This page walks you through what we're building, how we're building it, and what it means for you and your team. Read it once before you start. The form at the end takes 20–30 minutes.
Imagine a version of your department where the answers your team needs are already there — where institutional knowledge compounds instead of walking out the door, and where your judgment scales further than your calendar ever could. That's what we're building. With you. From your playbook.
The SOP doc says one thing. You know what actually happens. The bots need to learn from the second version, not the first.
The dropped handoffs, the informal workarounds, the tribal knowledge that lives in your Slack DMs and nowhere else.
What a bot should never do without asking. What's too sensitive to automate. That's your call to make, not ours.
40 questions across 10 sections. Takes 20–30 minutes. It's a reflection exercise more than a quiz — there are no wrong answers. The questions that matter most are about pain, not performance. What drains your time? What do you wish you didn't have to do? What do you keep fixing? Those.
Your right here: skip any question that feels uncomfortable. Write "prefer not to say." Come back to it in the consult if you want.
Oracle is the analyst bot. It reads the metadata of how your week moves — Gmail, Calendar, Chat, Drive, Sheets, plus your primary tool (Close, Monday, Zendesk, or QuickBooks). Who you meet with, which sheets you edit, how long things take. One read-only pass over a ninety-day window, then it stops.
Exclude any folder, label, or sheet before the scan starts. The output lands in your inbox before anyone else's.
Oracle delivers three artifacts: the Reality Report, the Time Map, and the Pattern Inventory. Your intake answers next to what the data actually shows. Time maps. Shadow workflows. The sheets you live in.
Mark anything you'd like adjusted, contextualized, or held back from the consult. The version we work from is the version you sign off on.
In person. Two hours, blocked. We walk through the form, walk through the Reality Report, and design three bots specific to your department: a private copilot for you, and two bots that serve your team.
You define each bot's job, what it's allowed to do autonomously, what it must always check with a human on, and what it escalates to you. Nothing ships without your signoff.
Your right here: veto anything. Rewrite anything. Say "not now" to any bot. We can do one bot or all three. Your call.
We don't ship to the whole department at once. First week, it's you + 2–3 people you trust. We gather feedback, adjust, iterate. Then we roll out to the rest of the team.
30 days after launch, we check in together. What's working? What's not? What should the bot stop doing? What should it start doing? The bots evolve.
Your right here: pause, modify, or kill any bot at any point. No bot is permanent. You own this.
Oracle is the analyst bot. It runs once before your consult, reads the metadata of how your week actually moves — calendars, inboxes, the tools you live in — and produces three things you'll want on your desk regardless of this project. They show up in your hands twenty-four hours before the consult.
What you said in the intake, side by side with what the data shows. Where they line up, where they don't. The conversations the consult is built around.
A visual of where your hours actually went over the last ninety days. Meetings, threads, sheets, calls. The shape of your week made legible.
Every task or conversation that happened five or more times in the window — ranked by time cost and how cleanly a bot could carry it. Your roadmap to leverage.
Oracle reads metadata, not message content. Read-only OAuth, ninety-day window, runs once. Exclude any folder or label you want left alone — same way you would with any analyst.
Then the bots for your department get built from whatever public information exists and a generic template. They'll be mediocre. Your team won't use them. The project will skip your department.
You won't be penalized. But you'll also have missed a chance to shape a tool that could meaningfully improve your week.
Every job has automatable parts. The CEO's job has automatable parts. What makes someone valuable isn't the automatable work — it's the judgment, the relationships, the ability to handle the unexpected. None of that lives in a bot.
If a bot picks up 20% of your week, leadership's expectation is that you spend those hours on higher-leverage work — the strategic blocks the calendar never seems to have room for. That's the whole point of the project.
That's the goal. A team that depends on a single person for answers is a fragile team. It means when you're sick, on vacation, or in a long meeting, everything stops. It also means you're spending your brain on things only you can answer — even when they're not things only you should answer.
Your value isn't in being the bottleneck. It's in being the person who defines how work gets done and where it's going. The bots take over the former so you have more time for the latter.
Only you and Alex. Not your manager. Not the CEO. Not HR. Not peers in other departments. Only two people.
If something in your answers surfaces a concern that would normally go to HR — a safety issue, a legal issue, something involving harm — Alex will tell you directly before anything else happens. That's the only exception, and it's narrow.
Please do. The whole design of this form asks about what's draining you, what's not working, what you'd change if you could. The more honest you are about the rough parts of your week, the better the bots will be at handling them.
Nothing on the form gets fed into your performance review. There is no "you said X in your intake, therefore Y" path. The intake is a design document for bots, not a record of you.
You pause them. You change them. You kill them. At any point. We review monthly, but you don't have to wait for a review — you can make changes whenever you want.
Every bot has an autonomy level you set in the consult: autonomous, draft-for-review, or escalate-to-human. You can dial any bot down or up. You can take any task back.
Then we slow down. The point of the pilot cohort is to find out whether the bots actually help before rolling out to a whole department. If your people push back, we listen. If the objections are valid, we change the bots or pull them.
Bots that people don't trust don't get used. So building trust matters more than building features. We'd rather ship a smaller bot that your team loves than a bigger one they resent.
Six months from now, your team is answering their own common questions without interrupting you. Your weekly calendar has fewer "quick questions" and more strategic blocks. Your inbox has fewer repeat topics. You're not the bottleneck for things that don't need you to be.
And the bots are boring — no drama, no runaway behaviors, just quietly handling the work they were designed for. That's the win.
If you still have concerns, reply to Alex's email and ask. If you want to book the consult before filling out the form — also fine. There's no "right" order. Do what feels useful.
Questions? Concerns? Hit reply on the email that sent you here. Everything is on the table.