As extraordinary AI becomes available to everyone, the real competitive advantage is what you give it to work with.
We’ve all had those mind-blowing moments watching what AI can do. And they keep coming.
Something that seemed impossible six months ago becomes commonplace. Models get smarter. Accuracy improves. New capabilities appear. The ground keeps moving.
Two years ago, customer advocacy software companies were doing enormous amounts of work just to get mercurial LLMs to produce consistent, accurate results. As a Salesforce partner, we spent the better part of a year trying to do exactly that with Agentforce. It was painful. The toolset was constantly changing.
Frustrating. Even when following best practices, behavior could be inconsistent and performance slow. And occasionally exhilarating, when something finally worked and we got a glimpse of what this new interaction model could become.
But there was no getting around one fact: it put us behind the curve.
At least that's how it looked at the time.
Then Salesforce released its CRM MCP connector and the landscape changed dramatically.
Suddenly, we could securely expose the data and workflows we’d spent more than 20 years building to AI. Within a few months, possibilities that would have required thousands of hours of agentic development became relatively straightforward.
AI didn't just give us new capabilities. It eliminated an enormous amount of development we thought we'd have to do ourselves. In a slightly perverse way, we benefited from letting the AI landscape mature.
And that experience raised a bigger question: If advances in AI can erase a technology gap that quickly, what exactly constitutes a sustainable competitive advantage in the AI era?
Businesses should use existing automation instead of AI reasoning when a process is repeatable, rules-based and already reliably executed by software. AI can provide the conversational interface while established workflows perform the underlying work more consistently and without consuming reasoning tokens.
As we explored what was possible, another issue was becoming obvious. Tokens cost money.
The initial fascination with what AI can do has inevitably been followed by scrutiny of what AI should do, and what it should cost. Token-maxing, out. Token-min-ing, in.
We already have extensive automation built around the most frequent customer advocacy activities: searching for advocates, submitting requests, fielding nominations, routing approvals, managing notifications and tracking activity.
These aren't theoretical workflows. They've been refined over years of actual customer use. And they're fast.
So we don't ask an LLM to reason its way through a process our software already knows how to execute? AI handles the conversation. Our existing automation handles the heavy lifting.
That gives users the fluid, conversational experience they increasingly expect, while mature, token-free software executes processes it has been refined to perform consistently and reliably.
There’s another benefit beyond cost.
A deterministic workflow doesn't wake up one morning feeling creative. It executes the business rules it was designed to execute. That matters when you're dealing with high-frequency operational activities where consistency, permissions, routing and accuracy matter.
The result is a hybrid model: use AI where AI adds something genuinely new, and use conventional automation where conventional automation is already exceptionally good at the job. AI doesn’t need to reinvent every wheel. Sometimes its highest-value job is knowing which wheel to turn.
No. AI changes where competitive advantage comes from. As foundational AI capabilities become broadly available, access to powerful models becomes less differentiating while proprietary data, mature workflows and domain expertise become more important.
If AI continues improving at anything close to its current pace, some of what customer advocacy software companies are painstakingly building today will almost certainly become standard model capability tomorrow. Features that require substantial development today may eventually be handled natively by an AI platform.
Capabilities being promoted as major differentiators now may become table stakes surprisingly quickly.
That's the great equalizer.
If everyone has access to extraordinary AI, AI itself becomes a weaker differentiator.
This is particularly interesting for established software companies. Historically, years of accumulated product development created a formidable barrier to entry. A newer competitor couldn't simply reproduce two decades of functionality overnight. AI changes that equation.
It can dramatically accelerate development, close certain capability gaps and give smaller companies access to technology that previously required substantial engineering resources. We experienced a version of that phenomenon ourselves when MCP dramatically shortened our path to delivering conversational AI experiences.
So what becomes the advantage? The answer is advocate data, deep advocacy program understanding, and the hard-won knowledge of how to translate that expertise into software that actually works.
AI becomes more effective in customer advocacy when unstructured information is combined with well-organized, structured advocacy data. Structured data gives AI context about customers, advocates, products, activities, preferences, relationships and history that unstructured information alone may not reliably provide.
AI is astonishingly good at piecing together unstructured information from call transcripts, emails, documents, surveys, support interactions and other sources. That's opening entirely new possibilities.
Information that historically lived in separate systems can now be examined together. AI can find relationships, identify signals and surface patterns that would have been extraordinarily difficult for humans to uncover at scale.
But the most powerful business applications will combine that capability with something less glamorous:
Well-organized, structured, continuously maintained customer advocacy domain data.
Unstructured data can tell you a great deal about what someone said. Structured domain data provides context about what it means.
Who is an advocate? What products will they endorse? What activities have they participated in? How frequently have they been asked? What topics can they credibly discuss? What relationships exist between the contact, account, opportunity, content and advocacy activity?
That's where AI starts with considerably more to work with.
Structured data gives AI explicit context, relationships and business meaning rather than requiring the model to infer them repeatedly from unstructured information. High-quality structured data can make AI results more reliable, relevant and actionable.
Once you get beneath the AI interface, some stubbornly old-fashioned questions remain.
That knowledge doesn't arrive with the latest model. It comes from understanding a domain deeply enough to know what needs to be captured in the first place.
In our case, that comes from more than 20 years of building technology specifically for customer advocacy, shaped by hundreds of customers, their requirements, enhancement requests, edge cases and changing practices.
We've also revised things along the way. A lot. That's important.
Domain expertise isn't simply knowing how something worked 20 years ago. It's the accumulated learning that comes from watching a discipline evolve and continually adapting the technology around it.
The result isn't simply software functionality. It’s accumulated domain knowledge expressed as data structures, relationships, business rules and workflows. And AI suddenly makes all of that more valuable.
When advanced AI is available to everyone, competitive advantage increasingly comes from what organizations give AI to work with: proprietary data, purpose-built data structures, proven workflows and deep domain expertise. Those assets can make otherwise similar AI models produce very different business outcomes.
For the past few years, much of the software industry has understandably focused on the AI layer itself. Which model? Which agent? Which copilot? Which framework?
Those things matter. But they are also changing at breathtaking speed. Today's breakthrough model becomes tomorrow's commodity service. Today's hard-won AI capability may be an API call next year. The assets underneath AI change much more slowly.
Deep domain knowledge. Purpose-built data structures. Years of accumulated customer data. Proven workflows. Understanding the exceptions and edge cases that don't appear in the glossy demo.
Those aren't as easy to download. And increasingly, AI can amplify their value.
AI is both. It equalizes access to sophisticated technology while increasing the potential value of differentiated data, workflows and domain expertise. As AI itself becomes ubiquitous, those underlying assets become increasingly important sources of competitive advantage.
AI is rapidly democratizing capabilities that once required enormous amounts of specialized development. Newer companies can close technology gaps remarkably quickly. That's real, and established software companies would be foolish to dismiss it.
But AI also amplifies the value of what already exists underneath it: structured data, mature workflows and deep knowledge of a particular business problem. That may ultimately be the more consequential change.
Because when everyone has access to extraordinary AI, simply having AI isn't much of a competitive advantage. The model isn't the moat.
Having something extraordinary to give it is.
.svg){kind=icon}
