Vendors and platforms
A lot of AI strategy turns out to be vendor strategy. Your data sits in SaaS products that make it hard to move, the big platforms want to be the layer between your people and every AI model, and every team is tempted to build its own.
The first three posts are a series on SaaS lock-in: how the moat works, what vendors could do instead and what the EU Data Act changes. The rest are practical calls: what Microsoft is really selling with Copilot, whether to use one AI provider or several, and when to build rather than buy.
6 posts, in suggested reading order
Why your SaaS data is so hard to move
SaaS vendors make your data hard to move, sometimes by design, and AI raises the cost of that. What it looks like and what to ask for before you sign.
5 min read · Updated September 2026
A playbook for SaaS vendors in the agent era
Lock-in is a weak moat once agents are the users. As a vendor I'd make leaving easy, build API-first, ship MCP servers that work and price on outcomes.
5 min read · Updated September 2026
What the EU Data Act means for SaaS and cloud contracts
Since September 2025, EU law makes cloud and SaaS providers help customers leave. What the switching rules require, the dates that matter and what to ask for.
5 min read · Updated September 2026
What Microsoft is really selling with Copilot
Copilot is Microsoft's bid to be the layer between your people and every AI model. The gains are real; know what you give up and what to ask.
6 min read · Updated September 2026
When multi-provider AI is worth the premium
You can now hedge AI lock-in without juggling vendors. The real work is choosing which layer to abstract and deciding what the insurance is worth.
6 min read · Updated September 2026
When a custom AI agent is worth building
Most teams should pilot enterprise AI they can buy, wired to their systems with MCP, before building anything. Building pays in a few narrow cases.
7 min read · Updated September 2026