Industry executives and experts share their predictions for 2025. Read them in this 17th annual VMblog.com series exclusive.
By Dom Couldwell, Head of Field
Engineering EMEA, DataStax
It’s no surprise that analysts are bullish on
Gen AI – Omdia predicts that spend on GenAI applications will
reach $58 billion by 2028, while Gartner found that GenAI would triple sales of servers
to power these systems from 2023 to 2028. However, while the companies that
provide GenAI are all investing heavily ahead of perceived demand, where are
the enterprise deployments that demonstrate where and how that GenAI spend will
make that significant difference to companies?
Prediction
#1 – Generative AI is on track for enterprise deployments
While Generative AI is due to attract huge
amounts of spending over the next few years, the number of production
deployments is still small. The amount of hype going on is significant, but the
market is certainly beginning to ask about results and real world projects.
It’s important to remember that Generative AI is still in its early stages of
adoption, with most businesses yet to launch their first production-grade
applications.
It’s also important to remember previous
big technology trends and the gap between new innovations being launched and
mainstream deployments by enterprises. Typically, enterprises took between two
and four years to take those innovations and build their own applications or
launch new services. As an example, Apple launched the App Store in 2008, and
it took until 2009 for Uber and 2010 for Instagram to launch. Mobile enabled
Spotify, Airbnb and Uber to disrupt the music, hospitality and transportation industries
and change how we think about and consume those services.
I expect 2025 will be the year when we see
companies actually launch and have to make good on their promises around AI,
both internally and to the market. Those companies that do this successfully
will see huge market impact.
Prediction
#2 – Gen AI will see the toolset it needs come through
Part of the reason why enterprises are
slow to experiment and deploy new technologies is that they don’t have the
tools and support systems in place to scale up. Gen AI technologies are still
developing with lots of innovation around Large Language Models, but the
returns are diminishing from the LLMs themselves. So the opportunity to
innovate around GenAI will come from how applications are designed and built,
rather than from the models themselves. For enterprises, they want to see how
they are building, the performance that they see back and how to measure those
AI-powered experiences over time. Without these tools, they are not willing to
launch into full-scale production deployments, as they can’t see where problems
exist or provide the audit trail for compliance or security.
Linked to this there is a lack of
expertise around Gen AI out there – finding the right people that can manage
and scale AI deployments is hard, simply because the number of people out there
with real-world experience is small. It’s a combination of those used to
working with very large enterprise level data sets for millions of users, and
those that have built on LLMs to design services. This is currently a very
small set of people, but it is growing.
To be ready for Gen AI, businesses need
better tooling, architecture, and observability systems to integrate AI
solutions effectively. This will come through in 2025.
Prediction
#3 – AI chatbot launches will succeed … or not
For most enterprises, their Gen AI budgets
will mainly be going to support services and customer applications. It’s easy
for AI influencers to denigrate AI-powered chatbots, but anyone who takes this
approach has never launched this kind of service in the real world. It is a
tough challenge involving getting data from knowledge bases into other
conversational content platforms, and then making that experience better than
older pre-programmed chatbots could deliver. You might think that is a low bar
to set, but the reality is that most people have not enjoyed those old-style
chatbots and they have to be convinced that GenAI can deliver too.
For enterprises, getting over this user
experience hurdle and delivering a service that thousands or millions of
consumers can use is a significant one. The goal is to enhance user
interaction, streamline information access, and improve support and engagement
through conversational AI interfaces. The prize here is happier and more
engaged customers, as well as reduced cost to support or sell to those
individuals over time. There will be successes in 2025, and there will continue
to be those that have problems due to poor model choice or attempts to break
models out of their guiderails.
For enterprises, these deployments will
have the eyes of the C-Level team upon them, as they have seen the hype on
spending and potential return and they want to see results. Tempering those
expectations and designing services that deliver real world results, either
through reducing costs, freeing up budget for re-investment, or through totally
redesigning the business process or workflow, will all come through in 2025.
Prediction
#4 – More developers will get hands-on with RAG
Retrieval Augmented Generation is becoming
ubiquitous in enterprise GenAI projects, because companies want to use their
own data. However, those projects also have to deliver to users and meet
specific use cases. This is where developers are having trouble, because they
need to increase the relevancy and reliability of their applications and they
have to think about the cost to deliver the service as well.
Over the next year, more and more
developers will take on GenAI services and look at how they can build these
services into their applications. This will help to expand the number of
developers with skills around RAG and GenAI, democratising AI and making it
more accessible. Once this takes place, those GenAI use cases will move forward
into production. This will help developers meet those user needs and find the
opportunities to build new applications that will meet what their CEOs are
looking for.
##
ABOUT THE AUTHOR
Dom
Couldwell is Head of Field Engineering EMEA at DataStax, a real-time data and
AI company. Dom helps companies to implement real-time applications based on an
open source stack that just works. His previous work includes more than two
decades of experience across a variety of verticals including Financial
Services, Healthcare and Retail. Prior to DataStax, he has previously worked
for the likes of Google, Apigee and Deutsche Bank.
