For years, logging into martech software meant entering a dashboard. Whether someone needed to analyze sales, review customer data, monitor campaign performance or manage an account, the dashboard served as the starting point. Software companies invested heavily in making those environments easier to navigate because getting customers into the application was central to how the product delivered value.
AI assistants are starting to challenge that assumption as people are growing accustomed to starting with asking a question instead of searching through menus and reports. They can tell an AI assistant what they want to know or accomplish and expect the technology to determine what information it needs to respond.
As that behavior carries into the workplace, software companies will see customers who may continue relying on their software while spending less time inside the application itself, changing how they expose their capabilities and even how they measure the value customers are getting from the product.
Software Can Become Part of the Conversation
AI assistants such as Claude and ChatGPT have already changed how people find information and complete everyday tasks. Enterprise adoption creates a different opportunity because these systems can increasingly connect with the specialized applications and proprietary data employees use at work.
Model Context Protocol (MCP) is one technology making those connections possible. MCP provides a standardized way for AI applications to connect with external data sources and software tools. Instead of requiring every interaction to begin inside a particular application, software providers can make parts of their products accessible through the AI environments their customers use.
To really visualize this, just picture an analytics application. A marketer traditionally might log in, open a report, select a date range, apply filters and compare several customer groups before finding an answer. With an AI interface connected to the underlying software, that marketer could simply ask which recently acquired customers are predicted to become high-value customers.
The analytics software still performs an important job behind the scenes, because it holds the relevant data, applies specialized models and provides capabilities developed for a particular use case, but the way the customer reaches those capabilities begins to change.
Proprietary Data Becomes Increasingly Valuable
Giving users conversational access to software also exposes one of the limitations of general-purpose AI.
A model can know a great deal about marketing or ecommerce without knowing what happened inside a particular company yesterday. It doesn’t automatically know how a brand calculates customer value or after what period they consider a customer to be lapsed. Useful enterprise AI depends on access to the information specific to the organization using it.
Software companies already sit on much of that specialized information. Customer data platforms understand customer behavior and business intelligence tools organize company performance data. Other enterprise applications contain similarly specialized information about the organizations they serve.
Connecting those systems with AI gives the model a way to work with company-specific information rather than relying solely on its general knowledge.
As AI models continue to improve, proprietary data and specialized software capabilities could become an increasingly important source of differentiation. The model may provide the conversational experience, while enterprise applications provide the foundational data and functionality that make the conversation useful to a particular business.
Answers Are Only Part of the Opportunity
The next stage of conversational software will also depend on what happens after the AI provides an answer. Knowing an insight is useful, but acting on it is where business software earns much of its value.
Imagine asking an AI assistant which customers are most likely to make another purchase, then using that answer to create an audience and make it available to another system, turning the conversation into a workflow. Similar possibilities exist throughout enterprise software. A user could identify a problem and initiate the next step without moving through several applications, because the underlying systems would continue to perform specialized functions while the AI assistant coordinates how the user accesses them.
Once AI can initiate actions, SaaS products start to look more like collections of specialized capabilities that an AI system can call when needed. An analytics product might identify an audience, a commerce platform might provide transaction data and a marketing platform might activate a campaign, and the best part is that the user doesn’t necessarily need to navigate each application for those products to contribute to the workflow. Then, software companies can begin designing workflows around what the customer wants to accomplish rather than the sequence of screens required to get there.
The Dashboard Still Has a Job
Dashboards are unlikely to disappear completely, simply because some tasks benefit from visual exploration or the ability to examine large amounts of information at once. Users will still need dedicated applications for many types of work, it’s just that their role has become less central.
A customer may open an application when deeper exploration is necessary while handling routine questions and actions through an AI assistant. Another user may interact with the underlying software regularly without opening its traditional interface at all.
SaaS companies have historically treated logins, sessions and time spent inside the product as signals of engagement. AI-mediated software complicates those metrics. A customer who rarely opens a dashboard could still be a highly engaged user if an AI assistant regularly calls the product’s data, models or workflows on their behalf..
Build for Where the Customer Works
SaaS companies spent years trying to become the application customers opened first. AI assistants introduce the possibility that no individual business application will consistently hold that position. Software still needs to provide trusted data, specialized functionality and reliable execution.
Companies preparing for this model should start asking which parts of their products customers need most frequently, which data can safely become available to AI systems and which actions users should be able to initiate conversationally. I think the dashboard will remain useful, but it just may no longer be where interactions begin.
