Research, deck, send

The situation

Tomorrow morning you have to say something credible about a competitor, or about a market someone wants you to move into, and you have nothing written down. You need research you could defend if pushed, a deck in your own template rather than a generic one, and the result in a colleague's hands tonight.

That is three sorts of work: digging, designing and sending. You describe the outcome once and Player One works out the shape.

What you need

Your template, if you have one. An old deck you liked, a template, a brand guide, or just a logo and two colours. Whatever you supply becomes the design: Player One has no house style of its own to impose. Without one it picks a direction and says what it picked.

A way to send it, if Player One is doing the sending. Email is the best supported route: it attaches the file, takes a subject and cc, and goes from your own connected mailbox. Connect one under Admin > Messaging (/admin/channels), or follow Connect a channel. Telegram, Slack and WhatsApp carry files too; with nothing connected, download the deck and send it yourself.

The conversation

Attach the template with the paperclip and say what you actually want.

"Research how Northlake are positioning themselves against us: pricing, the claims on their site, anything published in the last year. Then build me a ten-slide deck in the attached template for tomorrow's board meeting."

Expect a question or two first, about the angle and the depth. Then the assistant will say this is bigger than one conversation and offer to run it as a job: the research and the design are different work, done best by different agents.

"Yes, run it as a job, and let me see the plan before the deck gets built."

The job appears under Automations > Jobs (/jobs) and your chat comes straight back to you. Open it and the Plan tab holds the coordinator's plan: the questions the research splits into, and who is taking each. Work you asked to approve waits on the Tasks board with a Review Plan button. Approve it, or reject it with a reason that goes back to the coordinator.

The research lands first: the summary and key takeaways in your chat, the full report saved in the job's files. Every claim carries its source, anything only one source says is flagged as unverified, and where it found nothing solid it says so.

The deck is then drafted in your template and opens in the preview pane beside the chat, so you can page through the slides as they take shape. Every slide is rendered and inspected before it is handed over, which is how a clipped headline or a substituted font is caught rather than delivered.

Read it and push back in ordinary words.

"Slide 4 is doing two jobs, split it. Cut the last two bullets on slide 7 and give the pricing table its own slide."

You get a new version rather than a new deck: the parts you liked stay put. When it is right, ask for it to go.

"Email the deck to Priya with two lines saying it is a draft for tomorrow's board."

It confirms the address and the wording first: Player One does not send outward quietly.

What you end up with

An editable PowerPoint file, not a picture of one, in the job's Files tab and under Manage > Files (/workspaces). Ask for a PDF as well and you get one beside it.

The research report with its sources, saved next to the deck, so a claim on slide 3 traces back to the page it came from.

A record of who did what: the Team tab lists the coordinator and every agent that worked, each with a conversation you can open, and the Activity tab is the timestamped trail of every change.

Make it yours

Ask for a one-page brief instead of a deck, or as well as one. The research is the expensive part, and it feeds either.

Make it monthly. The same request can run to a schedule: see Schedule something. If it has to follow a set process every time, that is a flow rather than a task: Make it repeat covers the choice.

Add a reviewer. Ask for a second agent to check the finished deck against the research before you see it.

If it stalls

A job that hits a decision only you can make parks itself and asks. An amber card appears in the chat you started it from, and the job shows as Needs input until you answer there in your own words.

If the plan is wrong, reject it with a reason rather than letting it run and repairing the output later.

Stop, Resume and Close job are on the job page, and Stop asks you to confirm. Stopping is a pause rather than a delete: resuming brings back the tasks, the team and the shared folder as they were, and takes about half a minute to get moving. A frozen-looking job is sometimes waiting on capacity from your AI provider and will pick itself up.

Related

• Give it work, for when a job is the right size.
• Jobs, for the tabs, the states and how a job asks you things.
• Make things, for decks, PDFs and cited research.
• Create your first job, if this is your first.