Industry executives and experts share their predictions for 2025. Read them in this 17th annual VMblog.com series exclusive.
By Eoin Hinchy, CEO of Tines
As we head into 2025, we’re on the brink of
some major changes. Workflows are increasingly becoming mission-critical, AI
integration is overdue for a reality check and organizations are under mounting
pressure to prove value from their AI investments.
The following predictions capture where I see
the most meaningful advancements coming from in 2025. Critically, the next 12
months will be about making AI actually useful in the enterprise – and that
means proving its value in concrete ways that impact the bottom line.
Workflows will become the foundation of every enterprise
technology strategy
In 2025, workflows
are set to become the foundation that will make enterprise AI truly usable and
meaningful. They’re poised to become the single most important ingredient in
enterprise tech strategies, becoming the essential bridge between AI models and
actual business systems.
Workflows have always
played a big role in enterprise technology. In 2024, many large companies
invested heavily in advanced AI models, but many hit walls that prevented these
models from delivering real value. Now, as we enter 2025, we’re reaching a tipping
point where workflows are on point to become the “glue” that allows AI to
integrate seamlessly with companies’ proprietary systems and data. With
increasing pressure to demonstrate the ROI of AI investments, companies will
need to focus on measurable outcomes and quantifiable benefits.
Over the next year,
more and more companies will rely on workflows to connect to their systems,
authenticate securely, trigger complex logic, and move data between tools.
Foundational models will increasingly drive human decision-making on which
workflows to run-and this is where real traction will happen. So yes, workflows
have always been important, but now they will become the backbone of successful
AI initiatives.
The gap between AI’s promise and reality will widen
without better integration
AI’s progress will
continue, but unless we bridge the gap between promise and reality through
workflows, most companies won’t see the returns they’re hoping for.
In 2024, we’ve seen
the “delta” between AI’s promise and the reality of putting it to work in an
enterprise. Foundational models are getting smarter all the time, but they’re
only as good as the data they can access. In order to see this gap close in 2025,
they should be trained on proprietary business data-the real goldmine that
makes a company successful. This is what keeps AI from fully realizing its
potential in enterprise settings.
The true opportunity,
then, is in using workflows to close this gap. The gap between AI’s promise and
its real-world value lies in seamless integration with workflows. That’s
exactly what we’re seeing with clients who use workflows to connect AI to their
systems. The future of enterprise AI isn’t just about making AI smarter but
about making it relevant through integration.
Transparency will be the key to building trust in AI
Transparent workflows
will be essential to making AI trustworthy over the next year, allowing people
to look “under the hood” and see how decisions are made.
When it comes to AI
trust, transparency is absolutely essential. If users can’t see how an AI
solution came to a certain decision, they’re going to be skeptical about
letting it into critical parts of the business. That’s why workflows have a
huge role to play in giving users a transparent view of each step of the
process. If you ask, “What’s our annual recurring revenue (ARR)?” and the AI
spits out a number, workflows should let you dig into how that number was
arrived at. You’d be able to see which workflow ran, the query made in
Salesforce, and the raw results that came back.
Transparency builds
trust, especially in complex environments. For companies investing in AI in
2025, it’s this transparency that makes all the difference between a tool
that’s useful and one that’s just a black box.
Prompt engineering will become a core skill, not a
specialist role
The role of the
“prompt engineer” will disappear as everyone learns to interact with AI tools
as a basic skill, similar to using Microsoft Excel.
Prompt engineering is
not going to stay a specialized role. It’s a bit like how Excel used to be
considered a specialized skill, but now most people know the basics. Prompt
engineering is just the ability to articulate what you want clearly and
effectively-anyone can learn that. It’s not some ultra-technical task that
requires a dedicated job title.
Over the next year,
prompt engineering will start to feel like an essential skill across the
workforce rather than a niche role. Models are also improving in their ability
to interpret natural language, which makes prompt engineering even simpler.
Soon, I think we’ll all be doing it as part of our regular jobs.
AIOps teams will rise, but fully autonomous AI is still
years away
In 2025, we’ll see
the rise of dedicated AIOps teams to manage AI operations, while the appeal of
“agentic AI” (AI acting independently) won’t pan out as fast as some people are
predicting.
As companies adopt
more AI, there’s going to be a need for specialized AIOps teams. These teams
will handle everything from model deployment to managing quotas and security
for AI-driven workflows, sort of like how infrastructure teams manage data
pipelines. However, the idea of fully autonomous “agentic AI”-AI that runs
without human oversight-will remain a distant reality.
Real-world complexity
will likely prevent us from going fully autonomous in the near future. So many
situations require a human in the loop, especially when there’s potential for
errors. We’ve seen the most success when AI works with humans, each
complementing the other.
Autonomous AI without
human checks has led to more problems than solutions. So, for now, a balanced
approach where humans and AI work together is where we’re going to see value.
The push for
measurable ROI on AI investments will intensify
In 2025, it will no
longer be enough to just “adopt AI”-companies will need hard ROI metrics to
prove its value.
We’re now a couple of
years into the generative AI boom, and I think it’s fair to say that the
technology hasn’t yet lived up to its hype. CIOs and CTOs will demand concrete
metrics before approving new AI investments. Going forward, companies are going
to need hard ROI to justify spending on AI tools. Metrics like “80% of code now
touches AI” or “50% of customer queries are resolved by AI” are going to be
essential.
It’s no longer enough
to just demo an AI solution and assume it will add value. We need quantifiable
outcomes. And the companies that can show hard data on cost savings or
productivity gains are the ones that will actually see AI succeed in their
business.
Ultimately, 2025 will
be about bridging the gap between AI’s promise and its real-world impact.
Companies that focus on integration, transparency and measurable outcomes will
set themselves apart, while those clinging to outdated approaches will face growing
challenges.
##
ABOUT THE AUTHOR
Eoin Hinchy is the co-founder and CEO of
Tines, the trusted leader in smart, secure workflows. Born in Ireland, Hinchy
earned degrees in electronic engineering and computer engineering, then a
master’s degree in security and forensic computing at Dublin City University
and an MBA at Imperial College London. Hinchy began his career as a software
engineer on Deloitte’s security team before joining eBay’s global threat
management team. He rose to lead the company’s European security team, where he
dealt with a data breach that stole 145 million user records.
Hinchy’s experience there and as DocuSign’s
senior director of security operations gave him first-hand knowledge of the
problems that security teams face. In 2018, Hinchy and Tines co-founder Thomas
Kinsella set out to solve those problems by automating tedious workflows,
dramatically reducing the likelihood of incidents and ensuring teams throughout
organizations can respond to any incidents that do occur much faster – and
without writing a single line of code.
