AI bill shock is real. Here’s how small businesses should prepare

AI is often positioned as a way for businesses to save money thanks to its automation functionality. However, what isn’t being talked about as much is the hidden costs of AI.

In the same way, many early adopters of cloud technology experienced “cloud bill shock” from unforeseen costs, anaylsts are expecting the same to happen with AI.

As organisations ambitiously scale their AI operations, many may be caught off-guard by hidden costs. With the growth of AI models and datasets, the computational demands and associated costs are likely to skyrocket.

We already know that the hardware and power requirements to run large language models (LLMs) are enormous. And that will trickle down to smaller businesses that utilise the AI tools developed by big tech companies.

You want more AI? That’ll cost extra

“Most SMEs already have AI resources. Many in Australia and the Asia Pacific region are leaning towards Software as a Service (SaaS) solutions and top-tier providers have integrated AI into their offerings,” Dr Joseph Sweeney, an advisor for Intelligent Business Research Services (IBRS), said in an email to SmartCompany.

“Salesforce has Einstein, and Microsoft is in the game as well. HubSpot and other services also utilise these technologies. For most Australian SMEs, AI comes via their existing investments.”

According to IBRS research, there are three types of AI automation that have an impact on workplaces:

• Micro-tasks: These are tasks that occur thousands of times a second, such as those in self-driving cars, robotics, and security systems.
• Decision support: This involves AI combing through vast data to detect exceptions, trends, and make predictions.
• Task augmentation or elimination: The ultimate goal here is to eliminate human involvement entirely.

Dr Sweeney says the challenge will come when companies introduce additional AI functionality at an extra cost. And in some cases, it will be worth it.

“When you consider the savings that come from process automation, the costs often justify themselves. For example, digitising a simple process within a local council setting saves on average $97 (based on 2019-2020 research) each time the process is run, and a more complex process is in $300+ range,” Dr Sweeney said.

Dr Sweeney also said the most common ‘surprise’ expense companies will likely face while scaling up their AI implementations is low-code automation. Because once they get a taste, they’ll want more.

“The uptake, combined with low-code platforms, is swift, and this leads to exponential spending on the cloud services powering these AI functionalities. The challenge here is aligning the operational benefits with the IT department’s costs,” Dr Sweeney said.

And as AI regulations come in, there will likely be some extra costs there, too.

“There’ll be a cost associated with improving information asset management, ensuring that businesses understand the private, sensitive, and confidential information they hold,” Dr Sweeney said.

How to budget for unexpected AI operation costs

According to Dr Sweeney, cloud cost optimisation and cost control models can be useful for SMEs wanting to plan and budget for AI costs.

“The key lies in controlling the integration services costs. As organisations start to explore different AI services, it may save money (at least in terms of consumption fees) to adopt a multi-cloud approach,” Dr Sweeney said.

“However, there are also advantages to keeping all AI services within a single hyperscale cloud ecosystem.

At this early stage of the generative AI boom, many providers are selling at a loss or offering some services for free in order to capture market share. So it’s worth keeping that in mind.

“Once the initial wave passes, prices will undoubtedly rise. The most significant bill shocks will likely emerge from low-code process tools that allow users to build their processes outside the purview of the IT department,” Dr Sweeney said.