How enterprises actually adopt technology

Original CloudExchangers explainer. It describes established public practice and labels the desk’s judgment as analysis. It is not a report of a newly announced event.

The unit of adoption is a workflow

Enterprises do not adopt a technology. They change a workflow that already has customers, controls and a person who is blamed when it stops. A new database, a new model, a new cloud account: each is real only when a workflow uses it on a day that is not a demo. Microsoft’s Cloud Adoption Framework and AWS’s Well-Architected material are both, in different words, attempts to make that change repeatable. They are checklists for adults. They are not strategies by themselves.

Build, buy, or stop

Buying is rational when the problem is common and the vendor’s operation is better than yours will be. Building is rational when the workflow is the business and the commodity tools bend it into something you cannot sell. Stopping is rational more often than the roadmap admits, when the current system is merely unfashionable and still correct.

The hidden option is integrate. Most enterprise cost sits in the joints: identity, data definitions, the audit log, the way an incident is declared. A product that cannot say how it joins those is not a product yet. It is a prototype in a contract.

The review that matters

An architecture review earns its time when it can reject a design. Useful questions are stable. What is the system of record? What identity does the job run as? What is the restore? What data leaves the environment? Who is paged? What will this cost if it succeeds, not only if the pilot stays small? For anything that makes or recommends a decision with a model, NIST’s AI Risk Management Framework adds a further set: govern, map, measure, manage. Use it as questions, not as a certificate.

A review that cannot reject anything is a ceremony. People learn to route around ceremonies, and then the real architecture is whatever shipped.

Operations is the adoption

A system is adopted when the runbook exists, the on-call rotation knows the failure mode, and a change can be reversed. Training sessions do not do that. A week of the new system carrying a real, bounded workflow does. Widen the workflow only after that week has been dull.

Sponsorship and membership labels, including on a story like this one, are commercial furniture. They do not change the sequence. A vendor who wants a sponsored explanation still has to survive the same review. This publication’s membership label marks a longer piece. It does not lock it, because billing is not switched on. The operating point remains: do not confuse a label with a control.

Analysis

Enterprise technology is slow for reasons that are sometimes wisdom. The wisdom is the review, the identity, the restore, the pager. The slowness that is not wisdom is a queue with no owner. Tell those apart. Speed up the second. Do not skip the first because a competitor’s keynote was confident.

What this essay will not pretend

It will not offer a maturity score or a percentage of companies that “get it.” Those statistics are where this genre goes to lie. The sequence above is usable without them: workflow, decision to build or buy, a review that can say no, and a dull week in production.