Editor in Chief of Constellation Insights
Constellation Research
About Larry Dignan:
Dignan was most recently Celonis Media’s Editor-in-Chief, where he sat at the intersection of media and marketing. He is the former Editor-in-Chief of ZDNet and has covered the technology industry and transformation trends for more than two decades, publishing articles in CNET, Knowledge@Wharton, Wall Street Week, Interactive Week, The New York Times, and Financial Planning.
He is also an Adjunct Professor at Temple University and a member of the Advisory Board for The Fox Business School's Institute of Business and Information Technology.
<br>Constellation Insights does the following:
Cover the buy side and sell side of enterprise tech with news, analysis, profiles, interviews, and event coverage of vendors, as well as Constellation Research's community and…
Read more
Enterprise AI is evolving at such a breakneck pace that it’s tough to keep up. AI, notably agentic AI, generative AI and the mutations to come, is rewriting business and society in real-time. Constellation Research's AI Forum in Silicon Valley surfaced a bevy of themes about enterprise AI.
Here's a look at what matters right now.
What DeepSeek taught us?
Dheeraj Pandey, CEO and Co-Founder at DevRev, said DeepSeek's surprise reset expectations in the AI and technology market. "DeepSeek busts the myth that America innovates, Europe regulators and China imitates," said Pandey. "You can take a big thing, make it small, and it's disruptive."
Karen Silverman, CEO of Cantellus Group, said DeepSeek illustrates how assumptions can change overnight. "For many years we've been staring at that exponential curve stick straight up. That's what DeepSeek feels like. Overnight there can be a complete shift in assumptions," she said.
DeepSeek moments will also have repeatedly. "I think this is going to happen again and again," said Silverman. "We need to saddle up."
Future of models
Models will need to replicate more than language. "The idea that models are going to replicate human interaction is absurd. We need to start thinking more completely and contextually. When you think about how humans interact, language is just 10% of how we communicate," said Silverman.
Models will become more of a hierarchy of services, as well as adaptation and innovative memory systems. "I think the next move is going to be low data," said Silverman. The idea of massive training and expense is going to be diminished.
More from AI Forum 2025:
Security and AI applications
AI will need to be developed for security. Yaron Singer, VP of AI and Security at Cisco, said bad actors can extract training data from a model, put instructions in a document that the models can learn, and do a lot of bad things at scale. "We are spinning off a team to build state of the art AI for security applications," said Singer, who was CEO at Robust Intelligence that tested vulnerabilities inside of AI applications and added protection.
Attack surfaces with AI will become much larger. Singer said:
"What I think is interesting about AI security specifically, is that not only that the attack surface becomes much larger, but the nature of AI makes it such that the solutions are very different. You can very easily manipulate the classification piece and then trigger operations, right that cause unwanted behavior. It's so non-deterministic that traditional security solutions that we have just don't apply."
Cisco is creating reference architectures to secure RAG and integrated security systems for the AI stack. As technology evolves and innovations like reasoning models emerge so will the threats.
Singer said he was bullish on open models over proprietary, but acknowledged that there may be more security risks.
He added that the biggest security worry is that generative AI models will develop their own protocols that become risks. Singer said that the risks will become even more magnified as AI codes.
Herding AI agents
Agents will have to work together on tasks and automate across business processes, but orchestration layers need to be built. "We're all in on agents being the future," said Ed Macosky, Chief Product & Technology Officer at Boomi. "But the orchestration layer needs to bring them together into something intelligent."
Will AI agents replace humans?
Walter Sun, Global Head of AI at SAP, said agents in the enterprise will serve to do repetitive tasks and be a junior assistant for each persona. "You still need a human being to address exception handling," said Sun.
Seema Swamy, Head of Insights and Data at Walmart, said: "I actually see a very good role for AI in optimization, because there are so many things that humans cannot do it as effectively." In tasks that are less about optimization and efficiency, Swamy is concerned about biases and transparency. AI is data and efficiency, and humans are creativity and innovation.
Jana Eggers, CEO of Nara Logics, agreed especially in marketing where enterprises can have more standardization in branding. Humans will take on more creative roles with AI focused on optimization. "Stop focusing on making AI superhuman, and start focusing on making it super. Stop trying to make AI replicate us," said Eggers.
Wannia Hu, VP of Product Management at UKG, said AI agents are about freeing up humans not replacing them, but the burden is on the employer to alleviate fears. "It is unfair to place all of the mindset change on employees. Call on employers to think about reskilling, upskilling, and making that part of the day to day. It's absolutely critical," said Hu.
David Levine, Founder of PlanDataAI, said AI agents won't replace humans, but the fear is palpable. Levine describes AI as evolutionary and the human element is critical to implement genAI. "I'm going to say AI is evolutionary not revolutionary. If you tell people that they're more open to it," said Levine.
Hu said the biggest risk is entry level jobs where college grads are expected to come in at a higher level with little training ahead of time.
AI investments
Barney Pell, Venture Partner at Radical Ventures, said "I'm seeing AI coming to every niche, and now I'm seeing companies being evaluated on how good their AI is."
"AI is big business and the whole stack," said Pell. "There's this constant motion of companies and teams coming out of the bigger companies to start companies. The innovation will all come at the application level."
Efficiency as AI starts to run companies
David Giambruno, CEO of Nucleaus, is a master at cutting costs (as noted on CCE 2024). He said:
"The world is going to change super-fast, like super-fast. Once you automate your infrastructure, your infrastructure is not your code. Your code now runs everything that is from a bottoms up view. We are starting to train the AI to run companies. We're going to go through a Darwin event. You see the magnitude of change and the ability for the systems to run themselves, it is absolutely fascinating."
Sunil Karkera, Founder of Soul of the Machine, said voice interfaces will be the UI of the future. That reality will also mean less overhead while expanding margins. Karkera's company is built on AI agents.
Giambruno said that there will be casualties due to efficiency. He was decidedly less optimistic than others due to following the money. He said:
"I'll be blunt, right? 50% of the people don't make it. What I do is as much as a technology exercise is a financial exercise. It's not the technology that's hard. It's the financial engineering."
What's next for agents, adaptive inference?
Denise Holt, CEO of AIX Global Media, said AI will ultimately learn and adapt through adaptive inference. In a nutshell, adaptive inference means that pretraining won't be necessary because AI will learn more like a biological system. Holt highlighted active inference at CCE 2024.
She said:
"This means that IoT and AR, VR, all these emerging technologies now become interoperable, so IoT becomes sensory information for these agents to understand this live data. It's unstructured. They can now seek to make sense of it. They have a grounded understanding of the actual real physical world as it's changing over time. So they don't rely on pre trained data sets. They can learn in real time from real data, and they can learn and share this information with each other in the exact same way. So then you have this knowledge layer that begins to grow and evolve as a collective intelligence."
Barbat Hodjat, CTO of AI at Cognizant, said the intersection of language models and the transformer training model and active inference will be critical to scaling. The problem is far from solved. "There are some really interesting interdisciplinary approaches," he said.
Advanced AI compute and data centers go specialized
AI's power problem is being worked on, but the focus is more on choosing the right workloads for power. Sunny Madra, COO, President GTM, Operations and Supply Chain at Groq, said: “When it comes to energy consumption, we're not heading towards more efficiency or less energy. In fact, we're increasing the amount of energy. We look at that in terms of the racks that people are deploying and what they're going towards.”
Niv Zilberman, COO Aquatron DC Partners, said the industry is moving from general purpose processors. “We’re moving from general purpose compute and I'm just going to throw more compute at any problem into being workload driven. I think it is really optimizing not just the power and efficiency, but being able to deliver the accuracy in the right scale.”
AI as process automation tool
A panel of CFOs and finance execs outlined how they are using AI and in some cases process automation has driven cash flow improvements. The key takeaways:
- “Finance is a conversative function, but AI can be used to automate transactional-based functions. AR (accounts receivable) and AP (accounts payable) are the best AI use cases,” said Cathleen Nilson, CFO Xsolla. “AR and AP are a no brainer for any company of any size.” She added that AR and AP efficiency also improve cash glow for enterprises.
- Isabelle Wang, CFO Legion, is using AI to analyze the customer base and due diligence for internal project approval.
- Arnulfo Sanchez, Chief Accounting Officer at Datastax, said AI will eliminate monthly and quarterly close cycles. “Closes will be every day,” said Sanchez.
The panel wouldn’t use AI for tasks that require compassion, deal structure and deciding on big contracts and acquisitions.
The macro picture for AI
Navin Chaddha, Managing Partner of Mayfield, outlined the firm’s investment theses. He explained how AI is a once in a generation opportunity. Key points:
- “Our belief is AI is essentially going to team up with humans. I agree it will be humans in the loop to make them super humans.”
- “We think we are entering an era of collaborative intelligence, where AI will work with humans to change the way we work, live and play as humanity. Yes, in the short run, there'll be pain. Anytime a new technology comes up, there is pain, and the focus till now has been on, hey, let's cut costs. Let's improve efficiency. But what we are seeing with the new startups that are coming up, they’re actually being very smart. They are going after jobs that are not filled, jobs that humans don't want to do, or jobs that humans can't do.”
- “There are 30 million neglected small businesses. They can't hire knowledge workers. Now, suddenly, AI is an AI teammate. The technology is called agents, but the form it appears itself in, we call it a teammate, and that now expands the market, and it creates endless possibilities. Half the US economy is either 1099, or works in small businesses and now they can afford knowledge workers, which happen to be in the form of AI, and you pay for them when you use it.”
- The AI stack will be consumed by humans but also humans. “Every line of business function is going to have an AI teammate. Our thesis is AI teammates is a massive opportunity,” said Chaddha. “Today, the value is increasing to the hardware line, but it's going to very quickly move up to the app and the user layer.”
Data to Decisions
Future of Work
Next-Generation Customer Experience
Innovation & Product-led Growth
Tech Optimization
Digital Safety, Privacy & Cybersecurity
AI
GenerativeAI
ML
Machine Learning
LLMs
Agentic AI
Analytics
Automation
Disruptive Technology
Chief Information Officer
Chief Executive Officer
Chief Technology Officer
Chief AI Officer
Chief Data Officer
Chief Analytics Officer
Chief Information Security Officer
Chief Product Officer
