[DigitalToday reporter Chi-gyu Hwang] As leading enterprise software companies roll out so-called headless software platforms that can be used by AI agents rather than people, attention is focused on how the industry landscape could change.

In tech circles, headless generally means software that provides only back-end functions, without a user-facing screen (GUI, head). Interest is rising in how headless software could reshape the software ecosystem as major enterprise software companies such as Salesforce and ServiceNow recently rush to roll out headless-style flagship platforms aimed at AI agents.

ㆍSalesforce answers the future of CRM in the AI agent era with Headless 360 ㆍServiceNow launches Action Fabric, opening its platform to external AI agents

In this regard, Seema Amble (시마 앰블), a partner at Silicon Valley venture investment firm Andreessen Horowitz (a16z), drew attention by stressing elements that will become important in a headless enterprise software environment, saying software competitive advantage will inevitably change in a headless era.

In a recent post shared on the firm's website, she stressed that "next-generation systems of record are already changing into an agentic form that is not a repository that records human work, but captures context, initiates work and records data."

She highlighted the changes headless could bring to enterprise software, citing Salesforce announcements as an example.

According to her, Salesforce's move to open APIs and launch headless products can be summed up as a bet that, in the agent era, value lies in the data layer rather than the UI.

Salesforce's announcement also raises more fundamental questions, including what remains of Salesforce without the UI, and how that differs from a combination of a database and APIs that anyone can build.

Over the past 20 years, what Salesforce sold was dashboards, pipeline views and forecasting tools. The UI kept people tied to the product. It forced consistent data entry. It made sales representatives enter data they would not have entered unless they were using Salesforce. Many sales leaders insist on adopting Salesforce at new workplaces not because they like the UI, but because it has become ingrained, she explained.

But agents are shaking up this model because they can read and write underlying data directly without going through the UI. It naturally raises a question: In this environment, what remains and what disappears?

Amble said among the factors that kept customers attached to existing systems of record, those based on human habits and preferences will lose power in the agent era. "In an AI agent environment, ingrained habits (muscle memory), such as frequency of use and whether you read and write, cannot become a moat," she said.

Some things will remain in a headless era, she said. "Undocumented standard operating procedures (SOP) are still important in the short term. For agents to work correctly, workflow rules need implicit rules and practices accumulated over a long time within an organisation. Connectivity is also important. A CRM agent must weave together data for sales, billing and customer success. Compliance data will become more important. In interactions between agents, systems of record that serve as an identity and permissions layer are structurally hard to replace," she said.

Some factors will become even more important beyond what remains.

First is how easily a competitor can replicate an existing system of record. "The harder it is to replicate, the stronger the moat becomes. AI is lowering the cost of building 80 percent of a system of record anew. The remaining 20 percent - exception handling, approvals, compliance requirements and edge-case workflow - remains the standard that separates useful tools from true replacements for existing systems of record," she said.

Whether there is proprietary data is also important. "Defensible data that competitors cannot easily replicate or take is data the product uniquely generates, not data that was brought in. That includes observed behaviour, response rates, timing patterns, process outcomes and agent performance records. Data is now not a record but context that agents need to judge and act," she said.

Amble also pointed to whether a company owns the execution layer. "In the past, it was enough to store records. Now, products that can make the process run on its own - act, record results and use those results to make better decisions next time - will have a competitive advantage," she said.

Whether there is an execution element in the real world is also becoming important. "Software that does not stop at storing data or making recommendations, but actually sends people, delivers goods and completes services is far harder for competitors to catch up with. That is because it is connected to the real world, unlike pure SaaS that operates only on a screen," she said.

Network effects are also a variable. "Network effects have not been big in systems of record so far, but if agent-era systems are embedded in multilateral workflows such as buyers and sellers, employers and employees, and companies and auditors, network effects could become far more important," she said.