Giving a talk is a research, drafting, rehearsal, and delivery problem in one. Each part feeds the next, and each tends to live in a different app — research in browser tabs, drafts in Google Docs, rehearsals as voice memos on your phone, post-talk notes never written down. By the third or fourth talk, you're rebuilding context from scratch each time, which is why most working speakers feel like the talk is taking over their week.
This guide is about putting the parts of a speaker's work that benefit from being together — research, drafts, rehearsal recordings, post-talk notes — into one vault where the agent can read across them.
What a speaker's notes actually have to hold
A speaking practice over time develops repeating elements. Themes you keep coming back to. Stories you find yourself telling again because they land. Data points you cite repeatedly. Audiences you've spoken to before and might speak to again. Most speakers carry all of this in their head and slowly forget the parts they don't use weekly.
Four loads a speaker's vault has to carry:
- A research file per talk — the source material, citations, articles, and inputs you've gathered for this particular talk.
- A talk archive — every talk you've given, with the script or outline, the rehearsal notes, and the post-talk debrief.
- A story library — the anecdotes, examples, and case studies you reach for. Tagged so the agent can pull the right one for the right audience.
- An audience file — pages per recurring audience type or specific organization, with what worked and what didn't.
A vault holds all four together. Plain markdown for the text. Audio (rehearsal recordings, the actual delivery if you have it) transcribed with speaker labels. PDFs (research papers, conference programs, audience demographics) auto-converted to searchable text.
The talk page — one place per talk, deepening over time
Each upcoming or past talk gets its own page. The page lives the whole arc of the talk — research, drafting, rehearsal, delivery, debrief. Six months later when you give a similar talk, the page is the starting reference.
A typical talk page has sections for:
- Brief — what the talk is, the audience, the venue, the time slot, the ask from the organizer.
- Research — sub-pages for the source material you've gathered.
- Outline — the working structure.
- Draft — the actual script or detailed outline.
- Stories — links to entries in the story library you're using.
- Rehearsal log — recordings and notes from each rehearsal.
- Debrief — what landed, what didn't, what you'd change.
The discipline of one-page-per-talk makes the post-talk debrief useful. When you're invited next year to a similar event, you open the page from this year's talk first. The debrief tells you what to keep, what to change, what didn't work. You don't re-derive lessons you already learned.
The research pass — let the agent do the reading
Most talks involve research. A keynote on a current topic. A workshop that builds on academic work. A conference talk that needs to cite a recent study. The reading takes hours and the synthesis takes more hours.
The pattern: drop the research material on the talk's research sub-page. PDFs of papers and reports. Articles you've read. Books you're drawing from. PDFs auto-convert to markdown via docstrange so they become searchable text the agent can read end-to-end.
Then ask: "Read the research material on this talk's page. Pull the three findings that are most relevant to my central argument. List the citations I'll need to verify before stating each finding on stage. Note any contradictions in the research that I should address rather than skip."
You get a synthesis pass. Not a finished talk — a starting point. The agent has done the reading; you do the judgment about what to actually use. The same shape works for any research-heavy creative work; see how to use AI notes for memoir writing for a longer-form variant of the same workflow.
For citations specifically, the agent's pulling matters because the verification is on you. The agent can find the passage. The speaker still needs to read the passage in context before citing it on stage. A misremembered statistic or a misattributed quote becomes the moment everyone remembers from the talk, and not for the right reason.
The story library — anecdotes that compound
Most working speakers have a small set of stories they reach for. The client meeting that taught them something. The personal anecdote that opens a particular kind of audience. The historical example that frames a complex idea. Most carry these stories in their head and lose track of which ones they've used in front of which audiences.
A "Stories" sub-vault, with one page per story, captures them. Each page has the story written out (one or two paragraphs), the lesson it illustrates, tags for the kinds of talks it suits, and a log of the talks where you've used it.
When you're drafting a new talk, the agent reads the story library: "Pull every story I've used to illustrate the theme of long-term thinking. For each, note which audiences I've told it to and how it landed. Suggest one I haven't used in over a year that might fit this talk." You get a working list, with the institutional memory that prevents you from repeating the same story to the same audience.
For the story you've decided isn't working anymore, archive it with a note. The next time you draft a talk on a related theme, the agent reads the archive note and skips the story. You don't accidentally bring back the anecdote that's been retired for good reason.
The drafting flow — outline, draft, rehearsal
The drafting flow varies by speaker — some work from a detailed script, some from an outline, some from a few headline cards. The vault supports any of them.
For script-style speakers, the draft lives on the talk page as a markdown document. The agent helps with structural work that's hard staring at the page. "Read the draft. Where does the talk slow down? Are there sections that feel told rather than shown? What's the strongest moment, and is it in the right place?" You get observations to react to. Some you ignore. One is useful.
For outline-style speakers, the talk page holds the outline with sub-bullets for each section's intent. The agent helps with the structural pacing: "Each section is roughly how long? Where am I underweight on stories versus arguments? Which transitions feel forced?"
For the assembly part of drafting — pulling stories from the library, citations from the research, references from past talks on the theme — the agent does the boring work. "Draft an opening for this talk that uses the story about the client meeting in 2022, transitions into the central argument, and lands on the question I want the audience to leave with." You get a draft. You rewrite it. The blank-page problem is gone.
Rehearsal recordings and the agent that can read them
Rehearsal is where most talks get better, and most rehearsals get wasted because nobody captures what was learned.
The pattern: rehearse out loud. Record on your phone. Drop the audio onto the talk's rehearsal log. It transcribes with speaker labels — useful if you're rehearsing with a coach or a colleague who's giving feedback.
The transcript is searchable. Now between rehearsals, you don't relisten to forty minutes of audio. You ask the agent: "In yesterday's rehearsal, find the moments where I stumbled. Pull the sections where the wording felt awkward. Note any places where I said something that wasn't in the script — sometimes those are the better lines."
You get the relevant moments back, in plain text. You revise. The next rehearsal is shorter because it's targeted at the parts that needed work, not a full re-run of the whole talk.
A small habit that compounds: after each rehearsal, type three lines about what felt different this time. Over the run of the talk's preparation, those three lines become the most useful pages in the rehearsal log.
The audience file — what worked, what didn't, what to do differently
A speaking practice over years sees the same kinds of audiences repeatedly. The annual industry conference. The internal company offsite. The university lecture. The trade association keynote. Each kind has its own dynamics, and each specific organization has its own.
An "Audience" sub-vault, with pages per audience type and per recurring organization, captures the institutional memory. What worked with this audience before. What didn't land. What questions came up in Q&A. What the organizer asked you to emphasize that you'd otherwise skip.
Before a return engagement, the agent reads the relevant page: "I'm speaking again at the Acme annual conference. Pull what worked and didn't work last time. Note the questions that came up in Q&A. Suggest two things I could do differently this year." You start the prep with two years of context, not from scratch.
The post-talk debrief — three lines, two minutes
Most speakers skip the post-talk debrief because they're tired and the next thing is already on the calendar. But the debrief is the only place the actual lessons get captured.
Three lines on the talk page, within a day. What landed best. What landed worst. What you'd change. Two minutes of typing.
Over a year of talks, those debriefs become the most useful asset in the vault. They're the only place you've honestly recorded what worked in front of an audience versus what you assumed would. The next talk on a similar theme starts from the debrief, and it starts smarter.
For longer-form post-talk material — the article version of the talk, the LinkedIn post-mortem, the podcast clip — the same workflow applies as in content repurposing across platforms. The talk transcript becomes input for several other outputs; the vault holds them all.
Pricing, contracts, and the business side
A practical note: the vault holds the thinking about your speaking. It's not a contract management tool, an invoicing system, or a booking platform. The actual files stay in your billing tool. Contracts can be dropped in as PDFs for reference, but the legal source of truth lives wherever your e-sign service is.
For pricing context — past speaker fees, the rate range you charge for different formats, your standard rider — a small "Business" sub-vault is useful. When negotiating a new engagement: "Pull my last five speaker fees for keynotes at industry conferences. Suggest the rate range for this new prospect referencing what I charged for similar audiences." Grounded in your own pricing history. For storage and upload limits if your vault holds many large rehearsal recordings, see pricing.
A calmer way to give the talk
The talks get better when the system around them holds up over time. The research from this year compounds into the talk next year. The story you used last spring shows up in the library when you draft next fall. The debrief you wrote two years ago tells you what to do differently when you return to the same audience. The agent does the boring searches and the structural drafts. The speaker does the speaking.
Try Docapybara free — sign up, build a page for your next talk, drop in the brief and one piece of research material, and ask the agent for a draft outline grounded in both.