Businesses will pay all you want for agent AI, data tools will become a headache, AI agents will become more of a feature than a revolution, and physical AI will matter. These are some of the trends to watch in 2026.
Here are enterprise technology trends and predictions for 2026, organized by confidence level.
High reliability
Agent-based enterprise licensing agreements become the norm due to CxO pushback. In 2025, enterprise software vendors, worried about losing seats, introduced consumable models. The idea was that a hybrid approach would give CIOs the best of both worlds. But the consumption model was unpredictable, and CFOs and even CIOs wanted predictability. Enter your Agentic AI enterprise license.
Miquel Milano, President and Chief Revenue Officer of Salesforce, explained the rationale behind Salesforce's AELA (Agent Enterprise License Contract). “AELA is for customers who are already experimenting. They're ready to scale. They want to go all out, so we agree to a flat fee, so the risk is shared,” Milano said.
ALEA can be used in conjunction with Agentforce and other clouds. Of course, once you're all done with Salesforce, your next deal will allow you to better monetize the platform. We are likely to see similar deals from SaaS providers in 2026.
Here's the problem. SaaS providers can incur losses to ALEA while playing for renewal (if fully locked in). Milano focuses on customer lifetime value. “If a customer is smart, they can rob a bank. They can really make a lot of money out of it. They take risks because they want their customers to be successful. I price AELA in $5 million increments, and the customer adopts so much that the deal becomes too much for me. “I want a customer that's not profitable for me. If that transaction isn't profitable for me, that means I'm the happiest customer in the world. And I have 20 years to monetize that customer.” Milan.
Enterprise data charges and API economics will become a headache. Celonis is suing SAP over data access. In October, U.S. District Judge Vince Chabrier in San Francisco ruled that SAP should take the Celonis case. Towards the end of 2025, The Information reported that Salesforce is increasing the price of apps that use its data. CIO.com noted that these connector fees will likely trickle into your IT budget.
As agent AI is introduced and agents are connected, there will be multiple skirmishes over these data charges in 2026. Standard fees will be required to cover computing costs, but companies must keep in mind that they own the data. In some cases, your vendor may feel different.
Connection charges become a new cloud outlet for moving data. I would argue that data fees pose the biggest risk to scaling AI agents.
Agentic AI is a feature, what it really means is the speed of decision-making. This topic is Michael Ni's division, but the central theme is decision velocity, or how quickly small-scale decision trees and processes can be automated at scale. When Ni said in a video chat that agent AI was just a feature, I had two reactions. First reaction: How many pixels was spent on one feature?!? Second reaction: Mike has a point. The game is about speed of decision making and your mission in 2026 is to put the pieces in place to get there.
“What we're seeing is that the initial decisions are smaller decisions, and behind the scenes there's automation going on that leads to human involvement. Then you start breaking down those decision trees to get to five-fold, 10-fold improvements. We're seeing what leaders can really achieve with decision automation,” Nee said, noting that 2025 was the year many of the building blocks of decision velocity were introduced.
Forward deployment within the company is required. Software vendors are all copying Palantir, but almost every time executives talked about implementing agent AI, they brought up front-line engineers. In 2026, companies will realize they need their own forward deployment engineers to automate data, processes, architecture, and AI. These engineers know more about your business and industry than the engineers you rent from software vendors or service companies.
medium confidence
When the bubble bursts in 2026 or 2027, the AI market will split into two.. What will really grow in 2026 are concerns about capital expenditures and debt for AI infrastructure. But don't get distracted by AI plans. 2026 will highlight how the easy money of LLM training, ridiculous amounts of funding and fantastic remaining performance obligations is over. The circular economy of AI and the AI market dominated by OpenAI are completely different from the enterprise AI version. As AI and process automation converge, the productivity boom is just beginning.
Building is better than buying. The build-versus-buy debate continues, as AI agents make it easier to create applications that were previously purchased. Additionally, as customers begin to push back against the inflation of SaaS transactions, building looks like a great option. SaaS costs are rising every year. It's almost as bad as medicine.
In 2026, there will be a tipping point where companies believe that customized applications for their use cases are the way to go. The sunk costs of enterprise systems are abstracted using agent AI as a user interface.
AI benefits extend to more levels of the enterprise stack. Software vendors will demonstrate revenue and productivity gains from AI. On Wall Street, enterprise software vendors are finally joining the AI rally. Nvidia and AI infrastructure efforts will flatten out in 2026.
Physical AI has its moment. Manufacturing and industrial sectors are beginning to leverage physical AI to deliver real value. Additionally, advances in physical AI will begin to rival early LLMs. This focus on physical AI fosters optimism for robotics and edge AI applications.
Not unreasonable, but unlikely in 2026
- Nvidia stock ended the year flat to down, but the decline wasn't big enough to lead to a real sell-off.. Sales growth is starting to slow as hyperscalers rely more on custom silicon. The industry is making intensive efforts to break out of Nvidia's hardware and software bunker around AI.
- Advances in hardware and software have created an overabundance of AI infrastructure that doesn't require as much computing and energy. Wall Street rewards companies like Apple for not getting excited about building AI infrastructure. AI backlash grows as multiple NIMBY grassroots movements block plans to build data centers in small towns and rural areas. I said something similar in 2025, but to no avail.
- OpenAI knows it can't keep raising money forever They are pushing for fiscal austerity to show they can generate cash flow and make profits.
- With the early arrival of quantum supremacy, quantum use cases are becoming mainstream in enterprises. It also became clear that superconducting quantum computing was the clear technology winner, with the reality creating a mad scramble among companies focused on trapped ions, neutral atoms, annealing and other technologies.
- Meta retools its AI operations after it became clear that a new management team and focus were not producing results. Meta's AI operation is similar to the New York Mets, who pay a huge salary but don't win.
- High memory costs lead to buyer strikes for PCs, servers, and smartphones.
- 2026 will be the biggest year in history. Databricks, Anthropic, OpenAI, SpaceX, and Stripe are all headed for IPOs to go public. Two of these five headliners will be trading below their IPO price after three months.
- Usage of TikTok plummets under new ownership as US algorithm adjusts.
Scorecard from 2025
Here's a look at examples where my predictions for 2025 came true, and other areas where my crystal ball cracked. We have included the probability that the prediction will be performed in 2025.
On target:
- 2025 has been a more volatile year than usual.
- Productivity gains from AI have expanded within enterprises.
- OpenAI and Microsoft have definitely become adversaries.
- Companies will be excited to try new SaaS revenue models. Spending sounds great…until the bill arrives.
- ERP is under fire. ERP is not going anywhere and is becoming an abstraction.
Off base:
- Allegations surrounding vendor consolidation were never implemented.
- Agent AI brings about autonomous processes. We're obviously still in the early stages. The buy-side remains wary of lock-in.
- Building AI data centers will stall. In fact, it accelerated in the days after I made that prediction.
- NVIDIA's growth slows. Nvidia still faces the law of majority and more competition, but 62% revenue growth is pretty sweet.
- Edge computing becomes more important for AI workloads. It will happen, but it certainly won't happen in 2025.
