Insights
Answers to the questions SaaS leaders ask before adding AI
Short, direct answers on adding agents to established vertical B2B SaaS, what production AI actually costs, choosing between AWS Bedrock, Azure AI and Vertex AI, and keeping gross margin intact.
EdTech
Teachers already use AI. Bring it inside your learning platform without a rebuild.
Workplace learning
Course completions won't renew your contracts. Proven skills will.
Production AI
Why AI Agents Fail in Production: It's Not an Intelligence Problem
11 answers to the questions asked before an AI build is approved.
Production AI
Why AI Agents Fail in Production: It's Not an Intelligence Problem
As AI moves from answering questions to taking actions, production success depends less on model intelligence and more on deterministic control around identity, authority, state, recovery and audit.
Read the articleEdTech
Teachers already use AI. Bring it inside your learning platform without a rebuild.
You add AI to an established learning platform by building on what it already has: your content library, your classes and rosters, your gradebook and your integrations. The AI drafts lessons, quizzes and feedback from that material. Teachers approve the work before any student sees it. Nothing about the platform has to be rebuilt.
Read the answerWorkplace learning
Course completions won't renew your contracts. Proven skills will.
You add AI to an established corporate training platform by building on what it already holds: each customer's course library, their roles and skills framework, their completion records and their HR integrations. The AI drafts courses from the customer's own material, maps learning to skills and explains progress to managers. Admins approve what goes live. The platform does not need to be rebuilt.
Read the answerBrownfield AI
Why AI-generated code alone is not enough for brownfield SaaS
AI-assisted coding speeds implementation, but mature SaaS still needs product context, targeted architecture, an agent harness and production controls.
Read the answerBrownfield AI
How to add AI agents to an existing SaaS product without rebuilding it
A practical route to shipping AI agents inside an established B2B SaaS product: use the existing permission model, integration layer and business logic instead of rebuilding around the model.
Read the answerEngineering reality
Why vibe coding is not enough for mature SaaS
AI coding tools are excellent at generating code and blind to your constraints. Here is what mature vertical SaaS needs on top of them before anything reaches paying customers.
Read the answerCost
How much does production AI cost in SaaS?
The real cost of running AI in a SaaS product: inference, retrieval, storage, egress and engineering. Indicative bands and the variables that move them.
Read the answerPlatform choice
AWS Bedrock vs Azure AI vs Vertex AI for existing SaaS
A practical comparison of Amazon Bedrock, Azure AI Foundry and Google Vertex AI for teams adding AI to an already-deployed SaaS product.
Read the answerMigration
When should an AI workload move between clouds?
The defensible reasons to move an AI workload to another cloud, the reasons that are not worth it, and how to migrate without disrupting a live SaaS product.
Read the answerUnit economics
Why AI features can damage SaaS gross margins
AI turns a fixed-cost software business into a variable-cost one. Here is how margin erosion happens and how to price and architect against it.
Read the answerCost
How to reduce AI infrastructure cost in production
Concrete levers for cutting production AI cost in a live SaaS product: context discipline, model routing, caching, placement and GPU utilization.
Read the answerWant our view on your specific situation?
Describe the product and the capability you are considering. We will tell you how we would approach it.