" . . design for users you'll never meet . ."
Not long ago, every role I filled assumed a human at the end of the wire. Journeys, logins, dashboards.
Today, many ‘users’ never see a pixel: service principals, bots, schedulers, AI agents, queues and functions trading events at machine speed while people set intent and watch the dials. That shift is redefining what good Solution Architecture looks like .. and how I spot architects who can actually land it.
Designing for machine users
When your consumers are workloads, identity stops being a HR attribute and becomes a capability question.
"What is this thing allowed to do, for how long, and under which policy?"
The Solutions Architects who stand out talk about workload identity, short-lived credentials and entitlements that follow capability rather than team charts. Authentication and authorisation aren’t bolted on after the pipeline turns green; they’re first-class design elements.
Experience changes too. With non-human consumers, there’s no glossy interface to mask systemic flaws. ‘Experience’ is latency, throughput, backoff, idempotency and clear failure semantics. If a message is processed twice, what happens? If a dependency slows by 200ms, who notices first?
The customer, or your SLOs? Strong candidates ask these questions unprompted.
And then there’s topology. In this world, the diagram matters more than the demo. Good architecture reads like a truthful map: where data lives, how it moves, where congestion can occur, how you contain blast radius.
You must clearly narrate trade-offs. Consistency, availability; this queue drains at X per second; that cache is regional because of residency ..
Operability and guardrails (not meetings)
If you can’t follow a unit of work across boundaries, you don’t have a system .. you have guesswork. Observability is part of the product.
You must bake tracing, logs and metrics into the design from day one and describe the minimal telemetry needed to diagnose a cross-service failure at 2am without waking five teams.
Certain habits show up again and again in programmes that land value; event-first thinking to avoid tight coupling; versioned contracts because change is constant; policy as code so guardrails replace meetings; secrets that die young; cost treated as a non-functional requirement with real numbers, not a wish; failure handled gracefully with dead-letter triage, circuit breakers and compensations that were written before anything shipped.
The traps are familiar too.
Service accounts are created like visitor badges and never retired. One monolithic CI/CD sausage machine forces every workload through the same gate regardless of risk. Telemetry debt builds quietly until the first incident reveals that logs aren’t traces and metrics without exemplars don’t help. Multi-region is treated as copy-paste rather than a design problem with sovereignty, fail over and reconciliation to solve.
Hiring for substance and shaping the org around it
From my recruiting chair, the strongest Solution Architects make strategy tangible.
They frame outcomes as SLOs and error budgets, not slogans. They speak about identity like security practitioners but stay pragmatic: least privilege scoped to capability, automated rotation, break-glass with recording. They narrate failure paths with the same ease others describe happy paths (ask what happens when the queue backs up and they’ll tell you when autoscaling kicks in, who gets paged, and which compensating flow activates).
That means changing how we hire. Keep the stakeholder skills, but add a short systems conversation where the candidate walks a single message through their design and explains versioning, idempotency and the minimal telemetry to debug it. Listen for a cost story finance can follow: unit economics, cardinality limits, storage tiers, and why a million tiny writes might cost more than ten big ones. Look for diagrams you can implement without a legend.
None of this lands if the organisation around it stays human-scale while the systems go machine-scale. Teams that win tend to own by flow, not function: a business capability or event stream end-to-end (data, policy, runtime, reliability). Approvals move into pipelines where signed artefacts and policy checks are the control. Runbooks read like user manuals for the people who matter at 2am: the humans on call. Progression frameworks reward operability and risk reduction alongside feature velocity.
The human bit (ironically)
The rise of non-human workloads doesn’t remove people - it moves human judgement to where it has the most leverage.
The best Solution Architects are translators, turning strategy into constraints, constraints into patterns, patterns into code and incidents into learning. If you can do that (and tell the story cleanly) you’ll have more options than time.
From my side, the brief is simple.
Surface those people early, set the bar clearly, and run an interview loop that recognises substance over shine. If your ‘users’ never log in, you need architects who make invisible systems understandable, operable and economical. That’s the work I love to support .. and the talent I’m always keen to meet.
