Solving specific problems driving enterprise adoption of AI

Enterprise adoption of generative AI is real.

Some enterprises are deploying assistants and chatbots that employees can use to ask questions of data and receive responses. Others are on the cutting edge and building agents that can act autonomously to surface insights, make recommendations and even take on some repetitive tasks that previously were performed by people.

But those enterprises are exceptions, according to Ariel Katz, CEO of embedded analytics specialist Sisense.

Ariel Katz

Katz took over as Sisense's leader in April 2023, just six months after OpenAI's launch of ChatGPT marked significant improvement in generative AI technology and sparked a surge of interest in generative AI adoption. In his time as Sisense's CEO, Katz has witnessed generative AI go from hype to reality.

However, it's only a reality for certain enterprises rather than the majority, he said in a recent interview. Instead, many are holding off on putting generative AI tools in the hands of employees and turning over processes to machines.

A recent survey by data management vendor Ataccama found that lack of trusted data was one major factor preventing organizations from developing and deploying AI tools. Others have found that lack of technological readiness, cost concerns and insufficient personnel are among other barriers to AI adoption.

Yet some enterprises are successfully developing and deploying AI. One common factor unites them, according to Katz: The enterprises at the forefront of AI adoption have a specific need that AI is addressing, a problem that AI is solving for them.

During the interview, Katz discussed how he's seen AI evolve over the past two years, where enterprise adoption currently stands and how he expects it to evolve. In addition, he spoke about the relationship between embedded analytics and AI, analytics trends in addition to AI such as resurgent interest in on-premises data management, and how Sisense itself is evolving in this new era of analytics and AI.

Editor's note: This Q&A has been edited for clarity and conciseness.

You took over as Sisense's CEO right at the time when generative AI was becoming the dominant trend in analytics and data management. In the nearly two years that you've been CEO, how have you seen AI in the enterprise evolve?

Ariel Katz: One way to look at the evolution is going from machine learning models to create intelligence in specific enterprise domains such as forecasting and recommendations to LLMs [large language models], showing us what it means in the enterprise to interact intelligently with machines and go from insight to action much faster than before. The journey now is about how to enable enterprises that want full transparency about AI -- want to make sure it's governed properly, want to make sure that their data is protected -- and to enable individuals who want to make sure they won't lose their job because they relayed something to their superior based on an AI hallucination. The consequence of making a mistake with AI is orders of magnitude higher than [in the past], so people need to take their time.

As a data platform, the journey for us is to effectively build a bridge with agents that connect the different worlds of data-based human knowledge and AI without hallucinations, creating the trust bridge that connects the two in a way that is governed.

As you look at how enterprises use AI, what is the biggest trend?

Katz: I don't see super-clear patterns that I can relate and say, 'This is what is happening.' There has been a lot of hype that has resonated. But businesses need to have AI [as part of their strategy] to appeal to their customers. We're starting to see the formation of [standards] around AI, which shows that it's starting to become structured. There's ISO 42001, which if adhered to by an organization shows that they're either starting to build their business or relay features to customers using AI, and doing it in a responsible way. I think we'll see more of those [standards] going forward that show you're not only using AI, but also doing it in a way that is enterprise-friendly.

As far as adoption of AI, it's still early days. We have AI in our platform and access to LLMs, and I can see early adoption from some customers, but I can't say that the vast majority of customers are using AI. Among enterprises, unlike consumers, no one wants to be the first. I still see a lot of hesitation, but it's only a matter of time.

It's not necessarily about just being ready. Instead, it's about focusing on a problem to solve. Usually, vision in the enterprise comes from necessity.

Are there common characteristics you're noticing among enterprises embracing AI adoption?

Katz: It's not necessarily about just being ready. Instead, it's about focusing on a problem to solve. Usually, vision in the enterprise comes from necessity. A pattern that I'm seeing is BI evolving to being API-first, where it's about building applications. A few of our customers are starting to use the LLMs we're providing to solve a bottleneck in their organization. They have too few data scientists and data engineers that create models. With LLMs, they can build tools and serve them to front-end users. They can democratize their BI and their insights, and get them to the masses of their users by embedding the tools.

For them, it wasn't a concern about being early to AI. For them, it was about getting dividends from data and pushing boundaries versus their competitors. That's what will push AI to the mainstream going forward. It's about necessity.

Looking ahead to the rest of 2025, how might AI in the enterprise look?

Katz: I think we'll start to see more and more trusted agents. Agents are very relevant. That's where we're seeing enterprise workloads going because there are high-volume and high-value tasks that are too expensive to give to humans. It's about cost elimination and reduction. And more importantly, it's about accuracy and repeatability. Companies are coming up that are very lean in terms of human capital, but are very good at using agent- and AI-driven technology. I see that trend happening already in 2025.

I also think, more and more, we'll start to see agents that create more value from workloads such as sales enablement and supply chains. It will be a shift from a consumption model to a success-driven outcome. That's pretty radical. I don't know if that will happen within a year, but it makes sense as AI takes hold.

Earlier you mentioned embedding AI tools in applications to get them to end users. With Sisense historically focused on embedded BI, how do you view the relationship between AI and embedded analytics?

Katz: They're good friends.

BI is going to be more about how applications are built than about traditional delivery through dashboards and reports. That's fundamental. Embedded analytics reimagines BI so it's everywhere, so insights can get to every person to make data-driven decisions. To do that, developers need to be unleashed to build analytics into applications the same way that communication is being built into applications with Twilio today. That's how you get BI to the masses of users.

AI is going to be present throughout the entire data stack. For example, data modeling has been a big barrier for creators. I would argue that 99% of developers don't know how to build a decent SQL statement, unless it's a very simple one. The way to use AI is to democratize the way they access their data. Based on the intent, Sisense is taking an API-first approach so AI can build a model for developers. It's like Lego blocks. By understanding intent from an LLM, problems can be modeled. LLMs are solving the complexity of data modeling.

In addition to data modeling, what are other areas of the data stack where AI is playing a role?

Katz: Developers don't want to build applications as standalone dashboards. They want to bring analytics to you inside your workflow. [APIs] provide the Lego blocks for developers that allow them to embed analytics wherever they want -- within Slack, Marketo, sales documentation. AI allows you to go very quickly from descriptive mode, personally describing requirements, to the point where you can deliver insight to the application. And it affects everything in between, such as how you bring in data to address a problem, how you enrich data, how you create better models using not just enterprise sources, but also sources like weather and demographics, and how you bring it together to predict something like better customer churn.

Beyond AI adoption, are there other major trends you're seeing in analytics?

Katz: We're asking, what will BI version 2 look like? If you look at what Gartner and others are saying, it's starting to look like an old category with the same players all the time, and no one is shifting into the market. It's starting to feel frozen. Dashboards and reports are artifacts that are 60 years old. But the world is moving. We're talking about data volume [growth] and LLMs. However, BI is still the same, so I believe the next frontier is to bring analytics everywhere and make it a more natural experience in terms of how you interact with data.

There are natural language query and natural language generation, but beyond those, how do you put those in the hands of developers, because they are the ones who are going to connect insights and actions? I think APIs are going to be the next frontier. It's a lot about democratization of data, at the end of the day, and the API-first approach allows that.

How is Sisense responding to AI adoption and other analytics trends?

Katz: Definitely more use of LLMs, and definitely continuing to be smart about enabling both on-premises and cloud deployments.

We really believe on-premises is here to stay. There are many regulated industries that are not ready to go all in on the cloud. They believe in AI, but they want to run AI in their environment. They want to have control over where their data travels. That's not to say they don't use the cloud, but it might be a private cloud, which is more governed than a public cloud. They're not running on old mainframe hardware, and they're not necessarily an old data center, but they're not running on our cloud, and we still need to serve them. They are a very valuable set of businesses.

Trend-wise, if I look at the last two years, most of our new deals are coming in the cloud, but on-premises is not going anywhere. For us, it's about continuing to be modern and cloud-first while serving the on-premises use case in a way that is trustworthy.

Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than 25 years of experience. He covers analytics and data management.