# DoorDash Is Testing Lunch By Text. It's A Clue To How You'll Buy Software In 2027.

## Executive summary

AI agents are fundamentally shifting the value proposition of software, moving revenue generation away from traditional SaaS interfaces and toward the underlying data, application, and agentic layers. The ability to integrate and process data via a custom agent (the 'agentic layer') is becoming the primary differentiator. Companies must now position themselves not just as tools, but as trusted data custodians and workflow orchestrators to survive the agent disruption. For builders, this means focusing on creating highly structured, interoperable data schemas that can be consumed by multiple agents.

## Key takeaways

- Agent disruption makes old SaaS tools less sticky.: As agents handle information processing (e.g., reading contracts, writing kickoff plans), the need to open and use a dedicated, single-purpose tool diminishes, making the tool itself less valuable.
- The agent trained on proprietary knowledge is hard to replace.: The true value lies in the 'lessons learned'—the unique context and workflow knowledge taught to a custom agent. This knowledge, if housed in a custom harness, creates a significant competitive moat.
- The competitive set for software is broadening.: Major players (Meta, Salesforce, Microsoft) are all positioning themselves as control planes, leveraging their existing data distribution to manage and govern agents, rather than just selling the application.
- Value must be demonstrated across multiple layers.: To sell software in the age of agents, vendors must move beyond simple value demonstrations and prove how their solution impacts the data layer, the application layer, and the agentic layer simultaneously.
- The core value remains in structured workflows.: While interfaces may shift, highly compliant, structured workflows (like payroll) and the underlying data structure remain critical applications that cannot be easily replaced.

## Technical details

- Agentic Architecture: The concept of the 'agentic layer' is the ability for an agent to read, process, and act on data across different systems. The goal is to move the customer interaction from the UI to the agent, which is the future of software.
- Data Layer Strategy: The data layer is paramount. Vendors must position themselves as the trusted data source, allowing agents to access and process information, regardless of the interface used. This is a key focus for Meta (Muse) and Salesforce.
- Salesforce Koa: Salesforce announced Koa, a reasoning model based on the Neotron stack tuned for CRM. It is in pilot and trained on synthetic data, demonstrating Salesforce's attempt to position itself as the central agent gateway for CRM data.
- Microsoft Copilot/Autopilot: Microsoft is positioning its suite (Copilot, Autopilot) as a comprehensive control plane, leveraging its existing productivity tools (Outlook, Excel) and Azure infrastructure to manage enterprise data and AI access.
- Interoperability Standard: The Multiple Connection Protocol (MCP) is highlighted as a critical mechanism (pioneered by Anthropic) that allows agents to plug into data sources, making interoperability a key feature for any modern enterprise solution.
- Data/Workflow Integration Example: Crustdata demonstrated a workflow where a two-person recruiting business built a custom skill using Crustdata's APIs, combining structured data with proprietary selection criteria, bypassing a packaged SaaS tool.
- Compliance and Workflow: Workday's payroll agent example shows that even with advanced AI, high-compliance processes require robust, configured underlying systems and dependable execution, making the application layer resilient.

## Practical implications

- For Employees: Proactively document and socialize your unique, complex AI workflows and the data they rely on with your manager to ensure business continuity and secure investment.
- For Leaders/Buyers: Do not assume AI is a simple overlay. Conduct deep 'fingerprint usage' audits with teams to understand the actual, idiosyncratic ways AI is being used to drive outcomes, rather than relying on high-level vendor pitches.
- For Sellers/Builders: Focus on building solutions that are inherently modular, allowing the data layer to be easily accessed and consumed by multiple agents (multi-agent system). The goal is to sell optionality and trust, not just a feature set.
- For All Stakeholders: Understand that the conversation has shifted from 'Do you have AI?' to 'How do you manage the data, application, and agentic layers together?'

## Topics

AI Agents, SaaS Architecture, Workflow Automation, Enterprise Software Strategy, Data Governance, Nate's Newsletter, Meta Muse for Small Business, Salesforce Koa, Microsoft Copilot/Autopilot, Crustdata, Workday

Source: https://www.youtube.com/watch?v=0j8wy4jUlsw