Industry executives and experts share their predictions for 2025. Read them in this 17th annual VMblog.com series exclusive. By Kunju Kashalikar, Senior Director of Product
Management, Pentaho
AI and GenAI are rapidly transforming business processes
across every industry. According to a study by AI at Wharton, adoption rates more
than doubled in 2024 and are expected to climb
further in 2025. However, there's a big gap between adoption and
success. Gartner reports that 30% of internal AI projects are abandoned due to
poor data quality inputs. Companies that want to
fully harness the benefits of AI and GenAI in 2025 need to shore up their data
quality to avoid the "garbage-in-garbage-out" cycle that severely limits user
trust and adoption.
There are three main steps companies can take to get their
data fit for AI:
- Improve Data Classification: How data is classified, tiered
and stored is crucial for AI. Not all data is created equal, and just
having all data "on hand" for AI is cost prohibitive and limits potential
use cases. Re-tiering and automating data storage based on usage,
freshness and value frees up costs and resources, giving teams more time
to money to focus on value-added tasks. Strong data classification also
enables stronger governance and security, especially important when models
are being trained on PII and confidential information.
- Focus on Data Observability: Observability gives the business
a single pane of glass to see what is happening with data whether it is at
rest, in motion, in use in applications, tapped for BI reports or in ML/AI
pipelines, helping to avoid any potential quality and usage issues. Especially
with data's dynamic nature, organizations need to approach observability
as an active process and not a static view. If incorrect data enters a
pipeline or model, businesses need tools in place to automatically
mitigate issues based on policies, alert teams to data status changes for
potential downstream impacts and inform real-time response via team
members if needed.
- Adopt Data Products for Scale: Sourcing data ad hoc for
different AI use cases is labor and cost intensive. To scale AI,
businesses need to efficiently create data products and provide those in a
data marketplace shopping experience for defined use cases. Through data
products IT teams can have better control in organizing, classifying, and
ensuring quality data is fed to AI models for more reliable outcomes while
avoiding bias and inaccurate data that leads to mistrust of AI outputs.
Businesses today are in the early stages of
realizing the vast potential benefits of AI. AI success highly depends on the
quality and strength of the data being provided to models. Healthy, fit data
leads to the highest ROI on AI driven projects. We believe that in 2025,
companies will come even closer to ensuring their data is AI fit and that will
lead to more widespread successful AI adoption.
##
ABOUT THE AUTHOR
Kunju Kashalikar, Senior Director of Product Management, Pentaho
Kunju is a senior leader with extensive experience in driving product development from concept to widespread adoption. He leverages his technical expertise, strong development methodologies, customer engagement skills, and proficiency in big data and cognitive technologies to drive results. He specializes in delivering technology leadership by integrating open-source and cognitive technologies, user research, customer behavior analytics, A/B testing, and design thinking to build innovative products for both public and private cloud environments.
He has successfully led globally distributed, multidisciplinary teams to deliver industry-leading products with a significant revenue impact. His work includes facilitating design workshops and collaborating with business stakeholders, end users, and enterprise IT teams to design and implement Data Marketplaces, Data Governance, Data Products
He is passionate about building and mentoring global, cross-functional teams, fostering agile practices at scale, and embracing DevOps principles.