One of the more interesting use cases for generative AI is their application in AI “agents”, or apps that utilize AI to actually undertake tasks for you, and automate outcomes.
OpenAI recently moved into this next phase with its “Operator” agent project, which can interact with other websites to undertake specific tasks assigned to it. So, for example, if you wanted to find the best prices for home-delivered vegetables, then have them arrive at your door, without you doing any of the leg work, Operator can actually do that on your behalf.
It’s still something of a work in progress, but conceptually, the next stage of web interaction could be bots talking to websites, and finding the best deals and offers that align with your interests.
Which is just one of the ways in which dedicated, AI-powered agents will be able to undertake tasks that have traditionally taken hours (well, a couple at least) of research.
Meta highlighted another interesting use case for the same last week, in a post about how soccer (or football, depending on where you’re from) club Sevilla FC has developed a custom AI-powered process called “Scout Advisor”, which utilizes Meta’s Llama model, as well as IBM’s Watson, to uncover potential football talent, based on a range of parameters.
As explained by Meta:
“To make recruiting decisions, Sevilla FC needed to assess qualities like attitude, tenacity and leadership across a large volume of scouting reports. Without Scout Advisor, recruiters had to spend 200 to 300 hours analyzing a single shortlist of players, but now Sevilla FC’s recruiters can simply ask Scout Advisor a question about the soccer talent they’re looking for to see a list of matching players, including precise AI-generated summaries of their performance. IBM’s watsonx and Meta’s Llama enable Sevilla FC to bridge the gap between traditional human-centric and data-driven scouting in the identification and characterization of potential recruits.”
Again, that won’t be perfect, but it’s interesting to consider the potential of custom, confined AI-powered tools, that can facilitate more specific tasks, and save many labor hours by honing in on smaller elements within a given process.
The same could be applied to social media marketing, and uncovering prospects via AI tools.
To test this out, I asked ChatGPT to create a social media prospecting app that would enable me to create custom lists of potential prospects based on their social media posts, information in their bios, their most recent activity, location, demographic info, etc.
It gave me a Python code for an app that will scan social platforms (within each’s limitations) and analyze user content, then provide a dashboard to visualize and filter that information.
I ran a simulation in Replit, which gave me a basic overview of the data it could produce:
Now, this is X-specific data, which may not be as valuable as it once was in this respect, while this is also a very basic example, and not at the level of being an “agent” as such. But you can see how easily you could create a more complex system based on these tools, which would then enable you to automatically add prospect data to your database, provide a summary of key discussions within your niche, highlight emerging tangential trends and provide content suggestions based on them, or even generate entire creative options as a result, etc.
With access to relevant API keys, and some basic coding knowledge, you could build a custom prospecting tool that would save you a heap of research time, based on your own defined parameters that relate to your target clients.
That may not be for everyone, while there’s also evidence to suggest that relying on each platform’s AI tools themselves may deliver better outcomes than restricting them with your own prospect parameters. But the point is that AI agents will soon provide more capacity to get more done, in less time, within very specific, custom settings.
Which is a much more valuable use case for AI than, say, generating images of yourself as a medieval knight, or an alien, or whatever else Meta AI’s dumb in-stream prompts suggest.
And while these systems are still prone to errors, or “hallucinations” as the creators prefer, you can see how these tools, in the right hands, and honing in on the right information, could be hugely valuable in many contexts.
The trick is in data access, and getting live information from each platform at an affordable price for your business. That’s not going to be easy, with data access getting more expensive every day, but dedicated AI agents could provide all new ways to monitor the media, and uncover opportunities.
