The latest 'AI Swarms' trend simply means 'doing things at the same time'

The latest 'AI Swarms' trend simply means 'doing things at the same time'
With due thanks to ChatGPT for producing this image!

I've been deeply engaged in a load of research for one of my clients so I haven't had the opportunity to publish much recently.

I'm finding it fascinating to watch the ongoing evolution of the Conversational and Agentic AI space. I can't scroll up and down on my social feeds at the moment without hearing about AI Swarms and how parallel this and parallel that is going to change everything.

Well, yes.

But just to pour some rather cold British water on these wonderfully exciting terms: Let us be clear, we have had the option and the capability to run parallel activity for the longest time.

If we're splitting hairs then fundamentally you're limited by your CPU processing capability. But generally speaking, when we are talking about agentic swarms doing interesting things, we just mean stuff happens at the same time.

For all of those n8n Agentic hackers out there, I can appreciate that the idea of things working simultaneously will sound extraordinary and very exciting – especially when your process flow is built on a linear canvas.

However, I felt it was worthwhile documenting that for the majority of readers here, the base-level reality is that 'AI Swarms' (or, doing things at the same time) is either a feature we've already got in place or something that's already in the process of being built or at the very least, planned.

Ask any Conversational AI vendor about their platform capabilities and they'll be able to explain in some detail what they already do – and what they've got coming.

In many cases, the lack of parallel processing being implemented in production systems by 'Agents' or chatbots is simply down to the technical limitations of many legacy systems being orchestrated.

So whilst it would be wonderful to programmatically do 5 things at once when the customer has asked for a replacement debit card (for example), there is often a requirement to (at the very least) sequence the requests and in some cases, a mandatory need to batch requests.

I doubt we're going to see 'AI Swarm' as a feature request in many RFPs in the next few months, bt I do think it is a good idea to be dusting off your parallel, processing, frequently asked questions and expect to be asked about them ongoing.