In conjunction with our AI Product Series data, ETR Insights hosted a panel discussion between technology executives designed to gather feedback on the AI data and commentary around the AI lifecycle. These executives are busy integrating AI services like Microsoft’s Azure OpenAI and Copilot, Salesforce Einstein, ServiceNow’s Now Assist, Databricks’ Mosaic, and Darktrace, primarily focusing on productivity enhancements, cybersecurity, and strategic cost optimization. Encouraged by cost savings and efficiency improvements, these companies are funding AI by reallocating resources from other business areas—though formal ROI can be difficult to calculate—and our panelists warn of frequent disconnects between expected and realized returns, often resulting from internal politics or misaligned use cases. Broadly speaking, organizations are cautious and careful in deploying AI and the lifecycle is still weighed down by pricing concerns, with executives wary of premium charges, token-based pricing models, as well as potential vendor lock-in. Regarding AI’s impact on the workforce, these panelists presently foresee employee “transformation” via reskilling and redeployment, rather than outright job reduction. The discussion included integrating AI within Azure Active Directory; balancing proprietary and third-party AI models; token-based pricing; and the challenges of efficiently scaling AI infrastructure.
Vendors Mentioned: Adobe / Amazon / Anthropic / Automation Anywhere / CrowdStrike / Darktrace / Databricks / DeepSeek / Epic / Homecare Homebase / Meta / Microsoft (Azure OpenAI, Copilot, GitHub) / OpenAI / Palo Alto Networks / Proofpoint (Tessian) / Salesforce (Agentforce, Einstein) / SentinelOne / ServiceNow
Executives are ramping up investments in artificial intelligence, with spending increases reflecting growing confidence and a clearer vision of AI’s role in enhancing business efficiency. One CIO at a major enterprise described the company’s integration of Azure OpenAI services, now providing employees secure, personalized access to advanced language models. “It’s linked in with our Azure Active Directory. It knows who you are. We can pass other sort of metadata tags through from the Azure AI end, attributes, which is great.” Here, AI’s impact spans human resources, employee benefits, and cybersecurity; the company has also purchased Microsoft 365 Copilot licenses for about 10% of their staff and regularly uses ServiceNow Now Assist and Salesforce Einstein. “Our cybersecurity staff, we’ve been doing things in AI—probably unknown to people—for many, many years. We use systems like Darktrace and [Proofpoint’s] Tessian. They’ve got AI and machine learning built into them from how they were originally set up many years ago.” His company projects a 10% to 20% increase in AI-related IT spend in 2025, after seeing buy-in at the Board level, and is exploring more sophisticated applications of AI. “We’re looking really seriously at things like agents now, to take the assistants to the next level.”
“2024 was more of a pause year for many enterprises,” notes another CIO. “There has been a shift in this time frame in the past three months, that we’re going into much more of a regular investment year, going back to the trajectory that we saw in 2021 to 2023, a significant set of years.” Organizations are funding AI initiatives through cost optimization, Cloud Ops initiatives, and other architecture redesign and capital expenditures aimed at long-term gains. Despite broader economic pressures—including declining revenues—some continue reallocating existing resources towards AI initiatives on its demonstrable cost-saving benefits. “The AI budget has increased by taking away from some other parts of the business, to be honest,” says one Head of Data Science. “AI is bringing in a lot more value and helping us save those costs.”
ETR Data: In the November AI Product Series survey, 40% of respondents indicated plans to increase AI spending by 11% to 20% over the next 12 months. The most recent distribution has seen shifts, with some revising their plans downward into the 1% to 10% increase range, and others revising upwards to the 21%-30% and 31%-50% growth intervals. Overall, 75% of respondents plan to increase their AI spending by 1-30% next year. Only 4% indicated a decrease.
One panelist favors a balanced approach; they have consolidated onto Microsoft Azure and Databricks, but are also investing directly in proprietary models. “Specifically on AI, we’ve been working quite significantly on Mosaic, a product from the Databricks stable, as well as now more of the talk being around agentic AI, from Q3 of last year. This is everything from Agentforce for Salesforce, to ServiceNow’s Pro Plus tier.” Despite the predominant focus on text-based solutions, this team also anticipates significant developments within images and video. “Of course, there are pros and cons to it in terms of both upfront investments, as well as OPEX funding that you would have on both of those. But really, I see this as a positive for the ecosystem, where we are seeing more development, and this is really going to raise the floor for all capabilities within the foundation model space.”
While half-jokingly acknowledging that the product is largely just an update of Einstein, two organizations represented in this panel have been actively integrating Salesforce Agentforce, particularly now given its more open API. “[The API] makes it much more viable as a product or something useful for us, as building models in Einstein Analytics wasn’t the easiest solution. Putting everything in Salesforce Data Cloud doesn’t make sense, because then you take away the entire data lake concept that you’re building.” It is important to note, however, that this company is still enjoying discounted Salesforce preview rates. “I’ve looked at the offerings that we will have to pay eventually. It does seem to be something we’ll have to have a discussion or negotiation with Salesforce about, because we are not using the full components, and we’re never going to.”
Another executive cautions against inflated expectations: “When you do a POC the ROI is higher, and when you scale the ROI tends to drop, because your infrastructure costs tend to go up.” Internal departmental politics complicate ROI calculations, while poorly conceived initiatives often lead to disappointing results. “I’ve had a lot of business departments picking up AI use cases or partnering with external consultants, that six months down the line, their projects are not working out, the costs haven’t justified the returns, or there are no returns.”
Broadly, our panelists acknowledge AI’s potential impact on jobs, but they emphasize transformation rather than reduction. “My feeling is that with the agentic AI that’s going to start really hitting this year, I think we’re going to see what I would call ‘job impact,’ people either being redeployed, or people going away from the sort more mundane, repetitive tasks that they’re doing now.” Another executive outlines a strategic approach involving reassessing, retraining, and updating workforce skills to adapt to AI’s changes. “In the longer term, I think there are going to be more beneficiaries of these newer job personas and job profiles.” In healthcare, “Until we get the Westworld-type robots, our goal is to make our clinicians more efficient than any sort of staff reduction.”
ETR Data: Respondents in the AI Product Series survey disagree that AI spending and initiatives will slow hiring or reduce headcount, with around a tenth of respondents strongly disagreeing. Compared to the November survey iteration, there is growing disagreement that AI will slow hiring (from 2.9 to 2.8) or reduce headcount (from 2.8 to 2.7).
As businesses grapple with AI pricing models, executives are disinclined to pay premium prices for embedded AI features. One compared these pricing strategies to Amazon’s approach. “You get people to get used to it, and then jack up the prices.” Several raised concerns over the long-term viability of current pricing strategies. “There are incentives that are added on for adoption for many of these tools, but once those wean off, what are you going to pay as a real price for these and how is that going to be structured?” Another panelist believes that “Tokenomics,” or pricing based on token usage, requires scrutiny. “We’ve been having a lot of internal discussions around our token spend, and about how we want to become more compute efficient on our own infrastructure as well.”
One CIO worries about forced adoption of costly AI features, and their company might miss out on valuable functionality if they refuse to pay for premium offerings. “Before you always had the road map that would show you, here’s what’s coming with Adobe as an example, and now it’s going to be, am I losing some feature-rich functionality?” Another described frustration over additional charges for features he feels vendors should inherently provide. “I think one of the reasons we get it for these big enterprise solutions, is because we take advantage of their R&D, their ongoing road map, and their ongoing feature set. And to be fair, I sort of feel a little bit cheated when I get asked to pay for additional features.”