top of page

Post

Expert analysis, insights, and opinions

A Congregation of Thoughts from Enterprise Connect 2025



It was sad to say goodbye to the Gaylord gators yesterday afternoon (pictured above). 


While it is exciting for EC26 to be in Vegas, I will miss my 15 years at the Gaylord Palms Resort & Convention Center (one or two of which were virtual due to Covid). 


Coming back to the Gaylord was always like coming home. I knew all the restaurants and meeting spots; I had memorized the conference center layout and the most effective path through the waterfalls to get from point A to point B.


I also had a tradition for the last day of EC. I would meet up with my aunt and uncle, who live in Florida, and have dinner at the Old Hickory Steakhouse in the Gaylord Palms ahead of catching a late evening flight back home.


And so I found myself saying goodbye to the gators yesterday afternoon between when the Locknote session finished at noon (one of my favorite EC sessions!) and when the restaurant opened at 5:30 pm.


Caffeine as a Catalyst for Thought Congregation

While I will publish additional vendor- and topic-specific articles over the next week, I wanted to share an initial, high-level congregation of ideas related to EC25. Note: A group of alligators is called a congregation, hence the title of this article.


Having arrived home at 1:30 am and waking at 6 am to walk my husky Kai and my daughter's golden retriever Promise, both of whom are often featured in my EC presentations, it seemed this morning was a perfect time, fueled by caffeine, to gather and share a few overall impressions and takeaways from the conference.


In my session this year I discussed "playing fetch with a husky"
In my session this year I discussed "playing fetch with a husky"

Tell and Show

Many of the vendor messages sound the same. Customer Experience (CX) vendors all talk about agent assist, virtual agents, and AI-powered call summaries.


Employee Experience (EX) vendors all talked about meeting and call summaries, and agentic AI.


However, the difference in delivery across the EC25 keynotes was at times stark.


Authentic presentation styles and live (or semi-live) demos resonated more strongly with the audience, compared to heavily scripted and "stoyboarded" demos.


First person, "this is how I use AI", or simply "this is how it works", demonstrations were easier to understand compared to more complex scenarios where fictional characters were described in fictional situations (two keynotes created scenarios where their characters were operating in theme parks).


Articulating core principles, for example "responsible AI", and then showing how your organization translated the abstract into specific product features is more powerful than simply stating a principle that all your competitors also purport to adhere to. The proof is in the pudding and the devil is in the details. Tell but than show how.


Execution can Differentiate

While showing features that are currently available, in addition to talking about them, works for the here and now, as we look ahead, demonstrating the ability to execute on defined roadmaps will set vendors apart.


Scaling simple agents created in demos to production ready agents is not as easy as many believe or portray.


As the number of agents grows, the ability to coordinate (or orchestrate) the behavior of agents becomes critical.


If I ask a question, or define a goal, how will systems effectively and efficiently decide which agents to invoke? (Today some solutions make users explicitly reference a single agent at a time.)


When there are multiple agents that do the same thing, for example two HR benefits agents created by different people, what methods are used to determine the more effective agent to invoke?


When different participating agents return contradictory information, what process arbitrates between the possible solutions?


With respect to data, which as the modern cliché goes is the "fuel for AI", how will organizations effectively manage their reams of corporate documents to ensure out-of-date or incorrect information is not fed into AI systems? And what if a disgruntled employee decides to create erroneous information in order to "poison the well"? How quickly will this information be detected and deleted from customer-built LLMs or customer-indexed vector databases? 


Given the probabilistic nature of AI models, what tools will be used to monitor and ensure agents continue to deliver accurate information, especially as these virtual, and potentially agentic, agents are unleashed externally to serve customers?


All of these problems are solvable but none are easily solvable.


Vendors who invest in R&D, establish testing, build effective partnerships, and have the leadership to develop, prioritize, communicate, and deliver on a publicly shared roadmap will increasingly win both mind and market share. It will not be dreaming up a new "whiz bang feature"; but rather, the ability to execute that will become the key differentiator.


Hybrid Solutions for the Win?

Vendors can deliver CX and EX features but ultimately they need to provide value that exceeds implementation costs (which of course is the idea behind ROI, return on investment).


While full cloud solutions provide great agility, scability, and ease of deployment, it may be hybrid solutions with some components deployed on customer-owned equipment connected to some services in the cloud that provide the best blend of cost-predictability and data security.


Panel Sessions are Popular but can be Painful

It takes a good amount of time and effort to prepare 45 minutes of well-researched content.


I know this directly because Brent Kelly and I spent the last 3+ months working on and refining the content for our session on AI Assistants. And of course, because things change so quickly, we spent the days right before EC25, updating our slides to keep the slides accurate.


We discuss Zoom, Google, Cisco, Microsoft, and Salesforce AI 
We discuss Zoom, Google, Cisco, Microsoft, and Salesforce AI 

I spent a similar amount of time preparing my session that detailed the strengths and weaknesses, and best practices needed to leverage Microsoft Teams and M365 Copilot


Decoding licensing, optimization, and Copilot
Decoding licensing, optimization, and Copilot

In contrast, you can theoretically identify some target panelists, assemble a few questions, and meet for the first time 15 minutes ahead of a panel session.


Then, depending on your panelists, the session will either go well OR one or more panelist may spend 10 minutes giving their bio or answers will involve large quantities of repetition without specifics or depth.


Even when you spend time with panel prep calls, it takes special skill, and a little luck, to moderate an effective panel. As such, panel sessions, while great in theory, can be painful for both the moderator and the audience. Often the greater number of panelists the higher the risk of going off the rails.


Still, I was surprised when a handful of people, unprompted, thanked me for NOT delivering a panel session after my Wednesday morning Teams+Copilot session.


Sidebar -- on Thursday morning I attended a session moderated by Irwin Lazar of Metrigy which was a fantastic panel session. Irwin did a great job moderating and the panelists were knowledgeable, succinct, and very informative: Tabice Ward, Muhammad Muneer, Garth Landers, Shawn Rolin.


Irwin Lazar did a great job facilitating this session
Irwin Lazar did a great job facilitating this session

A fantastic group of panelists discussed AI security and compliance
A fantastic group of panelists discussed AI security and compliance

At the end of this session, I overheard several audience members comment that this was the best panel session of the week that they had attended.


The Congregation of People

For me, EC has always been about the people.


It was a true pleasure to collaborate again with Brent Kelly. I learn something each time I have the opportunity to work with Brent. Even though it is hard work to synthesize as much information as we do into a 45-minute session, working with Brent is something I look forward to every year.



Thank you to the vendor teams from Microsoft, Zoom, Cisco, Genesys, Concentrix, Five9, and more for your hospitality and conversations at various breakfast, lunches, and dinners.


And of course the spontaneous conversations are unexpected treasures. 


Special thanks to Eric Krapf and team for bringing us all together. I am confident that while in 2026 we may be congregating in a new place, the Enterprise Connect crew will once again create an environment to learn, connect, and deepen our industry connections.


See you in Vegas!

bottom of page