Practical notes from the engineers who build agents, copilots and AI products.
A churn model is only useful if it feeds an action; build it with the retention team, validate on time and act on the scores. A churn prediction model should be defined around a churn event the business can act on early, scored at the cadence the team works to, and judged on customers saved, not accuracy alone.
Automated reconciliation matches collections, deposits and settlements daily instead of a two-day monthly effort. The system takes every delivery-level collection from the driver app, matches it to bank deposits and UPI credits, flags the gaps the same day, and produces shipper settlement files without a spreadsheet.
Running TheEazy CXM, HXM and Zentrix in production is where our patterns for evals, multi-tenancy and cost control come from. We carry pager duty, pay our own inference bills and answer to our own customers, and every practice we bring to a client build was first tested on a system we could not walk away from.
A good roadmap sequences foundations before features, has stop points, an owner per phase and running costs alongside build costs. It is short enough to fit on two pages, honest about what is unknown, and it is reviewed against evidence every month rather than defended against it.
A delivery platform needs customer, merchant, rider and admin surfaces, real-time dispatch, UPI payments and a backend sized for festival peaks. Multi-vertical means one backend for food, grocery and parcels, with per-vertical catalogue and fulfilment rules on top, so the second vertical costs far less than the first.
Buy when a product covers 80% of the need, integrate when your data is the differentiator, build when the process is. The third option is the one most companies miss: a bought model or platform, connected to your own data, retrieval and policies, gives most of the benefit of building at a fraction of the cost.
An AMC should define response times, included hours, patching cadence, monitoring, reporting and, for AI, evaluation and cost reviews. It should also say who owns the code and credentials, what counts as a change request rather than maintenance, and how either side exits. Anything vaguer is a retainer, not a contract.
Submit early, prepare privacy labels, permissions rationale, test accounts and screenshots; compliance is scope, not an afterthought. Most rejections come from a short list: missing account deletion, unexplained permissions, payment rules, incomplete metadata and reviewers who cannot log in. Plan for them in week one.
A POC runs on your real data against agreed thresholds and produces production-style code; a demo runs on curated data and proves nothing. Buying on a demo is how teams lose a quarter discovering that the impressive thing they saw does not survive their documents, their edge cases or their volume.