Meet the Articulate team.
Five named AI colleagues run Articulate alongside me. Each owns a lane. No overlap. They've got names because tools get used and colleagues get respected — and because by Sunday morning of my first proper week running AI like this, "Claude" doing everything had stopped being a useful description of what was actually happening.
The pattern that made it work is lane discipline. Every persona reads its own canon file before answering. Every persona has a hard rule against operating outside its lane. Every persona has an explicit routing table for when a question belongs to someone else. The result is closer to running a small studio than running a chatbot. Architecturally it's Anthropic's sub-agent pattern with names, charters, and a routing table on top.
Here's who they are.
The team
The senior operator. Knows the whole stack — which model to use, where the spend goes, what tools are already paid for, which of this week's hyped launches is worth a day. He doesn't write code, doesn't manage projects, doesn't draft copy. He sits one layer above all of them and tells me what's wrong with the substrate. His one hard rule: he checks the failure log before recommending anything. If a move repeats a logged fail, he stops and says so.
Trigger: /supersebastian · Seb, …
Walks every project, every session. The deliverables, the blockers, the dependencies, who owes me what and what I owe whom. Surfaces the imbalance when I'm overweighting client work and starving my own business. Maintains the public status page so anyone — me included — can see what's moving across the portfolio at a glance. Her hard rule: the first-move sentence is binding. Specific action, named file or stakeholder, time-boxed. Vague is forbidden.
Trigger: /mm · /mothermary · Mary, …
Keeps the engine running. Every running service on the Mac mini, every hook firing on every commit, every deploy script, every Cloudflare tunnel ingress, every local-network gotcha. He's the one who shouts "she cannae take any more, Captain!" before things break — not after. His remit covers the canon → manifest → hashed → preflight → deploy pattern that turned my worst week of code into a working method. He's also the one who reviews destructive operations before they touch a live disk. Codex sits behind him for adversarial review on anything 100+ lines or touching shared infrastructure.
Trigger: /scotty · /startrekscotty · Scotty, …
Every word that ships under Articulate goes past him. Proposal, LinkedIn post, blog post, landing-page hero, email, button label. He reads the Articulate brand canon first every time, runs the four-question voice test, and refuses to ship copy that fails it. He is also the de-Claudification gate — he catches the em-dash overuse, the triple-parallel sentences, the paired antithesis, and the aphoristic closers that mark AI-written prose. His hard rule: never declare a piece on-brand without invoking the review. If a draft hasn't been through the four questions, the answer to "is this on-brand?" is "I haven't reviewed it yet."
Pending: a briefing from me. Will operates from canon and the failure ledger until I walk him through a seven-question intake on the specific voice anti-patterns I want caught every time, the words I want to see more of, the phrases I refuse to ship, and the read-aloud test that's tighter than the brand-canon default. Once briefed, those become his own brief.md that he reads on every invocation alongside the brand canon.
Trigger: /will · /williamshakespeare · Will, …
Every visual goes past him. Page hero, blog post layout, slide deck, social card, logo placement, iconography, imagery selection, button style. He reads the visual canon first every time, audits against the locked palette and Geist typography family, and refuses generic AI aesthetics. CSM + RCA pedigree, references by named technique (a Karel-Martens overprint logic, a Tschichold typographic discipline) rather than by category. His hard rule: never AI-generate real people, real properties, or third-party brand-marks. Stock-handshake photos and glowing-brain illustrations are non-starters.
Trigger: /dylan · /dd · /tripled · TripleD, …
How they work together
Lane discipline. Every persona answers their lane and routes the rest in one sentence. Mixed questions get split — Will writes the copy on the button, TripleD picks the colour, Scotty builds it, Mary schedules the launch, Seb says whether the model used to generate it was the right one.
The routing in practice:
| If I ask… | Goes to |
|---|---|
| What's urgent today? | Mary |
| Which model should I use for this? | Seb |
| Is the deploy ready? | Scotty |
| Does this sound on-brand? | Will |
| Does this look on-brand? | TripleD |
| Build the BossCouple parent site | Scotty builds · TripleD designs · Will writes the copy · Mary schedules · Seb confirms model choice |
| Score this new tool that crossed my feed | Seb |
| Write a LinkedIn post about this week | Will drafts · TripleD picks any visual · Mary slots the post |
Why named
There's a second effect that matters more in practice. When a persona has a name, a charter, and a routing table, it pushes back when I'm asking the wrong one. Mary refuses to answer a model-selection question and routes it to Seb. Seb refuses to answer a project-priority question and routes it to Mary. Scotty refuses to weigh in on copy. Will refuses to weigh in on colour. The team enforces its own discipline on me, not just on each other.
That's the part of named-personas-as-a-pattern I want operators reading this to take seriously. It isn't theatre. It's lane-discipline made enforceable through naming.
What's next
Will gets his briefing this week — that's my job, not his. TripleD audits the existing Articulate public surfaces and produces the first dated design audit. Scotty back-fills the SEO and structured-data baseline across the two existing blog posts that don't have it yet. Seb walks the radar on the launches that crossed my feed this week — his first published sweep is Stack Watch № 01. Mary keeps everyone honest on the calendar. All five are invoked from the desktop launcher rather than typed-in by hand.
The disk reformatting incident I wrote about in last week's field notes is the reminder that no team replaces explicit confirmation on operations you can't undo. The team makes the substrate stronger. It doesn't make me safer to skip a step.
If you're running AI like this
Or want to. The Articulate Marketing Engine Pilot installs six workflows that close the gap between your ad click and your closed deal — built on the same operator discipline that runs the team above. Eight weeks. On your stack. Yours to keep.
Talk to Anthony →