The Chatbot Era Is Over. Here Are the Three Skills That Replace Prompting.

If your AI workflow is still "open a tab, type a question, copy the answer, paste it somewhere" — you're using a version of AI that the people generating the most output from it have already left behind. The gap between how they work and how most professionals work is now measurable, and it's growing fast.

Five signals, one conclusion

On June 25, Ethan Mollick co-authored an OpenAI research paper (arXiv:2606.26959) using internal Codex data. Agentic AI users grew more than fivefold in the first half of 2026 — and the fastest growth wasn't among engineers. The median OpenAI legal employee now generates 13 times more monthly output via Codex than in January. Researchers: 50 times more. Mollick's conclusion, in his own words: "The chatbot era is over."

These aren't people who learned to prompt better. They changed what they hand over — from questions to tasks.

The same week, Andrej Karpathy called Anthropic's Claude Tag launch "the 3rd major redesign of LLM UIUX" — AI moving from a tab you open to a persistent Slack teammate with organisation-wide context. Anthropic's Economic Index (June 26) showed 93% of Claude conversations now produce a concrete artifact, and one-third of surveyed users expect AI to handle most of their work within 12 months. Google shipped computer use directly into Gemini 3.5 Flash (June 24), enabling agents that click screens and fill forms at roughly one-fifth the cost of GPT-5.5.

Five independent signals. Same week. Same direction.

The three skills that replaced prompting

Delegation. Scoping what an agent should do before it starts, rather than correcting what it did after. That means defining inputs, outputs, permitted actions, and the specific points where the agent must pause and confirm. If you can't write a two-paragraph brief for an agent task, the agent isn't ready to run on its own.

Supervision. Knowing when to intervene and when to let it work. Always-on agents like Claude Tag have ambient context your one-off prompts never did. An agent that reads your channels and drafts responses is useful; an agent that sends them without review is a different proposition entirely, and the supervision design is what separates the two. Supervision isn't reviewing every sentence — it's defining the threshold at which a human must check before work continues.

Workflow design. Building multi-step processes where AI handles execution and humans handle judgement. Not "write this email" but "here's our response process for supplier complaints: step one, flag and categorise; step two, draft based on complaint type; step three, human review before send." The Codex users generating 50 times more output aren't prompting more. They've built workflows with defined handoff points.

Serena: from 40 manual summaries a day to one review queue

Serena manages client intake at a 90-person legal services firm. Her day used to start with 30 to 40 intake emails that needed reading, summarising, categorising by matter type, and routing to the right attorney — all before 10am. She'd been using AI for the summarising step only, one email at a time, copy and paste.

She redesigned the whole process. An agent now reads intake emails, assigns a matter category, drafts a three-sentence summary, and flags urgency. The output drops into a Slack channel for her review. She confirms the routing and presses send. The agent never sends anything — that step is non-negotiable.

Her throughput more than doubled, and her routing accuracy improved, because she now has time to actually check the categorisation instead of rushing through summaries. The delegation design took two hours. The supervision decision — what the agent can do without her, what needs confirmation — took thirty minutes and was the most valuable thinking she did that week.

She'd stopped treating it like a search bar.

Marcus: what happens when you skip the design step

Marcus runs operations at a 15-person architecture practice. He deployed an agent to organise project files by naming convention. Three days in, it moved files that were live in client presentation folders, breaking shared links mid-project. His client noticed before he did.

The error wasn't the agent. It was the scope. Marcus hadn't written the one rule that would have caught it: never move files referenced in active project folders.

The scope document he now requires before any agent task includes the question that was missing — what should this agent never do without asking me first? He treats agent setup the way he treats a new hire's first week: supervised on day one, spot-checked by day five, running independently on defined tasks by week two.

The one thing

The chatbot era taught you to prompt. The agent era requires you to manage — scope the task before it starts, define the supervision threshold rather than reviewing every output, and build the workflow that keeps you at the right decision points.

That's harder, worth more, and something you can start practising this week.