For AI product manager · 9 Grok Bots

An AI-product back office that measures quality before you ship it

Grok Bots for a solo AI product manager: turn model failure cases into a versioned eval set, draft specs that pin the model and the acceptance evals, score every version before rollout, and track cost, latency, and drift — all as drafts you approve.

Hire the team See the bots
● Drafts only — nothing sends without your approval · Uses only your own data
After checkout
Your setup link arrives by email
We set up your fleet
After checkout, Workplace AI provisions your fleet on our Cursor Teams workspace and follows up soon. Watch for your Cursor invite and Grok Bot download link — then accept, install, and go.
9 bots·3 domain specialists·Success metric: Eval-backed change rate and roadmap currency·Category: AI roles
The problem it targets

Owning an AI feature alone between decisions — triaging feature asks and reading raw model transcripts to find where the assistant goes wrong, hand-building eval sets from screenshots and Slack threads, writing AI-feature specs that have to pin a model, a prompt, guardrails, and pass/fail evals from a blank page, eyeballing new prompt or model versions for regressions, and reconciling token cost, latency, and quality drift from exported telemetry — so real failure cases get lost, specs ship without an acceptance eval, a prompt change silently regresses, and the weekly quality-and-cost read is a scramble.

What you get

From day one

1

Every AI feature request captured and each real model failure case logged from your own transcripts, so the pattern is visible and nothing is lost in a channel.

2

A versioned eval set with a written rubric, an AI-feature spec that pins the model, prompt, and guardrails, and a version review that scores each candidate before rollout.

3

A cost, latency, and drift read per feature and a one-page weekly AI roadmap and eval scorecard, every number traced to its source.

Meet the team

9 bots, each with one job and clear approval points

Read-only bots watch and prepare; drafting bots write but never send. You approve everything that leaves.

1

Chief of Staff

coach
Drafts for your review — you approve before anything leaves
Your morning brief, meeting prep, fleet roundup, and after-meeting coaching.
Approves: none — nothing leaves without you
2

Signal & Failure-Case Triage

intake
read-only
Watches & prepares — never sends
Capturing feature asks and logging real model failure cases from transcripts
Approves: none (nothing leaves; the log is internal)
3

Eval Set & Rubric Curator

researcher
read-only
Watches & prepares — never sends
Versioned eval sets and pass/fail rubrics drawn from real usage
Approves: none (nothing leaves; the eval set is internal)
4

AI-Feature PRD Drafter

drafter
specialist
Drafts for your review — you approve before anything leaves
Specs that pin the model, prompt, guardrails, and acceptance evals
Approves: every spec before it is shared
5

Model-Behavior Version Reviewer

specialist
specialist
Drafts for your review — you approve before anything leaves
Case-by-case version diffs against the eval rubric, regressions flagged
Approves: every rollout; any guardrail regression
6

Launch-Readiness & Rollout Coordinator

coordinator
Drafts for your review — you approve before anything leaves
Drafted launch-readiness and rollback checklists and go/no-go notes
Approves: every message; every flag flip and rollout
7

Cost, Latency & Drift Tracker

tracker
read-only
Watches & prepares — never sends
Per-feature cost, latency, and quality-drift reads from telemetry
Approves: none (nothing leaves; thresholds proposed)
8

AI Roadmap & Scorecard Reporter

reporter
read-only
Watches & prepares — never sends
Weekly roadmap and eval scorecard with every number traced to its rows
Approves: none (nothing leaves; the PM commits)
9

Social & Referrals

growth
Drafts for your review — you approve before anything leaves
Draft posts and outreach to your own audience, credited and approved by you.
Approves: sends
The handoff — one shared chat
Triage→Curator→Drafter→Reviewer→Coordinator→Tracker→Reporter
What the fleet does in week one

All read-and-prepare — you check its work before anything is trusted

What it does not do

The guardrails are the product

  • ×It drafts and scores; it does not ship. It never promotes a prompt or model version to production and never flips a feature flag — that stays with you and engineering.
  • ×It works only from your own exported transcripts, telemetry, and opted-in contacts. It does not reach anyone you have not chosen or send an update on its own.
  • ×It never touches production or live settings, moves money, or commits the roadmap — the call is yours.
  • ×It reports what the eval covers and never claims a quality or cost result it cannot trace to a source.
Eval-backed
change rate and roadmap currency
How you'll know it's working

One number, measured over time

Share of prompt and model changes that were scored against a versioned eval set before rollout, share of shipped AI-feature specs that carry an acceptance eval, and whether the weekly scorecard reflects live telemetry rather than memory

Before day one

What you fill in — the bots never guess these

failure-case-log
the header contract for the failure-case and feature-ask log every bot reads.
eval-set
the shape of a versioned eval case and its rubric; the only source of the eval bar.
telemetry
the header contract for your exported token-cost and latency telemetry.
Pricing

Cancel anytime.

Workplace AI provisions and runs your CrokPot fleet on our Cursor Teams workspace. We invite you on your own email — you accept, install Grok Bot, and we handle the rest.
$200/mo
  • Your whole fleet provisioned and installed for you
  • Cursor Teams seat included — paid by Workplace AI, never billed by Cursor
  • Onboarding + ongoing training
Hire the team
Your Cursor Teams seat is included and paid by Workplace AI. Drafts only, using your own data. No earnings, time-saved, or outcome guarantees are made.
How access works

You buy a CrokPot fleet from Workplace AI — not a Cursor plan

CrokPot runs on Workplace AI, Inc.’s (WAI) Cursor Teams workspace. WAI keeps the unpaid admin seat and pays Cursor for your seat. You join as the invited user on your real email and work through the installed Grok Bot fleet.

Your card is charged by WAI via Stripe for the CrokPot fleet ($200/mo, auto-renewing monthly) only — never by Cursor. If you cancel your fleet subscription, the invite/seat ends with it and is not portable to your own Cursor account.

FAQ

What you get, who owns the seat, and what happens after you pay

From the team behind Ancestry — CrokPot is built by Paul Allen, a co-founder of Ancestry.com.

Drafts only — you approve everythingYour own data, never shared$200/mo · cancel anytimeProvisioned & supported by Workplace AI

What do I actually get?

A pre-built fleet of Grok Bots for your trade, provisioned and installed for you on Workplace AI’s Cursor Teams workspace, plus onboarding and ongoing training. Every bot drafts; you approve.

Who owns the seat?

Workplace AI, Inc. (WAI) owns the Cursor Teams account and pays Cursor for your seat. You’re invited as a paid user on your own email. The seat isn’t portable to a personal Cursor account.

Do the bots act on their own?

No. The fleet is drafts-only, working from your own data. Nothing sends, files, or is trusted until you review and approve it — that’s the whole design.

Can I cancel?

Yes, anytime. It’s an auto-renewing monthly subscription at $200/mo. If you cancel, your invite/seat ends with the subscription.

What happens after I pay?

We follow up soon: our team provisions your fleet and invites you on the email you used at checkout. Watch for your Cursor Teams invite and Grok Bot download link, then accept, install, and go.

Training

New to Grok Bots? Get your team trained.

Live sessions and replays to get you and your coworkers productive with your fleet fast.

Explore Grok Bot training