# AI Personal Assistant for Ops Pros: Slash Commands URL: https://commandergpt.app/lp/ai-personal-assistant Type: landing Locale: en Published: 2026-10-29 Updated: 2026-11-01 --- > CommanderGPT is an AI personal assistant built on slash commands. Research, summarize and draft with saved prompts, 30 days of memory and shareable team playbooks. *AI personal assistant for ops professionals* ## An AI personal assistant you command CommanderGPT turns your repeat prompts into slash commands. Research, summarize and draft in seconds, with 30 days of context memory. ## The assistant work that eats your week A personal assistant handles prep, lookups and first drafts. Each of these maps to a command you can run today. ### Research on demand Run /research with a prospect, a competitor or a topic. You get a sourced brief before the call, not a blank tab and twenty minutes of searching. ### Summaries you can act on /summarize turns a long thread, doc or transcript into the decisions, owners and open questions. Paste it, run it, forward the result. ### First drafts in your tone /draft-email writes the first version using the context already pulled. You edit two lines instead of starting from a blank page. ### 30-day context memory The assistant remembers prior work across sessions for 30 days. A returning contact or a running project does not restart from zero. ### Multi-model choice Pick Claude, GPT-4o or Gemini per command. Use a fast model for sorting and a stronger one for anything a customer will read. ### Chains, not one-offs Link /research, /summarize and /draft-email in the Workflow Builder. One trigger runs the whole sequence in order. ## From repeat prompt to saved command 1. **Write the prompt once** — Brief it like a new hire: the goal, the format you want, what to skip. Use a real task from last week, not a toy example. 2. **Save it as a slash command** — Name it, assign a model, and type / to see it filter in live. Your long prompt is now three keystrokes. 3. **Chain commands into a workflow** — Connect commands in the Workflow Builder so the output of one feeds the next. Add Slack, Notion or Linear where your work already lives. 4. **Share it with your team** — Run /share to publish the chain as a Team Playbook. Edit it once and everyone gets the new version on the next run. *Use case: sales ops* ## Meeting prep in three commands An account executive has ten minutes before a deal review. Before the playbook, prep meant five tabs, a CRM export and a half-remembered thread. With a saved chain, /research pulls the company context, /summarize condenses the last month of notes, and /draft-email prepares the follow-up with open questions flagged. Time the old way for one week, then compare after two weeks on the chain, so the number you report is yours. The assistant does not know your pipeline unless you paste it in, and it will not judge deal quality for you. It applies your rules and shows its work, so you can fix the rules when one is off. - One chain replaces five tabs of prep - Follow-up drafted with open questions flagged - Rules stay editable text, not hidden settings *Use case: customer success ops* ## One setup for the whole team A CS team rotates through the front line each week, and each person arrives with a personal prompt and a different tone. The tool was never the problem, consistency was. One person builds the triage and reply chain, tests it on past tickets, then shares it as a Team Playbook. When escalation rules change, the owner edits it once. Be realistic about limits: the assistant only sees what you give it, so anything decided on a call needs a short written note before the chain runs. Humans still read and send every reply. - Same tone and rules for every teammate - Versioned playbook, one edit reaches everyone - Human review before anything is sent ## Where CommanderGPT fits against the usual picks | Capability | CommanderGPT | ChatGPT | Raycast AI | |---|---|---|---| | Saved custom slash commands | Yes, core feature | Yes, with custom GPTs and presets | Yes, from the command palette | | Chain commands into a workflow | Workflow Builder | Manual, prompt by prompt | Limited, extension dependent | | Share a setup with a team | Team Playbooks via /share | Shareable GPTs | Per-user by default | | Choose the model per command | Claude, GPT-4o or Gemini | OpenAI models | Several models | | Context memory across sessions | 30 days | Memory feature | Not the focus | | Plugin ecosystem and ubiquity | Smaller | Largest of the three | Strong on Mac | ## Questions ops leads actually ask ### What is an AI personal assistant, and is CommanderGPT one? An AI personal assistant takes work off your plate: research, summaries, drafts, prep. CommanderGPT does this through slash commands you run on demand, such as /research, /summarize and /draft-email. It is a workspace you direct, not an autonomous agent that acts on its own. You decide when it runs and you review the output before anything leaves your hands. ### How is this different from just using ChatGPT? ChatGPT is a strong general assistant and has the bigger plugin ecosystem. The difference is repeatability. In CommanderGPT you save a prompt as a named command, chain commands into a workflow, keep 30 days of context memory, and share the result with a team. If you retype the same long prompt several times a day, that is where the gap shows up. ### Which models does it use? You choose per command. Claude, GPT-4o and Gemini are available, so a quick classification step can run on a fast model while a customer-facing draft runs on a stronger one. Model quality changes often, so test your own commands on your own material before you commit a workflow to one model. ### Does it remember what I worked on yesterday? Yes, within limits. Context memory carries across sessions for 30 days, so a returning prospect or an ongoing project does not start from zero. It only knows what you gave it. If a decision happened in a call nobody wrote down, add a short note before you run the command. ### Can my whole team use the same assistant setup? Yes. Build a command or a chain once, test it, then share it with /share as a Team Playbook. When you edit the playbook, the next run uses the new version for everyone. This is the part that matters for ops teams, where consistency across five or fifteen people is the real problem. ### Will it send emails or book meetings for me? No. It prepares drafts, briefs and research, and you send. It does not act as an autonomous agent with access to your calendar or outbox. For most ops work that is the right trade: the judgment call stays with a person, and a wrong draft costs you ten seconds, not a customer. ### How long does setup take? Plan on about an hour to get your first useful command running: write the prompt as you would brief a new hire, test it on three real examples from last week, fix what is off, save it. Then connect Slack, Notion or Linear if your workflow lives there. The time saved per run is worth measuring in your own context. ## Set up your first assistant command Take the prompt you retype most often. Save it as a slash command and run it on three real examples this week. *Call to action: Try CommanderGPT* ## FAQ ### What is an AI personal assistant, and is CommanderGPT one? An AI personal assistant takes work off your plate: research, summaries, drafts, prep. CommanderGPT does this through slash commands you run on demand, such as /research, /summarize and /draft-email. It is a workspace you direct, not an autonomous agent that acts on its own. You decide when it runs and you review the output before anything leaves your hands. ### How is this different from just using ChatGPT? ChatGPT is a strong general assistant and has the bigger plugin ecosystem. The difference is repeatability. In CommanderGPT you save a prompt as a named command, chain commands into a workflow, keep 30 days of context memory, and share the result with a team. If you retype the same long prompt several times a day, that is where the gap shows up. ### Which models does it use? You choose per command. Claude, GPT-4o and Gemini are available, so a quick classification step can run on a fast model while a customer-facing draft runs on a stronger one. Model quality changes often, so test your own commands on your own material before you commit a workflow to one model. ### Does it remember what I worked on yesterday? Yes, within limits. Context memory carries across sessions for 30 days, so a returning prospect or an ongoing project does not start from zero. It only knows what you gave it. If a decision happened in a call nobody wrote down, add a short note before you run the command. ### Can my whole team use the same assistant setup? Yes. Build a command or a chain once, test it, then share it with /share as a Team Playbook. When you edit the playbook, the next run uses the new version for everyone. This is the part that matters for ops teams, where consistency across five or fifteen people is the real problem. ### Will it send emails or book meetings for me? No. It prepares drafts, briefs and research, and you send. It does not act as an autonomous agent with access to your calendar or outbox. For most ops work that is the right trade: the judgment call stays with a person, and a wrong draft costs you ten seconds, not a customer. ### How long does setup take? Plan on about an hour to get your first useful command running: write the prompt as you would brief a new hire, test it on three real examples from last week, fix what is off, save it. Then connect Slack, Notion or Linear if your workflow lives there. The time saved per run is worth measuring in your own context.