Agentic CX: Lucidya’s Era of Enterprise AI is already here

Every enterprise I talk to says some version of the same thing about their AI agent rollout: Confident in the boardroom, quieter in production.

 

That gap is not an accident, and it is not closing on its own. Gartner projects that more than 40% of agentic AI projects will be cancelled outright by 2027, citing cost overruns, unclear return on investment and inadequate risk controls.

 

To be clear, that’s not a knock on AI agents, which, if you know what to look for, do their job well. I read the number as a description of what happens when one good component is asked to carry an entire customer relationship by itself.

 

Stop thinking agents, start embracing the ecosystem

 

Most of the agents behind that number were performing the job they were built for. The failure sat elsewhere: Nothing underneath them was sensing what was really happening with a customer, holding their context, coordinating channels, or learning from the last interaction.

 

At Lucidya, we built the platform as an ecosystem: Sensing, understanding, engaging, resolving, and governing, working as one. The agent lives inside resolution, the hardest part to get right, and only as good as the parts feeding it.

 

The rest of the market is reaching the same conclusion from a different starting point. Salesforce, already deep into its own agent platform, signed a $3.6 billion deal this June to acquire Fin, a customer-service AI firm, following earlier acquisitions of a data foundation company and an analytics startup. HubSpot moved the same week to acquire Warmly, built to capture buyer intent before a lead reaches a form. Both companies already had agents. What they were assembling was everything an agent needs to work well: Data, context, and signal. The largest platforms in this category are building, through acquisition, what we built as one ecosystem from the start.

 

Intelligence before the ticket arrives

 

Most AI agents are trained almost entirely on their own transaction history: past tickets, past chats, past complaints. That means they only learn from customers who already made the effort to reach out after something had gone wrong.

 

Our Agentic CX Platform is trained on billions of data points that go well beyond tickets: Public conversation, media coverage, reviews, and market signals across more than 22 channels, the places where customers describe a problem to everyone except the company that could fix it. The most useful thing an agent can know is not how to answer a question asked a thousand times. It’s recognizing the pattern in what customers haven’t actually asked yet; the complaint forming weeks before it becomes a ticket; the sentiment shift before the call volume spikes.

 

The customers who set the real bar

 

This matters most in the most sensitive sectors. We work with governments, banks, telecom operators, and retailers, often in markets where a failure becomes public record rather than an internal metric. A government running citizen-facing AI at national scale has usually already completed the diligence most enterprise buyers are only beginning to formalize. That combination, at that scale, in some of the region’s most linguistically complex markets, is genuinely rare.

 

For a company adopting this kind of system, the difference shows up before a crisis does. Leadership stops measuring success by how fast tickets close and starts seeing what’s forming in the market before it reaches the inbox: a competitor’s misstep, a sentiment shift around a launch, a regulatory conversation gathering steam. Customer teams stop being the last line of defense and start operating with the same forward visibility as market intelligence, because the same data feeds both. Over time, the organization moves from responding well to being rarely surprised, and in regulated industries, that is close to the entire job.

 

Trust you can verify

 

None of this is worth much if a buyer has to take our word for it. This year Lucidya became ISO/IEC 42001 certified, the standard built for AI management systems, alongside compliance with PDPL, GDPR, SOC 2, and NIST’s risk framework. I would rather a customer test those claims directly than take our description of them.

 

The more rigorous test comes during a genuine crisis, when an outage or public safety event requires reaching an entire population within the first hour rather than the first week. That is when the gap between a system built to resolve and one built to respond becomes visible to everyone, including the public.

 

The future this points toward

 

Scrutiny of AI agent claims will keep intensifying this year, and it should. Companies still assembling their ecosystem through acquisition, or still treating the agent as the whole solution, will spend the next 18 months catching up. The ones that built the full system from the start will spend that time further ahead: Closing cases faster, keeping more customers, and winning the regulated contracts that go to vendors whose results hold up when a buyer checks them.

---

By William O’Neill, Chief Revenue Officer, Lucidya

---

Learn more about Lucidya’s Agentic CX Platform here