Practical notes from the engineers who build agents, copilots and AI products.
Sovereign AI keeps models, data and operations under Indian jurisdiction; for enterprises it means private deployment and local providers. In practice: open-weight or Indian models served in Indian regions or your own data centre, Indian contracts, and operations you control. It is a procurement stance, not a tool.
Indian delivery at senior quality runs a fraction of US rates; the gap narrows for AI specialists, and outcome-based pricing matters more than hourly. Compare vendors on what a finished, evaluated system costs and who owns it afterwards, not on the rate card, since rates hide seniority, rework and management overhead.
A useful SLA states response and resolution by severity, cover hours, escalation path and what happens when it is missed. It also defines the severities in plain language, says whose clock is used, and reports performance against itself every month, so both sides can see whether the promise is being kept.
Introduce a tenant key, isolate data by row or schema, centralise config and billing, migrate customers in waves behind a façade. It succeeds when the tenant key is enforced by the database rather than remembered by developers, and when each wave has a rollback path that keeps the old instance intact until sign-off.
Custom personalisation on your own event data replaces subscriptions with an owned system and an A/B-proven lift. It runs as a small service beside your store, reads events and the catalogue through Shopify's APIs, serves recommendations to your theme or headless front end, and reports lift over a holdout you control.
Ship to a chosen cohort, measure adoption and evals, iterate twice, then release generally with the metric that proves it. AI feature rollout works when the flag controls the model and prompt version as well as the UI, the cohort is chosen for feedback quality, and general release waits for a number, not a date.
Launch agents in shadow mode where they propose and humans approve, then grant autonomy intent by intent as evaluation evidence accumulates. The agent earns each permission with a record of accepted proposals, not with a demo, and the people whose work it touches decide when it is ready.
Self-host when residency, contracts or volume demand it; open-weight models match public APIs for most retrieval and extraction tasks. If none of those three pressures applies, an API is cheaper and simpler. If one does, plan for GPUs, a serving stack, an evaluation suite and the people to run them.
Patch critical vulnerabilities within days, dependencies monthly, and run scanning in CI so nothing waits for an incident. Add a quarterly slot for framework and runtime upgrades, keep a tested rollback for every patch, and report what was patched and what is outstanding each month so the backlog never goes invisible.