Qlik's agentic AI study - 97% have budget, 18% have deployed. Here's what works

Qlik has released its third annual Artificial Intelligence (AI) study, conducted with Enterprise Technology Research (ETR), capturing how enterprises are navigating the transition from AI experimentation to agentic implementation. Based on a survey of over 200 enterprise technology decision-makers, the report shows 97% of large enterprises have committed budget to agentic AI — with 39% planning investments exceeding $1 million — yet only 18% report full deployment.

In a discussion about the findings, James Fisher, Qlik's Chief Strategy Officer, says:

I'm still surprised by the gap that we see between expectations and commitments to AI and particularly agentic AI, but still a lack of adoption across the enterprise. Lots of good talk, but turning that into measurable action that is actually causing a benefit or resulting in a benefit, is still not where I think it should be.

This represents a significant evolution from Qlik's 2024 research, which showed 37% of organizations had formal AI strategies. That figure has now jumped to 69%. However, 46% of respondents believe it will take three to five years to operationalize at scale.

Budget pressures and fragmented funding

Fisher frames the deployment gap within broader enterprise constraints:

IT budgets overall are consistently under pressure in the enterprise, and when you're trying to increase investment, exponentially increased investment in AI and agentic solutions when generally budgets are flat, or even some cases declining from an overall enterprise IT perspective, then you're having to start making difficult decisions and trade-offs.

While 56% have dedicated budget for AI innovation, 60% are still pulling from IT/Technology budgets, and 42% from line of business funds. A further 79% agree agentic AI is critical to their organization's three-to-five-year strategic plan.

Data foundations and skills gaps

Data quality, availability, and accessibility emerged as the primary barrier to adoption, cited by 56% of respondents. Perhaps most revealing: while 77% claim confidence in distinguishing agentic AI from other tools, only 42% believe their organization has the internal expertise to design and deploy without external support. Fisher says: 

It has been a persistent barrier, even pre-agentic, as a persistent barrier to, alongside data literacy, to good use of analytics, good use of traditional AI, good use of generative AI and now as a foundation for agentic.

Integration with existing systems ranks second at 49%, followed by lack of internal expertise at 48%. Only 13% mentioned multi-agent systems when defining agentic AI; most focused on autonomous decision-making (61%) and task automation (49%).

Fisher takes a measured view:

This is still relatively nascent in terms of the grand scheme of things, and that transference of skills out into the wider ecosystem takes time. I don't think we should be overly surprised or overly critical that we've got an opportunity to develop skills.

Security and governance

Cybersecurity vulnerabilities topped deployment concerns at 61%. IT Operations is cited as the primary target area for implementation by 72% of respondents. Legal and compliance exposure ranks third among concerns at 51%, while 47% cite lack of explainability and auditability.

Fisher emphasizes how the stakeholder landscape is evolving:

Cyber security teams, data security teams and governance teams have always been involved. I think there is now an increased seat at the table for the one that you didn't mention, which is the legal teams.

On governance, Fisher observes:

The starting point for all of this comes down to policy. That sets the foundation for the governance of how all of our people work with AI in doing their job, whether that's delivering a piece of technology, or whether that's producing content, or whatever those use cases may be.

Where deployment works

Following our conversation, Fisher provided customer examples illustrating successful deployment. The common thread: each built AI on existing data foundations rather than attempting transformation.

A North American specialty chemicals distributor stood up a generative AI assistant for sales and customer service within two months, connecting it to existing document repositories in its Qlik environment and SharePoint. Roughly 40 people now use it daily for natural-language queries about complex product data. The company reports it can now hire commercial talent from outside the chemicals industry, confident AI will bridge the product-knowledge gap during onboarding.

A global industrial manufacturer with 3,500 employees took a similar approach with unstructured technical content – technical manuals, project documentation, and internal notes. Setting up each new knowledge base took approximately 15 minutes plus indexing time. The speed convinced leadership it could scale AI use without a large data-engineering project upfront.

An Asia Pacific entertainment group operating cinemas and theme parks improved attendance forecast accuracy from roughly 70% to over 90% using predictive capabilities layered onto their existing Qlik Cloud Analytics deployment. Those forecasts now drive labor scheduling and operational planning in near real-time.

A European food producer built AI-powered demand forecasts for premium organic meat products, reducing forecast deviations to around 1%. This cut over-production, reduced costly downgrades of organic meat to conventional, and lowered storage costs while supporting sustainability goals.

Fisher says:

Right now today, to prove value from AI and agentic, I believe we have everything we need. The question is, how distributed is that across the Enterprise Architecture? How distributed is that across the vendor ecosystem, and what does it cost for an organization to pull that together?

He adds:

When you're holding a hammer, everything kind of looks like a nail. I think we're slowly beginning to figure out which pieces of technology to use in which particular use cases.

My take

I keep going back to the headline numbers – 97% budget commitment, 18% deployment. What did the 18% do differently?

The customer examples suggest an answer. A chemicals distributor gets 40 people using AI daily within two months. A manufacturer sets up knowledge bases in 15 minutes. An entertainment group improves forecast accuracy from 70% to 90%. A food producer cuts forecast deviations to 1%. None required multi-year transformation programs or resolved every data quality issue first. They connected AI to existing data foundations and proved value quickly.

The definitional confusion may explain the wider gap. Only 13% mentioned multi-agent systems when defining agentic AI – arguably its distinguishing feature. Most conflate it with autonomous decision-making or task automation. If organizations don't understand what they're buying, it's harder to scope deployments that deliver.

Fisher emphasizes policy and governance as starting points. The customer examples add another dimension – bounded scope. The chemicals distributor didn't attempt enterprise-wide transformation – it connected an AI assistant to SharePoint and existing Qlik data in weeks.

After three years of research showing persistent implementation challenges, these examples offer rays of light for organizations willing to start small.