10 Data + AI Predictions for 2026

I think most enterprise data and AI teams can agree, 2025 didn’t quite go to plan. 

Deploying AI to production was difficult. The P&L impact was low. (“disaster” might be more accurate).

Add to the mix a lack of AI literacy at the executive-level and slowing performance improvements at the model level, and 2025 ended the year marked less by AI deployments and more by AI disappointments. (I’m looking at you MIT report.)

While the headlines might focus on the latest model releases or benchmark wars, it’s clear to anyone using these technologies that better models are far from the most transformative developments coming in 2026.

The real change is playing out in the trenches — where data scientists, data engineers, governance leaders, and AI/ML teams of all stripes are building these complex systems for production. And the headwinds that prevented broad AI success in 2025 will be the market factors steering the ship in 2026. 

So, with that in mind, here are my top ten data and AI predictions for 2026—and what they mean for the next era of AI.

1. Data + AI leaders will rise

If you’ve been on LinkedIn at all recently, you might have noticed a suspicious rise in the number of data + AI titles in your newsfeed—even amongst your own team members. 

No, there wasn’t a restructuring you didn’t know about.

While this is largely a voluntary change among those traditionally categorized as data or AI/ML professionals, this shift in titles reflects a reality on the ground that Monte Carlo has been discussing for almost 2 years now—data and AI are no longer two separate disciplines. And if organizations want to see AI succeed at any level, they need to stop treating it that way.

From the resources and skills they require to the problems they solve, data and AI are two sides of a coin. And that reality is having a demonstrable impact on the way both teams and technologies will evolve in 2026. 

2. AI-ready data will be the biggest topic of 2026

You’re going to be hearing the word “foundations” a lot this year. And there’s no foundation more critical for enterprise success in 2026 than AI-ready data. 

At the outset of the AI-everything bubble, LinkedIn’s carnival-barkers would’ve had you believe that every enterprise with a song in their hearts and a ChatGPT subscription could brute force their way to value with AI. But here at the beginning of 2026—just 2 years into that journey—the talking heads have grown suspiciously quiet. 

That’s because AI will only ever be as useful as the first-party data that powers it. Organizations might have spent the last year ignoring it—but it was only a matter of time before that bill came due. 

The big questions teams are asking today 

• What data does an agent actually need? 
• How do I make it useful for AI?
• How do I govern effectively? 
• And how do I know it’s trustworthy for production?

And to do that, we need to move beyond traditional notions of data quality, to establish tools, and standards, and processes to improve the health and performance of the data pipelines that will ultimately feed our AI. In 2024, Gartner said that 60% of the market would adopt data observability by 2026. It’s 2026, the need to deliver, trusted, governed, and semantically rich data has never been more prescient. (Well done, Gartner.)

Look, AI-ready data isn’t sexy. It won’t deliver a flashy demo or a splashy headline. It’s table-stakes data management. It’s a cost-center. Which is precisely why executives ignored it for so long. But in the face of a 95% AI failure rate, ensuring that your data is governed, trustworthy, and semantically rich isn’t just a nice initiative—it’s an existential priority.

One of my AI predictions is that AI-ready data investments will eclipse investments in agent development for 2026. Foundations are the next (and previous?) frontier for data and AI teams. Data quality programs will evolve to become targeted AI-ready data programs, and AI development budgets will be weighted heavily toward tooling and process over development. And you can bank on that. 

3. Teams will prioritize productivity over pilots

The pendulum is swinging. AI clearly had a value problem in 2025, and the blame rests (at least in part) on the executive team calling the plays.

“We still have a lot of folks who believe that AI is Magic and will do whatever you want it to do with no thought.”

That’s a real quote I heard late last year, and it echoed a common experience for data + AI teams:

• An executive with low AI-literacy sets the priority
• Project fails to provide value
• Pilot is scrapped
• Rinse and repeat

Companies spent billions on AI pilots in 2025 with no clear understanding of where or how their AI would drive impact—and it had a demonstrable effect on not only performance but AI enthusiasm as a whole. That’s going to have real implications going into this next year.

No, I don’t believe AI investments will disappear—I have strong conviction about that—but I do believe those investments will become much more intentional. The days of spinning up random pilots on a whim are in the rearview—and delivering measurable value will be the first, second, and third priorities in 2026. Those use-cases that do get new budget will include a real business case, real ROI, and the infrastructure and operational rigor to support it.

I think tools that deliver meaningful (and measurable) productivity gains will be the areas that see the most attention. Monte Carlo’s Observability Agents are a prime example of an AI use-case that goes beyond speculation to deliver meaningful productivity gains. Choosing the right tool—whether that’s an out-of-the-box agent or just a regular old dashboard—for the right problem.

To state the obvious, this should have been the trend for 2025 already… but you know what they say, sometimes we have to fall in a hole before we can run. 

4. Agent observability will become a non-negotiable for AI deployment

Consumers won’t use what they don’t trust. And the truth is, most agents today just aren’t very trustworthy.

Outputs are non-deterministic by nature. Pipelines traverse systems and teams with limited oversight. And even the smallest issues in data, embeddings, prompts, or models can lead to dramatic shifts in a system’s behavior.

Traditional systems engineering taught us to validate inputs, test logic, and monitor outputs. Yet in AI, that discipline often falls away, replaced by model-centric tooling and disconnected metrics. The reality on the ground is that most readily available solutions loosely defined as “AI observability” only deal with a subset of the problem. They can give you tell you when an output goes wrong, but they’re at a loss to tell you what where or why it happened. As the emergence of data + AI titles suggests, data and AI are one system—and if we want to make them reliable (let alone adoptable), we need to be able to manage them that. That’s means adopting solutions that provide visibility into the inputs and the outputs.

To put it more clearly, instead of chasing internal metrics like model confidence scores or prompt formats, data leaders should anchor reliability on the boundaries of the system:

• Inputs: The data, context, retrieval results, and services that feed the AI system.
• Outputs: The responses, recommendations, decisions, and actions that the AI produces.

In the same way that data observability has become essential for scalable data reliability, I predict that solutions like Agent Observability that unify observability workflows across the agent lifecycle are going to become essential for agent deployments in 2026 and beyond.

5. AI Governance will take shape

Ungoverned AI isn’t just a reliability risk — it’s financial, reputational, and everything in between. A good governance strategy for AI-readiness should address these 5 questions at a minimum:

• What data can we use?
• Under what circumstances?
• What are the risks associated with it?
• Do we have the right documentation to know where the data came from?
• Do we have the right access controls to prevent someone from accessing the data who shouldn’t?
• What can and can’t AI automate? What shouldn’t AI automate?

The problem with AI isn’t just the inputs, it’s the outputs too. People won’t use what they don’t trust. And with the rise in both approved AI failures and targeted prompt attacks, enterprises can’t continue to invest in what they can’t regulate, validate, and ultimately enforce. 

The problem is that AI development has been moving faster than the standards around it. I think 2026 will see the formation and adoption of more standardized frameworks for AI governa

6. Unstructured data will become a first-class citizen

If pragmatism is the tenor of the season, activating nascent datasets is about as low-hanging as fruits come. 

Most AI applications already rely on unstructured data — like emails, documents, images, audio files, and support tickets — to provide the rich context that makes AI responses useful.

But while teams can monitor structured data with established tools, unstructured data has long operated in a blind spot. Traditional data quality monitoring can’t handle text files, images, or documents in the same way it tracks database tables. 

Solutions like Monte Carlo’s unstructured data monitoring are addressing this gap for users by bringing automated quality checks to text and image fields across Snowflake, Databricks, and BigQuery. 

Looking ahead, unstructured data monitoring will become as standard as traditional data quality checks. Organizations will implement comprehensive quality frameworks that treat all data — structured and unstructured — as critical assets requiring active monitoring and governance.

It’s practical. It’s useful. It’s 2026. 

7. The Revenge of the ML model

Given my previous predictions, this one feels like a bit of a no-brainer, but I think it bears stating plainly.

One of the messages in the field right now is that LLMs are powerful… but they’re being used to do things that ML has been doing well for years. And as teams prioritize use-cases over technology preferences, MLis primed to make a spectacular comeback. 

LLMs are incredible. I love LLMs. But LLMs are expensive, their outputs are non-deterministic, and they’re orders of magnitude more difficult to validate than traditional data products. Which means that if you don’t need to use an LLM… well, you probably shouldn’t.

There are plenty of things that you can only do with an LLM—sentiment analysis, relevance scoring, etc. And I think LLMs are going to shine all the brighter in 2026 for those use cases. But you shouldn’t bring a bazooka to a book reading. And heading into 2026, I don’t think nearly as many teams are planning to. 

There’s no denying what LLMs CAN do. The sign of a good AI team? Knowing what they SHOULD do. Like JNCO jeans in the 2020s, ML models are back in vogue. And I wouldn’t be surprised if we see a few more overnight LinkedIn title pivots to own it. 

8. Context engineering will become a core discipline

Input costs for AI models are roughly 300-400x larger than the outputs. If your context data is shackled with problems like incomplete metadata, unstripped HTML, or empty vector arrays, your team is going to face massive cost overruns while processing at scale.

What’s more, confused or incomplete context is also a major AI reliability issue, with ambiguous product names and poor chunking confusing retrievers while small changes to prompts or models can lead to dramatically different outputs.

Which makes it no surprise that context engineering has become the buzziest buzz word for data + AI teams in mid-year 2025. Context engineering is the systematic process of preparing, optimizing, and maintaining context data for AI models. Teams that master upstream context monitoring—ensuring a reliable corpus and embeddings before they hit expensive processing jobs—will see much better outcomes from their AI models.

9. Enterprise teams will choose simplicity over performance

The AI model hosting landscape is consolidating around two clear winners: Databricks and AWS Bedrock. Both platforms are succeeding by embedding AI capabilities directly into existing data infrastructure rather than requiring teams to learn entirely new systems.

Databricks wins with tight integration between model training, deployment, and data processing. Teams can fine-tune models on the same platform where their data lives, eliminating the complexity of moving data between systems. Meanwhile, AWS Bedrock succeeds through breadth and enterprise-grade security, offering access to multiple foundation models from Anthropic, Meta, and others while maintaining strict data governance and compliance standards. 

What’s causing others to fall behind? Fragmentation and complexity. Platforms that require extensive custom integration work or force teams to adopt entirely new toolchains are losing to solutions that fit into existing workflows.

Teams are choosing AI platforms based on operational simplicity and data integration capabilities rather than raw model performance. The winners understand that the best model is useless if it’s too complicated to deploy and maintain reliably. The path of least resistance will claim the crown in 2026.

10. Conversational BI will be hot—but it needs a temperature check

Speaking of programs that drive value—data democratization has been trending in one form or another for nearly a decade now, and conversational BI is just the latest chapter in that story.

The difference between conversational BI and every other BI tool that came before it is the speed and elegance with which it promises to deliver on that utopian vision for even the most non-technical domain users. 

The premise is simple: if you can ask for it, you can access it. It’s a win-win for owners and users alike…in theory. The challenge (as with all democratization efforts) isn’t the tool itself—it’s the reliability of the thing you’re democratizing.

But again, this all hinges on the data being ready for that much access. The only thing worse than bad insights is bad insights delivered quickly. Connect a chat interface to an ungoverned database, and you won’t just accelerate access—you’ll accelerate the consequences.

Data + AI Predictions Wrap Up: The future of AI starts with your data—but it ends with ROI

If too much hype was the fuel, bad data was the match that lit the fuse. Here at the start of 2026, GenAI may well go down in history as the quickest descent into the trough of any Gartner-reviewed technology trend. 

But here’s the truth—that’s not inherently a bad thing. It’s only a bad thing if leaders don’t understand how to adjust course. Remember, that the hype cycle isn’t a measure of a technology’s value. It’s only an indication of how realistic the market’s expectations are about that technology. 

The good news is that when a market finally does move past the hype, we tend to come out on the other side with a better understanding of how to deliver meaningful value. And I think 2026 is the year we’re coming out the other side. 

Yes, decisions will be more measured. Spending will be more thoughtful. But at the end of the year, I believe more enterprise teams will emerge with a meaningful AI deployment, founded on real, secure, and trustworthy data. And that’s a year-end worth celebrating. 

If 2025 was the year the party got a little out-of-hand, 2026 is the year we grow from it. And all that maturing gets me very excited for 2027.

Our promise: we will show you the product.