Implement now, own the system of work later
Enterprises buying AI need hands-on help, not just a chatbot and a credit card checkout. Services-led growth and forward deployed engineers are emerging as the decisive edge for complex, agentic applications that aim to replace human workflows and become the new systems of work. The payoff: early margin trade-offs that compound into moats, pricing power, and ultimately software-like gross margins.
Points clés
- Lightweight, wrapper-like AI agents frequently break when attempting to replace core workflows in systems of record like Salesforce.
- Despite complexity, AI products in sales, support, and legal are scaling fast, often reaching $5–20 million in ARR within their first two years.
- PLG darlings (Atlassian, Slack, Figma, Notion, Dropbox, ChatGPT, Cursor) contrast with systems-of-record winners (Salesforce, ServiceNow, Workday) that relied on heavy implementation and services.
- Market caps underscore the model: Salesforce at ~$254B, ServiceNow ~$194B, and Workday ~$63B significantly outpace PLG peers.
- Implementation initially depresses margins: at IPO, ServiceNow posted 63.2% gross margin and Workday 54.1%; Salesforce reportedly burned $52M to generate $22M in revenue before its partner ecosystem matured.
- Over time, moats restore margins: by 2024, ServiceNow reached 79% gross margin and Workday 75%.
- AI can now streamline integration work itself (e.g., agentic browsers for systems without APIs, conversational agents orchestrating workflows), accelerating services-led deployments.
- Agents are becoming active coworkers, requiring expert services to redesign roles and processes; forward deployed and solutions engineers are central to success.
- Example: Decagon fields “Agent Product Managers” who implement and tune support agents with customers to drive autonomous resolution.
- Even model providers are leaning in: OpenAI listed 22 of 311 open roles as forward deployed or solutions engineering, signaling enterprise implementation as a competitive battleground.
À retenir
Stop worshipping the 80% gross margin slide in year one and start worshipping your customer’s messy workflows. Hire forward deployed engineers, pick a focused ICP, document everything, and automate the boring parts like your runway depends on it (because it does). Get onsite, wire into systems, and own the data ingestion layer—then you can enjoy your beautiful margins later; for now, embrace the glamorous world of implementation… yes, your TAM deck can wait.
Sources
