Technologies that will 100% be labelled in some places as AI Agents or Agentic in some places.

01. Simple Reflex AI Agent

“Simple Reflex AI Agent” they’re real, and they are ideal for people who make rules engines to say that they do that, and they really aren’t wrong. A rules engine is something that given a prescriptive set of inputs, applies a prescriptive set of decisions and results in a prescriptive set of outputs. They can even say that they add “symbolic reasoning” to help you create “neural symbolic solutions”, which aren’t untrue statements but context is absolutely everything. So this means that a Rules engine vendor can say they have an “AI Agent” solution, and people who knows the rules engine can now say they are indeed building AI Agents. They aren’t, but the stickers will be there.

02. Workflow Engine

Next up are the workflow engines which can do the same trick as rules engines, but this time focusing on a flow with decision points. Workflow engines have an additional bonus feature that they can say they are “an orchestration framework for Agentic solutions” a statement that isn’t false, but again will mean claims that the product itself is Agentic AI and that the demos you saw 5 years ago were actually agentic demos and that actually the company was ahead of the curve. Not true, but look at those stickers.

03. Traditional ML Vendors

Next up will be the most honest of the sticker vendors, those people who sell ‘traditional’ ML vendors. I say “most honest” in that at least they are selling AI, but in one sense also the most dishonest as they have the least excuse to not understand the actual shift. These folks will claim that Gen AI has “failed” and that for AI Agents to succeed you need to actually embrace developing your own models and doing your own training, and that this is the differentiation that AI Agents is all about. This group will absolutely be championed by Data Scientists who don’t want to learn new ways of working, so its liable to be one of the hardest ones to argue about. The claim will be made that by making the investments in traditional ML you’ll be able to achieve more than those spending billions on Agentic AI frameworks and Gen-AI models. It won’t, but look at that sticker.

04. New Stuff Vendors

The other sticker that is absolutely going to be used is people integrating a 3rd party solution into their solution, charging you a markup and saying that is their solution. This is at least getting you the new stacks and solutions, and hopefully they’ve done a good job of tuning it and guard-railing it for their solution, and hopefully they’ve rearchitected that application or stack to make AI successful. So here I’m absolutely not talking about those people who incorporate a 3rd party solution and have done the investment and re-architecture of their solution to make Agentic AI actually succeed on their platform. I’m talking about how people will “fake it to make it”.

05. Sidekick AI

The first way companies will do this is by slapping on an AI side-kick, something that doesn’t really work within the main transaction flow, but might look like it does because it is integrated into the UI. This sort of solution might provide access to a knowledge base, or provide you with some ‘content [re]writing services’. All this is doing though is making an API call at certain points and doing something that you could do yourself (probably cheaper) and then rebilling it to you with a markup. This AI Sidekicks are the same as cutting and pasting content from the app into Chat GPT and asking it to do something. If its added into the knowledge base then its great that they’ve done that, but this really is just an effective search engine, its not bringing you the idea of digital workers or AI collaboration. But there will be stickers

06. Demo Vendors

Another thing that vendors will do is present a demo, claim its a product and then when you try and buy it you’ll find out it comes with an extensive implementation project, almost like the demo isn’t generally applicable and is actually more of, at best, a framework, and most likely access to a set of people who can build things, sometimes they’ll even throw in a “niche consultancy” who might look like people who’ve just spun out from the software company or might be the people who built the demo, but nooo its absolutely not that. This is a real product that totally exists, it just requires a bit of configuration that looks like an implementation product. But here, have a free sticker.

07. App is just a tool vendor

One of the pieces that gets towards being valid but you need to peel back the sticker a bit is where someone has done some changes and is using a decent Agentic framework and isn’t just building a sidekick they are actually building Agents that use their application and deliver new functionality. The problem becomes if they haven’t made the re-architecture decisions and are just treating their own application as a tool for the AI. Effectively here they’ve created a startup inside their own company to create an agent for their application, without their application providing anything in itself. So here they are hoping that they will bind you to their platform because of their internal startup. It’s a pretty smart approach to take, but you should be clear that is what is being done, because you might get locked into the app because of the AI. But hey, this is a pretty cool sticker

Why spotting stickers is required ?

So the last point is why you should care, because if you don’t you’ll spend money needlessly, you’ll overpay for software and most importantly you won’t come close to achieving the business opportunities. There is a new world where I starts taking decisions and working as part of business teams, this will be the single largest shift in technology infrastructure since the advent of multi-tiered applications. This will transform every sector in significant ways.

But you won’t get there if you are just collecting stickers.