Industry executives and experts share their predictions for 2025. Read them in this 17th annual VMblog.com series exclusive.
By Zuzanna
Stamirowska, CEO and Co-founder, Pathway
Operating in an emerging technology, such as
artificial intelligence (AI), is exciting, and the charging pace of change over
recent years has brought ample opportunities. It cannot be forgotten that even
as we enter another year, AI is in its relative infancy, and the rate at which
new techniques and uses are being uncovered is unlikely to slow down.
In 2025, we will see the impacts of what I
believe to be the biggest disruption to AI since its inception. It will also be
crunch time for those leading AI transformation in businesses as exploration
budgets are replaced by pressure to make returns on investment (ROI). We go
into the next year equipped with a clearer idea of what AI can and can’t do,
allowing greater focus on reasonable, feasible use cases and making deceptive
buzzwords and jargon a thing of the past. These are my leading predictions for
next year.
o1’s
ascension
The transformative impact that OpenAI’s o1
model is going to have on the AI, data and tech scene is not being widely
spoken about now, but it is set to be the biggest shake-up since the beginning
of AI. Its capacity for deeper reasoning and complex problem-solving is
something that no other player has a comparable solution to just yet. 2025 will
be the year that users get to grips with the best applications for the enhanced
abilities of o1, but its full potential likely won’t be realized until further
down the line.
o1 certainly isn’t the be-all and end-all of
large language models (LLMs) as it is slow and expensive, so there is plenty
of space to innovate. The fight to beat o1 is sure to be the
industry’s next great battleground. It will be particularly interesting to see
how smaller solution providers seize this opportunity.
Live AI
will help overcome scaling issues
Data management, along with privacy and data
updates, is the most pressing obstacle that organizations are facing as they
scale AI. Data management challenges aren’t always clear in the prototyping
stage but cause significant headaches when moving to production. This is
because businesses operate in complex and dynamic environments and to gain
maximum value from LLMs, they must harness hordes of constantly evolving private company
data by contextualizing, indexing and retrieving it at
operational speed.
Enterprise budget holders are going to
scrutinize ROI on AI projects more closely in 2025 as projects move beyond the
proof of concept (POC) stage. There will be less leeway for projects stalled
due to snags that weren’t clear in prototyping, so businesses will have to
explore better approaches to take LLMs into production. Live AI is therefore
likely to become a greater focus than ever before as it will overcome common
data management limitations and allow businesses to deal with fresh structured
and unstructured data in smarter ways, leading to more powerful outcomes.
The key
is in the use case
Much of the hype around AI until now has been
based on what various architectures can do and how powerful they are. In 2025
we will see a progression past this mindset to a focus on finding the
techniques that deliver the best results with the data and infrastructure
available, especially in organizations
like banks which deal with sensitive data hosted on-premises and
have taken a long time to find ways to allow external systems to interact
around firewalls. We will see organizations maximizing their investments in AI
by locking in use cases, doing extensive persona work and considering the end
user’s journey through the technology.
There is an understanding now that AI
transformation needs to move at a greater speed. Businesses will therefore
increasingly focus on identifying feasible use cases facilitated with the
resources they have right now before developing the governance policies that
enable use cases to be expanded securely.
Bye
buzzwords
More businesses are hiring leaders who
understand the abilities and limitations of AI right down to the code and this
is helping disperse the culture of jargon masks that has become all too common
in AI transformation projects. With more dedicated and experienced leaders,
buzzwords and exaggerations will become harder to hide behind.
As a result, we will see success from the AI
solution vendors who act as collaborative, accountable partners that provide
real results, while those that don’t deliver will disappear.
Agility
and authenticity in 2025
As AI enters a new stage of development in
2025, agility and authenticity are going to be key. OpenAI’s o1 brings new
levels of reasoning and a new era of exploration, but now our view of what AI
can do is much clearer, leading to greater transparency and tangible outcomes.
Success will favor those who can seamlessly integrate advanced technology with
their existing datasets and iron out data management issues before entering
scaling phases.
##
ABOUT
THE AUTHOR
Zuzanna Stamirowska is CEO and Co-founder of Pathway, the
company building live AI systems that think and learn in real-time as humans
do. Customers, including NATO, La Poste, and a Formula 1 team, trust Pathway to
lay the foundations for AI transformation without needing to replatform.
