%20(1).1150616f.png&w=64&q=75){kind=small}
Blog
Insights & updates
Perspectives on agents, deployment context, and the systems teams build around them.
Change Management · Forward Deployed Engineering · Context
Change management is a context problem
Change management is usually framed as a people problem. In practice, it stays hard because the reasoning, decisions, and workarounds from one deployment rarely reach the next.
NPNikhil PillaiAugust 20, 20266 min read
Forward Deployed Engineering · Context · Knowledge Graphs
FDEs Connect Everything. Problem Solved? Not Quite.
Connect Slack, Drive, contracts, meeting notes, and calls to an agent and answers arrive instantly. But an FDE's real context problem is not access to information. It is understanding a deployment's evolving decision history without losing what the whole team should learn.
NPNikhil PillaiAugust 11, 20265 min read
Forward Deployed Engineering · Deployment Context · AI Engineering
Context Is Changing, and You Didn't Even Know About It
Forward deployed engineering is now a default motion, but its context is neither narrow nor stable. The role combines engineering, customer success, and decision-making, and the existing toolchain was never built to hold that shared picture together.
NPNikhil PillaiAugust 6, 20267 min read
Finance · Insurance · Domain-Specific Agents
The Interesting Case of Domain-Specific Agents
When an AI agent is deployed to handle specialized, high-stakes work in finance or insurance, the extent of silent failures in production is staggering, and standard evaluation frameworks are blind to them.
NPNikhil PillaiApril 16, 20266 min read
Evals · Sourcing Agents · AI Quality
What's Missing With Your Sourcing Agent
When you rely on an AI agent to source companies, founders, or prospects, the extent of silent failures in production is staggering — and your standard evaluation frameworks are blind to them.
NPNikhil PillaiApril 13, 20265 min read
Evals · Support Agents · AI Quality
What Evals Miss Once Your Support Agent Goes Live
Eval frameworks are genuinely useful. But there is a class of production failure that even a well-configured eval setup does not reach — and it is quietly eroding your customer trust.
NPNikhil PillaiMarch 22, 20265 min read