free, Next year. So I want to mention this quickly. I have a very small coaching community that I've been helping. I'm also hiring for recruiters in my own team who we also upskill with AI, and I'm also looking to acquire recruiting businesses to help strengthen my own team. So if you're stuck with any of the stuff that I'm talking about or you want a little bit of handholding help, then there's a few ways that you can reach out to me. I'll leave some details below where you can contact me on LinkedIn. So let's crack on with it. So here's the part that the AI hype videos, they tend to skip, and that is that somebody has to pay for the electricity. Now, it's just funny that I mentioned that. Just think about this for a second. The 9 biggest cloud companies are on track to spend around $830 billion on infrastructure this year. And the international experts around data centers, energy agencies specifically, they believe that energy usage for data centers is going to double by 2030. That's about 945 terawatt-hours. I only know that because I work with HV engineers. And I want to put that into perspective. That's a bit more than the whole of what Japan uses today. So power, cooling, chips, land, fuel— none of that is going to get cheaper anytime soon. And right now, a lot of AI is priced to win users, i.e., making money comes a little bit later. None of these platforms— GPT, Claude, whatever— none of them are profit-making companies at the moment. Their stock shares are going through the roof, But nonetheless, they're not making any profit, they're not making any money, and that's when prices move. So how do they get it back? We've already seen it. If you put AI in the subscription and raise the price, that's basically the move that Microsoft made in July. Now I just want you to think about that— why your software would rather you use their AI than your own. Because when you plug in your own Claude account or whatever other platform that you use into their platform, you pay Anthropic. They don't get anything extra. But when you use their built-in agent, they control the cost and they keep the margin, and your own agent is basically lost revenue to them. And what this means for an agency owner is pretty simple, is that your software bill goes up whether you touch the AI or not. And I'd expect API usage to get more expensive too. So just do an audit in the moment. If you're paying for AI in Microsoft, in Google, in LinkedIn, or a cloud subscription, and you're paying for all of that layered on top of each other, then you're paying for 4 times for roughly the same kind of brain. And I know everybody has their preferences and they work in slightly different ways. But pick where your AI lives and switch the rest of them off, especially if you can, because you're going to pay for AI whether you see it or you use it or not. So just make sure that you only pay for the bit that you actually use. And then the third thing I think happens is one that actually worries me a little bit. The doors you use to connect your own AI basically start getting smaller, and this has already started. If you go back to '23, Twitter, Reddit started basically charging serious amounts of money for API access. In 2025, Salesforce changed Slack's API terms to restrict bulk exports of Slack data and stop it being used to train AI models. And the pattern actually becomes consistent because once data becomes valuable to AI, then the platforms, they typically change or just lock everything down so that you can't get access to it. So my forecast is basically tighter rate limits and separate AI usage, either pricing plans, API keys, and some connectors only on premium plans. And LinkedIn has never been relaxed about automation. So recruiters will feel this first. So where am I putting my bet? Literally in my own businesses, 2 places. First is the data. Our candidates and client data has to live somewhere we own and can export. If your CRM is the only place your data exists, you're renting your business. You need to have a backup. And the second is that the process, which is the AI agent, is now cheap, and building it and maintaining it and running it as it is, is right now the cheap part of the work. However, What actually took me years at Responsum is knowing what good actually looks like on a desk, like which CV goes to which client, how a candidate gets qualified, when to pick up the phone instead of sending a message. A platform update can't copy that because it lives in how we work and they never see it. So that's one of the areas. The other part is that I'm trying to keep my software subscription small. If a vendor shuts down or shuts its doors or stops us using API access, unless we pay some stupid amount of money, then I want to be able to swap out that tool in a week, not take a year before I can actually fix the problem itself. And another thing is that the platforms, they can copy your agent, they can't copy how your agency thinks. So that's my view on it. That's the high-level version. And a lot of this comes down to one question, which is which tools are worth paying for and which ones are basically just charging you and waiting before they increase the price and then they're going to milk you. I go through this exactly, what I use, what I'd cut, and I do that in this video. So if you're wanting to stop paying 4 times for the same thing, then watch this video and I'll see you over there.
