--- title: Modes of AI work description: Companion framing to the AI scale (chat, context, clockwork, colleague) that treats the four levels as modes of work — the scale is for your skills, not your workflows. Skills climb the scale once; workflows never climb. No metaphor by deliberate choice (2026-08-26) — every candidate ranks the modes, and the ranking is the misunderstanding. The job picks the mode. Third mode renamed from automation to clockwork (2026-09-09) to end the collision with the AI Fluency framework's Automation modality; the artifact it produces is still called an automation. category: patterns updated: 2026-09-25 --- # Modes of AI work **Modes of AI work** is a companion framing to the AI scale. The scale — chat, context, clockwork, colleague — describes the natural progression most people follow as they mature with AI: starting with conversational use, then building custom assistants, then designing automations, eventually working alongside an AI colleague that takes on problems itself. That progression is real and worth naming, because it tells someone with no prior framing where they currently sit and where they could go next. But the progression is **additive**, not **substitutive**: someone operating at the colleague level still uses chat constantly, still has custom assistants for ongoing work, still runs automations. Each level you reach gets added to your repertoire; you don't climb away from earlier ones. The one-line reconciliation: **the scale is for your skills, not your workflows.** Your skills climb the scale once — each mode teaches what the next one needs. Your workflows never climb it — each workflow gets the mode that fits the work. ## No metaphor — a deliberate position The framing carries **no standing metaphor** (decided 2026-08-26). This is a position, not a gap. Every candidate tried or considered failed the same way — it ranked the modes: - **Gears** (rejected 2026-07-31): "stuck in first gear" and "kick into high gear" pejorate the low end — the idiom reintroduces exactly the shame the framing exists to remove. - **Transport** — walk/bike/train/plane (adopted 2026-07-31, dropped 2026-08-26): walk→plane is ordered by speed, cost, and prestige in everyone's head, so the metaphor invited the ladder misread and then needed a patch sentence ("nobody stops walking") to undo its own implication. Two of four mappings carried no meaning — a bike says nothing about loaded context, and a plane makes the human a passenger, not a delegator. And "the trip picks the mode" selects by distance, a single axis — which quietly re-imports the ranking. - **Kitchen** — knife/meal prep/slow cooker/sous-chef (considered 2026-08-26, passed on): the strongest mappings of any candidate, but still an ordered learning progression at its core, and four items from four different categories. The standing insight: **any metaphor ordered enough to be memorable orders the modes, and the ordering is the exact misunderstanding the framing exists to remove.** The failed metaphor hunt is itself usable content (see short 16) — it demonstrates the point rather than decorating it. In practice the modes are presented plainly: four names, one-line definitions, and real week-of-work examples showing all four in use at once. Dignity for the basic mode comes from the facts, stated directly — chat is not something you outgrow; practitioners at every level use it all day — not from imagery. The most common failure pattern around the scale is still shame (practitioners apologizing for "still just using ChatGPT," as if chat were something left behind); the correction is the additive-not-substitutive point itself. Key line: **"The job picks the mode."** (Retired with the transport metaphor: "the trip picks the mode," "nobody apologizes for walking.") ## The four modes (also the four levels) ### Chat You ask, AI answers in real time. Conversational, ephemeral. You bring the question, the AI brings the response, and you stop when the answer is good enough or your question changes. Chat is the right mode for one-off questions, exploratory thinking, debugging unfamiliar problems, and any work where the next move depends on what AI just said. The investment per use is near zero; the same is true of value retained between uses, because nothing persists. People at the chat level use AI mostly here. People at higher levels keep using chat for the work that fits this mode — and there's a lot of it. ### Context Your context — role, style, reference material, constraints — gets loaded once. The assistant arrives knowing the things you'd otherwise have to re-explain. You drive each conversation, but the assistant is a persistent collaborator you keep returning to for a category of work: drafting emails in your voice, editing against your style guide, planning content against your brand. Investment goes up because you maintain the loaded context; value also goes up because the assistant gets sharper with use. People at the context level have built one or more custom assistants. They still use chat for ad-hoc questions; the custom assistants handle their recurring categories of work. ### Clockwork The AI runs without you initiating each execution. The trigger is a schedule (every Friday morning), an event (when an email matching a filter arrives), or a manual on-demand kick (you say "run the weekly digest" and walk away). You review the output afterwards. The mode is a saved instruction plus a recurrence plus an output destination. Investment is in the design and the source bounding; value compounds through the schedule itself, because it runs whether you remembered or not. People at the clockwork level have one or more workflows running without their initiation. They still use chat and custom assistants for everything that doesn't fit the clockwork mode. The mode is called clockwork; the thing you build in it is still called an automation, the same way the context mode produces a custom assistant. ### Colleague The AI works the way a good colleague does. It notices problems before you raise them and takes them on. It works with you and with others on your team, rather than only answering you. And it iterates on its own work until it thinks the result is ready for your review. You review what it hands over; you don't drive each step. The mode is initiative plus collaboration plus iteration to a standard. Investment goes into the standing brief that lets it judge what matters and what "ready" looks like, and into the escalation rules that tell it when to bring something to you instead of deciding alone. Value comes from delegating both the execution and the judgment about when the work is done. This is the highest-leverage mode and the most demanding to design, because the AI has to hold your priorities well enough to act on them without asking. People at the colleague level have at least one such colleague running — an AI chief of staff is the canonical example. They keep using chat, custom assistants, and automations for the rest of their work, because the colleague mode is overkill for most tasks. ## The scale is for your skills, not your workflows The most common misreading of the scale is that you climb it and leave the earlier modes behind. The truth is that maturity is measured by how many modes you have working in concert, not by which one you've reached. A founder who built three custom assistants a year ago and a scheduled automation last month has accumulated capability — but they still answer ad-hoc questions in chat all day, because that's what chat is for. The reason the modes get named in scale order is that the learning progression is real. Chat doesn't require any setup; building a custom assistant requires having context worth loading; designing an automation requires the human-review patterns developed at the context level; designing an autonomous loop requires understanding the verification standards that make automation safe. Each mode builds on capabilities developed at the previous one. Skipping ahead is possible but rarely advisable for work the user hasn't already done at lower levels. But once accumulated, all four modes coexist. The right design question for any new workflow isn't "what's my current level" — it's "which mode fits this work." The job picks the mode. ## Choosing a mode for a specific workflow After a workflow has been identified and decomposed, the mode for it follows from a few questions about the work itself. **How often does this happen, and on what trigger?** If the trigger is "when something occurs to me," chat is the right mode — there's no recurring pattern to automate. If the trigger is "every Friday" or "when a new ticket arrives," clockwork fits. If the trigger is "I notice this category of work and want to do it well each time," context (a custom assistant) fits. **Who initiates each execution?** If the human starts each run, the mode is chat or context. If the trigger is external (time, event), the mode is clockwork. If the AI is expected to notice the work itself, take it on, and bring it to you when it judges it ready, the mode is colleague. **Where does the output land?** Chat output stays in the conversation. Context output is the conversation plus whatever the human copies out. Clockwork output goes to a destination the human checks — a folder, a channel, a doc. Colleague output is work brought to you for review when the AI judges it ready, with the iteration history usually left behind. Choosing the destination is part of choosing the mode. **What's the failure cost of one bad run?** High-cost single runs — legal advice, client communication, irreversible actions — push toward chat or context, where the human reviews each step. Low-cost single runs — a digest, a draft, a list of suggestions — make clockwork or colleague modes safe to design. The cost is what determines whether the human stays in the loop per-run or per-batch. ## How this changes the design conversation Naming the mode early focuses the design questions on what actually matters. Each mode has different design considerations: **Chat**: how do I phrase the question so the answer is useful? What context do I provide? Where does the AI fall short on this kind of question, and how do I verify? **Context**: what reference material does the assistant need loaded? What context stays the same across all conversations, and what gets supplied per-conversation? How does the assistant get better with use? **Clockwork**: what sources does it read, and are they bounded? Where does the output land? What triggers a notification? What happens if a run fails silently? **Colleague**: what does it need to know to judge what matters? When does it bring something to you instead of deciding alone? What does "ready for review" look like, and how many iterations before it stops and asks? Naming the mode after a workflow has been identified produces concrete follow-up questions. Naming a level without a workflow attached produces vaguer reflection, which is useful for orientation but not for design. ## If you know the AI Fluency framework The AI Fluency framework (Dakan and Feller, taught by Anthropic) names three interaction modalities: Automation, Augmentation, and Agency. Those describe how the AI behaves in a given exchange. The four modes here describe what kind of work is being done and what the human's relationship to the AI part is. The two schemes answer different questions and coexist without conflict. Roughly: chat and context cover their Automation and Augmentation, since a human initiates each exchange and either instructs or co-works. Clockwork and colleague sit in their Agency, since the human configures behavior once and the AI acts on future work without being asked each time. The mapping is approximate by nature — context mode configures a persistent assistant, which is Agency-flavored, yet the human still opens every session, which is not. The word **automation** is the trap. In the AI Fluency framework it means the AI executing a task on direct instruction right now, such as drafting one email. In this framing it used to mean the opposite end of the spectrum, a workflow running on a schedule or trigger without the human present. Anyone holding both vocabularies would picture the wrong thing. The third mode was renamed **clockwork** on 2026-09-09 for exactly this reason. The artifact built in that mode is still called an automation, which is unambiguous because it names a thing rather than a mode. ## Terminology note This framing was called **"shapes of AI work"** until 2026-07-31, when it was renamed to **modes** (plain, non-hierarchical, needs no explanation). A transport metaphor (walk/bike/train/plane) served as standing imagery from 2026-07-31 to 2026-08-26, then was dropped — see "No metaphor" above. When drafting new content, use "modes" with plain definitions and concrete examples; no transport imagery. "Shapes" and transport lines survive only in already-published material. The third mode was called **automation** until 2026-09-09, when it became **clockwork**. The rename was not a metaphor decision — clockwork names the mechanism plainly (set it once, it runs, you review what comes out) and it does not rank against the other three. It resolves a genuine collision with the AI Fluency framework, where "automation" means close to the opposite. Note the one weakness: clockwork sounds time-only, so definitions must say explicitly that an event trigger counts. Published material carrying "automation mode" is corrected as it is touched, not retroactively. ## Related pages - See: what-is-a-workflow — What a workflow is and how its steps split between AI and the person, the step that comes before choosing a mode - See: ai-workflow-redesign — Methodology for identifying which workflows in someone's life are good candidates for which mode - See: scheduled-automation — Detailed pattern for the clockwork mode, including cross-tool implementation - See: ai-fluency-framework — How these four modes line up against the three interaction modalities in the AI Fluency framework, including the automation-versus-clockwork naming collision - See: example-image-generation — Concrete example of a colleague mode with self-evaluation before human review - See: example-email-drafter — Concrete example of a context mode with loaded style guidelines - See: agent-design-principles — How mode choice affects agent design, especially the "use the dumbest agent that can do the job" principle for each mode