Feature
Generative AI is going to take a "cold shower" in 2024.
That's one of the takeaways from UK-based CCS Insights most recent set of predictions on the connected world and its landscape for next year. The 85 predictions covered in the report focused on personalized experiences, collaborative convergence, sustainability and of course, artificial intelligence (AI). The emergence of generative AI unsurprisingly was a big focus, but according to CCS, the optimism of recent months across the industry is starting to be replaced with a little more caution.
Generative AI Slowdown
“Generative AI has a cold shower in 2024 as the reality of cost, risk and complexity replaces the hype of 2023. The hype of 2023 has ignored several obstacles that will slow progress in the short term,” the report reads.
“The cost of deployment is a prohibitive factor for many organizations and developers. Additionally, future regulation and the social and commercial risks of deploying generative AI in certain scenarios result in a period of evaluation prior to roll-out.”
Other signs of a generative AI slowdown have become apparent in recent weeks. Gartner reported in August that the technology had reached the Peak of Inflated Expectations, the step just before the inevitable decline into the Trough of Disillusionment.
Industry commentators have also stated it could be as long as five years before productivity gains from generative AI become apparent. The question for enterprise decision makers then, is whether the high costs of investment are worth it and whether their organizations can actually afford it. With Microsoft Copilot and Google Duet both costing an extra $30 per user per month, the question isn't theoretical.
The Nuances Around Generative AI Costs
CCS Insight's predictions around the sobering 'reality check' for generative AI in 2024 ring true to the complexities and challenges this cutting-edge technology presents. However, the question of affordability is a nuanced one that cannot be pigeonholed into a binary answer, Peter Wood, CTO at Spectrum Search, said.
The initial outlay for integrating advanced AI systems can be daunting, Wood continued, but what needs to be factored in is the long-term return on investment.
The value created by automating menial tasks and generating deeper data insights can free up human resources for more strategic roles far exceeds the upfront costs, he noted. Moreover, advancements in AI have the potential to open up new revenue streams and business models that can offset the deployment expenses.
The landscape of AI affordability is rapidly changing, he said. With the advent of cloud computing and more accessible AI toolkits, smaller companies can now take advantage of 'AI-as-a-Service' platforms, thus lowering the barrier to entry. “The democratization of AI is well underway, making it increasingly feasible for businesses of all sizes to integrate these technologies without breaking the bank,” he said.
He also points to the potential costs of not implementing AI responsibly. Future regulations around AI ethics and data governance are likely to impose financial penalties and reputational risks that could far outweigh the costs of initial deployment, he said.
The question, then, is not just whether companies can afford to deploy generative AI, but also whether they can afford the potential fallout from not implementing it thoughtfully. The costs are not merely financial but extend to societal impacts and ethical considerations.
Related Article: Are You Giving Employees Guidelines on Generative AI Use? You Should Be
Additional Generative AI Costs
The costs associated with using generative AI aren't straightforward and are often multi-layered, said Anna Wang, CTO and co-founder of Searchlight. The most obvious cost being the API costs of using foundational models from language labs like Open AI and Anthropic.
“I believe that this cost should rank low in the list of concerns,” she said. “Just as most business these days don't think twice before trusting and paying cloud providers like AWS and Google Cloud to power critical infrastructure, businesses will simply fold in the cost of using generative AI APIs as part of their R&D budget.”
Businesses also need to take into account costs around positioning and competitive differentiation. If they're currently leaning heavily into using generative AI, they have a crowded landscape of offerings to choose from, many of which look the same.
In order to commercialize generative AI products and return profits, businesses need to consider how their generative AI offerings provide more value than what is free and publicly available.
A third source of cost ambiguity relates to compliance with generative AI regulation. This bucket has the biggest possibility for ballooning by far. However, it takes time for this regulation to take shape and for enforcement to begin, so Wang doesn't see this as an immediate dampener on generative AI adoption.
Rather, Wang predicts the cost of using generative AI will get lower in 2024 as the amount of tooling and public resources around using generative AI continues to grow.
“Instead of needing to hire specialized talent around ML, deep learning and other such technologies, most businesses can transfer existing engineers on their R&D teams into investigating this new technology while keeping headcount costs relatively flat,” she added.
Regulation Is the Wildcard
Although the costs of training LLMs is high, many companies won’t have to train their own LLMs, said Charles Chow, Lumen Technologies head of marketing for Asia Pacific.
“There is an extraordinarily high cost associated with training large language models. While few organizations could afford to deploy a personalized AI tool, the reality is they probably won’t have to,” he said. “Recent research shows that smaller, task-oriented AI tools are much cheaper to develop, train and deploy.”
Instead of developing large-scale models that can “do it all,” he believes that in 2024, the focus will change to honing AI tools and models to smaller use cases. “Instead of a ChatGTP model, we’ll see specific platforms that help platforms optimize a small part of their business. While cost will always be a factor, these will fade with time as we improve the AI training process,” he added.
He shares Wang's belief that regulatory issues is the real wildcard here. Without concrete legislation, AI companies have been self-regulating. Some of the LLMs developed by industry giants aren't even bothering to disclose training data sources, he notes. When governmental legislation does land, the AI landscape will have to shift to meet it.
“This shift will force companies to be much more cautious with how they use AI and where they deploy it. Instead of the hype of 2023, we’re looking at a careful, delicate, and regulatory-dependent 2024,” he said.
Webinar
Sep
25
CMS Briefing: A Live Look at What’s Next in AI-Driven Platforms
Learn how leading organizations are using AI‑driven tools to publish faster, personalize smarter and stay secure.
Register
Webinar
On demand
Ready or Not: How Data-First Organizations Are Unlocking Agentforce Potential
Learn how to cut through the noise, activate Agentforce and build a Salesforce AI strategy that actually delivers.
Watch Now
Webinar
On demand
AI in Customer Service: Faster Resolutions, Happier Customers
Don’t let rising demand burn out your team. See how to build a smarter, more resilient support org.
Watch Now
Webinar
On demand
From Hype to High-Impact CX Strategies That Actually Scale
Turn buzzworthy AI and outsourcing trends into measurable CX wins with fresh 2025 data.
Watch Now
Webinar
On demand
Insights to Action Rethinking the Contact Center for Real Business Impact
Join our exclusive webinar to hear CX executives share their innovative strategies for transforming service delivery.
Watch Now
Webinar
On demand
Accelerating Healthcare Ops with AI
Sign up to see how your peers are putting AI to work — and making it stick.
Watch Now
Related Article: AWS's Diya Wynn: Embed Responsible AI Into How We Work
Decisions, Decisions
Generative AI is primed to have a great year in 2024, even as we careen towards the trough of disillusionment, said Ryan Ries, generative AI and machine learning practice lead for Mission Cloud.
Experimentation is already well under way as people test the potential of the application with free services like ChatGPT, Bing and Bard. The costs start entering with the next step, as businesses customize generative AI solutions for their own applications.
But businesses have a number options out there to deploy these kinds of solutions, ranging from using API calls from tools like Amazon Bedrock or spinning up models with tools like Sagemaker Jumpstart, Ries notes.
The cost to run these systems depends on the number of users. An effective deployment to a small number of users could cost as little as $20 a month for 1 million tokens up and down, he said. When the number of users grows, he said the conversation should shift to deploying foundational models in your own ecosystem. Small companies with many potential users should probably consider whether they can use a smaller specialized foundation model, he added.
The currently available models are here to stay, he said, so even with any upcoming regulations most customers will not be disrupted or have to worry about losing all the capabilities they just built out.