- 18 Sep, 09:14Media
Moved daily budget from Campaign A to Campaign B
From A $1,400 · B $900 to A $1,000 · B $1,300
Approved byMarketing lead
- 18 Sep, 09:12Media
Lowered the daily budget on Display remarketing
From $260 a day to $180 a day
Approved byMarketing lead
- 18 Sep, 09:06Creative
Added four headline variants to Retargeting
From 3 ads to 7 ads · new ones paused
Approved byAuto-approve policy · set by Marketing lead
- 15 Sep, 07:30Media
Paused ‘Autumn carousel 3’
From Live to Paused
Approved byPause policy, under $100 a day · set by Marketing lead
- 11 Sep, 09:20Lifecycle
Changed the wait before the win-back flow, still a draft
From 30 days to 21 days
Approved byOwner
- 11 Sep, 09:08Media
Moved daily budget from Brand search to Shopping
From Brand $600 · Shopping $800 to Brand $450 · Shopping $950
Approved byOwner
Work gets done.You stay in charge.
Agents draft, build and propose. A named person approves anything that reaches an audience or spends money, every write lands paused or as a draft, and every change can be reverted.
Product preview
A week of work, and the moment it stops.
Watch the work arrive, and notice where it halts and waits for you. Select an agent to see what it may do alone and what it may never do without a person.
Product preview · scripted demonstrationAgents working
- Starting…
Who runs these workflows today
Nobody can buy an agent from Troy yet. The same work is done for clients of Troy’s team by strategists, and that work is where each agent comes from.
- 01
A strategist runs it
Troy’s strategists run client marketing on an internal console assembled from third-party tools, the platforms’ official connectors and AI workflows.
Available now · Troy’s team - 02
Three times makes a playbook
Every strategist task is logged with its type and its minutes. A task repeated three times becomes a written playbook.
- 03
Ten runs makes an agent workflow
A playbook run ten or more times that saves two hours a month becomes an agent workflow behind the approval queue. Each strategist decision to approve, edit or reject a draft is the test of whether the agent is good enough.
- 04
Only then, a feature
Only after that does it become something a customer can switch on. The platform is extracted from real work rather than built on speculation.
Invite-only from 2027
The approval queue
Set the rules. Watch where the work lands.
Each tool has a policy: auto-approve, always ask, or ask above a threshold. Change one, or move the cap, and nine sample actions re-sort in front of you.
Product preview · illustrative actions
$500 a day
One cap for every tool that changes spend. The marks are the sample budget moves.
Runs 3, Asks you 5, Refused 1
Runs3
No question asked, and it still lands paused or as a draft.
Pause ‘Autumn carousel 3’. No purchases in 14 days.
Spends $35 a day
Under your threshold of $100 a day. The ad is paused, and the pause can be reverted.
Add four headline variants to the Retargeting ad set.
Ad set budget $120 a day
Auto-approved. Lands as a paused ad.
Add two video cut-downs to the Prospecting ad set.
Ad set budget $640 a day
Auto-approved. Lands as a paused ad.
Asks you5
Waits for a named person. An approval can wait for days without losing state.
Pause ‘Brand film, 15 seconds’. Frequency is past the limit you set.
Spends $310 a day
Above your threshold of $100 a day.
Move $400 a day from Campaign A to Campaign B for seven days.
Moves $400 a day
Inside your $500 cap, and spend changes always ask.
Raise Prospecting from $400 to $520 a day.
Adds $120 a day
Inside your $500 cap, and spend changes always ask.
Draft a three-email win-back flow for lapsed customers.
Segment of 2,400 people
This tool always asks.
Draft a price-drop email to every subscriber.
Segment of 48,000 people
This tool always asks.
Refused1
Never runs, and is never queued for later.
Move $1,200 a day from Brand search to Shopping.
Moves $1,200 a day
Over your $500 daily cap. Refused, not queued.
Holds whatever you set
- Every write lands paused or as a draft.
- Nothing over the cap runs. It is refused, not queued.
- Spend changes are never unattended in the first version.
The change log
Every change, who approved it, and a way back.
Revert any entry. The original stays where it is, and the revert is written above it as a change of its own.
The rules
Rules the agents run under.
They apply to every agent, on every account, whatever policy you set.
- 01
No number appears in a report unless it came from an API response.
Every figure in the weekly report traces back to the response it came from.
- 02
No claim of cause without an experiment.
An agent may say what changed. It may say why only when a test was run.
- 03
No spending authority without a cap.
Caps are hard, and they are enforced at the platform level, not only in Troy’s code.
- 04
On Meta, agents batch changes weekly instead of optimising daily.
Frequent edits reset an ad’s learning phase, so an agent that tinkers every day makes the account worse. Meta, about the learning phase
What agents will not do
- Change a budget on their own. Every spend change is proposed, approved by you, and capped.
- Change spend unattended. Not in the first version, at any size.
- Send an email or a text. Agents only draft. A send needs your approval and your confirmation that the list has consented, Troy never imports lists, and agents write no text messages to begin with.
- Make photoreal AI people or voices, or talk to your customers as a chat agent.
- Hold your media money. You pay the ad platforms directly.
- Replace your CRM, your email service or your social calendar. Troy connects to your tools.
On CRMArena-Pro, a benchmark of agents on business tasks published in May 2025, structured workflows exceed 83% success while multi-turn, open-ended work sits near 35%. That gap defines the product. Agents draft, build and propose. A person decides.
CRMArena-Pro benchmark
Questions
When, how much, and who says yes.
Invite-only from 2027From $299 a month per workspace · indicative
By invitation from 2027. Until then the same workflows are run by Troy’s strategists for clients of Troy’s team, which is available now. A free Troy Score is the way in to either.
From $299 a month per workspace, with unlimited seats. The price is indicative until agent costs are measured. A strategist on call is about $1,500 a month on top. There is no percentage of media spend and no hourly billing, and you pay the ad platforms directly.
Because Troy’s cost per agent run has not been measured, and we will not guess at it. No credit or per-action price is published until the cost has been measured on at least 500 real runs. Until then the workspace is the fee and usage passes through near cost.
A named person on your side. With Agents, you approve everything that ships or spends, and each tool asks as often as the policy you set for it says. Every approval is recorded with who gave it, and a request can wait for days without losing state.
Every change is written to a permanent change log and can be reverted. A revert restores the setting. It cannot return money that was already spent, which is why anything that changes spend asks first and stays inside a hard cap.