[Blog] Considering AI agents? Note their cost

TL; DR

AI agents may be a hot thing right now.

But in my investigation of their fit for operators, I found at least four challenges that (IMO) currently get in the way of adoption.

Here's the first one: Cost!

What tools did I consider? Why?

AI tools are coming and going fast. By the time you read this, the list I include here should maybe change.

I'm not an expert on the latest list. So I'm really thankful to Ahmet Açar of Attain. He's an AI whiz, who shares loads of insightful content about AI in general and agents in particular. My list of tools picks up his suggestions (as of January 2025). Thank you Ahmet!

Why this list? Because it showcases a breadth of tools, from those suitable to total AI beginners to true experts. It gives you a sense of the spectrum.

The operations leaders for whom I write may typically fall closer to the "beginner" end of the spectrum. They don't need to be technologists for their work–and typically aren't. But some of you may also be quite sophisticated. Who am I to assume? 😄

(Worth noting: I have no relationships with any of the companies listed or any of their competitors. I have no incentives to recommend or not recommend these or any others. I am discussing them based on their publicly available information, ahead of using them myself, based on my own editorial perspective, without influence from others, except where acknowledged. The information is as of February 2025 and may change by the time you read this.)

Price shock is real

Ok, so I found the list of tool suggestions inspirational. Here was not an endless morass of tools but a smart, curated list.

So I dove in, to learn more about the tools ...

only to reel at the pricing.

My first impression, on the companies' homepages, was quite positive. Yes, indeed, these tools should make a big difference for anyone wanting to automate workflows or implement AI agents.

But once I switched to the pricing page, I realized that the "typical" $20/ user/ month cost only applied in few cases.

Yes, the automation platforms Make and Zapier cost that much or less. And so did one of the AI agent-centric platforms, Relevance. But for the other tools, pricing was sharply higher.

(Note: I'll talk more about the difference between workflow automation and agents. But for now, just assume that workflow automation is slightly older tech that you would expect to be a bit cheaper than AI agent platforms.)

Multiply it for your number of users and add any incremental usage costs or higher-priced plans, and this stuff can get expensive fast.

And that's just one tool. You may need several to automate your various use cases. So costs may multiply.

Costs may make sense but be hard for COOs to stomach

So are agent companies just milking users? Not necessarily.

AI agents, by their nature, can cause more server usage and cause higher costs. As premium solutions, they may also command premium pricing, in acknowledgement of their higher value to users.

In other words, it may all make sense and offer good value-for-money.

And of course, you are partly considering this technology to reduce or avoid costs. In comparison to salary costs, this could mean a great deal no matter which tool on this list. (It all depends on framing and what comparisons you use after all. 😁)

But if you're a COO considering whether to bring AI agents to your team's operational work, you may still hesitate. Value or no, this is still a bunch of money, especially while this is all still experimental and you don't have a robust business case yet.

Operations fit matters too

We'll talk LOADS more about the capabilities of AI agent and automation platforms to meet the needs of operators. But for now, here's what you would find too at first glance: An estimate of platforms' agent capabilities and Operational focus, as judged from the self-descriptions that they offer on their websites.

I already have reason to take those descriptions with a grain of salt. But you have to start somewhere!

Based on this, the tools that most explicitly focus on Ops use cases are:

• Make
• Relevance
• Zapier
• n8n

Those four tools differ in their solution approach:

• Make and Zapier are workflow automation platforms. They don't (so far) offer AI agents, as far as I can tell.
• Relevance and n8n offer AI agents. In Relevance's case, that is their focus. In n8n's case, the focus is actually workflow automation. But per Ahmet Açar, the platform also works well for agents. (Haven't tested that yet.)

Finally, there's also the question of how "technical" the solution is, meaning, practically whether one needs to use programming to get things done, or whether the tool is possible to operate via a visual drag-and-drop interface like most websites and apps.

Here:

• Make and Zapier have always been fairly non-technical. That still means that they can get putzy. I've used Zapier a fair bit. In my personal, limited experience, it can get somewhat complicated to get it to do exactly what I want.
• n8n appears to be more technical (require programming).

And so that leaves us with one platform that checks all the boxes:

What to do about costs

Costs should not be your only consideration of course. I'm just splitting them out as one of four roadblocks to AI agent adoption by operators.

But let's say that cost is the key blocker that's keeping you from getting started with agents. If so, I'd suggest the following:

T.I.S.C.

Further reading

Açar, Ahmet. "I spent nights testing these AI agent and automation platforms". LinkedIn. (Jan. 2025)