5 Data Science Buzzwords, Revisited

A lot has changed since we first explored data science buzzwords in 2023, with the biggest advancement being the growth in adoption of AI. According to the 2025 Data & AI Leadership Executive Survey, conducted by a partnership between Data & AI Leadership Exchange and DataIQ, “89.0% [of respondents report] that AI is expected to be the most transformational technology in a generation, up from 64.2% in last year’s survey.” GenerativeAI is gaining a lot of ground, in particular. Last year, the survey reported that only 5% of organizations had put GenAI into production at scale, while this year, 24% have done so.

Even more noteworthy, the survey goes on to state “AI is having a ‘halo effect’ on data, with 94% of respondents agreeing that interest in AI is leading to a greater focus on data.” This finding reinforces what we’ve known all along: data is a mission-critical asset that requires investment and attention. 

But when it comes to data, there’s a lot of buzz. Some of the terms are the same as they were in 2023, while a few others have emerged. Let’s take a look at the buzzwords making waves in the market today.

What’s New and Noteworthy
‍As machine learning and large language models (LLMs) become mainstream and the data mesh fad loses steam, we’re adding a few new buzzwords to our list.

‍Entity resolution
Entity resolution addresses challenges related to the reconciliation of records across and within datasets by detecting and matching records that are the same, despite differences in spelling, formatting, associated attributes, and other discrepancies. And while not necessarily “new,” entity resolution is back in the limelight, especially as companies look to further their adoption of AI using clean, consolidated, trustworthy data.

But as many of you know, entity resolution is hard work. From messy source data and the inability to scale to changing business context and missing external content, companies are discovering that the process of resolving entities to deliver trustworthy golden records is easier said than done. Until now. 
‍
Unlike rudimentary tools of the past, AI-native master data management (MDM) solutions combine advanced AI with human feedback to automatically identify and resolve inconsistencies across key business entities and the data sources where they live. With AI-driven entity resolution as a core part of a modern MDM solution, companies can overcome the limitations of their legacy, rules-based MDM solutions by delivering the flexibility to adapt as their needs and strategies change. Said differently, AI-native MDM puts the management and control of data into the hands of people who need it: the data engineers, data consumers, and data stewards who use data to drive business growth.‍

Data governance
‍Similar to entity resolution, data governance isn’t new. But what companies are realizing is that without proper data governance, their efforts to turn data into a mission-critical resource will fail. 

Data governance, at its core, is a set of policies, processes, standards, and roles aimed at enabling organizations to maintain accurate, consistent, secure data as it moves throughout the data lifecycle. Done well, data governance not only ensures companies remain compliant with regulations, but it also enables them to feel confident in their data so they can use it to make informed decisions. For many years, source-based governance was all the rage. But recently, we’re seeing a trend towards consumption-based governance. And this makes a lot of sense. 

Consumption-based data governance arms data consumers with the right tools so they can feel secure in the fact that they are using data in a safe, appropriate way based on the information access policies defined by their organization and the consents of the data owners. And while roles like data scientists don’t own data governance, their productivity relies heavily on how well - or not - it was implemented. 
‍
Agentic AI
Agentic AI is generating a lot of buzz, with some speculating that the market for it is up to ten times greater than the SaaS market. But what, exactly, is agentic AI?
‍
According to a recent article in Harvard Business Review, agentic AI “refers to AI systems and models that can act autonomously to achieve goals without the need for constant human guidance. The agentic AI system understands what the goal or vision of the user is and the context to the problem they are trying to solve.”

Instead of operating reactively, agentic AI is proactive, adopting goal-oriented behavior either defined by users or learned through experience. As new information becomes available – or as circumstances change – agentic AI can adapt, even without constant human intervention. 

When applied in the context of MDM, solutions like Tamr are, in fact, agentic AI solutions. Data teams apply Tamr’s AI/ML capabilities to assess data problems, and then Tamr automatically matches and clusters data and then unites, fixes, and enhances it. This agentic AI approach is much more efficient than traditional, rules-based approaches that require teams of humans to perform a myriad of manual steps just to improve the data. 

What’s Still Buzz-worthy
‍While many buzzwords come and go, some remain relevant, even as the data landscape changes. 
‍
Data products
‍Ask three people to define “data product” and you’ll likely receive three different definitions. But despite the many interpretations of what a data product is and is not, one thing is clear: they are here to stay. 

Tamr defines a data product as a consumption-ready set of high-quality, trustworthy, and accessible data that people across an organization can use to solve business challenges. They’re concrete, recognizable, easy to find, and easy to use, which means users are more likely to realize value from them. 

At Tamr, we believe that data products play a critical role in an MDM program. Not only do they provide a pre-configured, packaged application that delivers the best version of your enterprise data to all consumption endpoints, but they also make master data for critical entities such as customers, suppliers, and healthcare providers tangible for everyone across the organization. 
‍

Tamr’s CEO, Anthony Deighton, said it best:
“At its core, every business is a data business. Which is another way of saying every business should have data products and think about managing their product – which they might think is software or retail or healthcare. But it’s not. It is, in fact, the data. And they should manage that asset like a product.”

‍
Data quality
‍Last but certainly not least, our final buzzword is data quality. While data quality may feel out-of-place on a buzzwords list, we felt it was important to include given the critical importance of high-quality, trustworthy data. 

The challenges associated with delivering a single, authoritative, accurate version of a business entity’s data across multiple data sources and datasets (aka, a golden record), are not new. Despite valiant efforts, millions of dollars, and many years’ worth of time, few companies have actually delivered high-quality data that users can trust. 

So why is that? To start, many companies relied on traditional master data management (MDM) solutions to help them operationalize their data. However, as they’ve learned, this approach is a mistake. By focusing solely on technology, they skipped the valuable steps of assessing, improving, and reviewing their data with the people who know it best. Not only did this oversight exacerbate their data quality issues, but it also made them incredibly difficult to fix. 

Instead, companies should think about MDM as a journey that starts with knowing where you are – and where you want to go. Once they have this assessment in hand, then they can clean the data and put it in front of end users to gather feedback and build trust. Once they confirm their data is trustworthy, only then should companies begin to operationalize their data by connecting it to key business systems. 

As a result of following the MDM journey, companies can finally answer questions like “how many customers do I have,” “which providers deliver the specific set of services I need,” or “what markets have the highest potential for revenue growth?” with confidence. And because they’ve improved data quality before they operationalize it, they’ll realize higher degrees of success. 

As we’ve said before, data buzzwords come with a lot of hype. And while it’s natural to approach the buzz with a bit of skepticism, in the end, much of the hype isn’t just noise – it’s a reflection of the real opportunities waiting to be explored as you embark on a journey to deliver data everyone can trust. It’s up to you to decide which ones are worth embracing – and which ones aren’t worth the hype.