BRBrookeCustomer experience
Resolves support tickets, handles refunds and delivery enquiries, and keeps customers informed around the clock.
GorgiasGmailShopifySlack
Agents
No two agent workforces look the same.
We build the agents your business actually needs. Some roles come up in almost every operation, so we start from those. The ones that matter most are usually the ones only you have.
Want the list instead of the story? Every capability, grouped →RoleWhat it coversOrigin
CXTickets, refunds, delivery enquiriesCommon
MERCHStock cover, reorder, product performanceCommon
MKTGCampaigns, reporting, paid media briefsCommon
TRADEDaily summaries, targets, commercial riskCommon
YOURSThe role only your business hasBuilt with you
Start from a common role · or have one built and trained for work nobody ships out of the box · every agent lands on the same permissions, approvals and audit · start from a common role · or have one built for you ·
01 · Common roles
Where most businesses start.
These are the roles we build most often. Nearly every operation has work of this shape, which is why we begin there. What you actually deploy gets shaped around your systems, your policies and your way of working, and it won’t look identical to anyone else's.
LALachieMerchandising
Watches stock levels, flags coverage risks, drafts reorder recommendations and monitors product performance daily.
ShopifyCin7GA4
JEJessMarketing
Drafts campaign copy, reports on performance, manages content calendars and briefs paid media with data backed recommendations.
KlaviyoMeta AdsGoogle Ads
RIRikoData and reporting
Generates trading reports, pulls channel performance, flags anomalies and delivers intelligence before the team asks for it.
ShopifyStripeSlack
MEMelTrading operations
Runs daily trading summaries, monitors targets, escalates commercial risks and briefs leadership before the morning standup.
ShopifyXeroSlack
SCScoutLearning and memory
Watches what works, records learnings across the workforce and surfaces improvements that make every other agent sharper over time.
All systemsMemory
02 · Three ways to get one
Take one off the shelf, have us build it, or build it yourself
Most businesses use all three over time. Whichever path you take, the agent lands on the same permissions, approvals and audit.
Path one
Out of the box
Deploy one of our agents this week. Connect it to your systems, set its permission stances, and let it start narrow.
Path two
Custom, built by us
Defined with you during onboarding, or whenever the need comes up, along with any connectors it needs that don’t exist yet. We build it, then train it over weeks the way you would train a new hire. Estimated per build.
Path three
Built by your team
Your own people build agents and skills inside the platform, within guardrails you set. No engineering required.
03 · Built with you
Shaped to your business, in weeks not months
The platform is the product. Onboarding stands up your instance, connects it to your stack, and defines the custom agents and any new connectors you’ll need. Each build is then estimated separately.
StepWhat happensWhen
01Tell us the workflow. A thirty minute call. We sketch the handbook, the tools and the permission stance. You see the shape before you commit.Day 1
02We build it, wired to your stack. Custom connectors to internal tools, ERPs, warehouses and partner APIs. The handbook tuned to your tone and process.Week 1
03We train it, like a new hire. Weeks of shadowing and correction, with permissions widening as trust builds. This is where a generated agent becomes an agent.Weeks 2 to 4
04Live, and yours. Your agents, your memory, your IP, exportable at any time.Ongoing
Anyone can generate an agent in an afternoon.That’s a graduate on their first morning.
Asking an AI to write you an agent gets you day zero. The weeks after that are what turn it into something you would trust with a customer, and almost nobody does them. That’s the part we do.
04 · Always on
They work when nobody is watching, and across days.
An agent can run on a schedule, react to an event, and schedule itself. That last one is what lets work span days rather than dying with the session it started in, and it’s the reason a workforce is different from an assistant.
On a schedule
The performance brief at 07:00 every weekday. The weekly pack before the Monday meeting. Set once, runs forever.
On an event
A ticket arrives, a stock level drops, an order fails. The agent responds without being asked.
Its own follow ups
An agent can schedule itself. In five minutes, tomorrow morning, or every Monday. Work that spans days doesn’t fall through the gap between sessions, because the agent is what carries it, not a conversation.
With approvals
Anything consequential still pauses and asks a person, in the channel your team already uses.
TimeAn agent across three daysState
MONDeploy is queued behind the inventory sync. I will check back once it finishes, and again first thing tomorrow if anything looks off.Self scheduled
MONDeploy succeeded. Order sync running clean, no anomalies in the first hundred events.Checked
TUEOvernight run clean. Nothing to escalate.Checked
FRIFiled the follow up it set for itself on Monday, into the weekly report.Closed
EVERY
Morning
since setup
since setup
00
Times anyone
had to be there
had to be there
05 · Memory in motion
They consolidate while you sleep
Every night an agent runs a quiet review pass over what it learned during the day. Every morning it comes back sharper than the night before, without anyone retraining a model.
Day · active
Working memory fills
Sessions log facts, decisions, open threads, customer signal and follow ups. Everything captured, nothing yet curated.
Night · consolidation
Reviews, refines, organises
A scheduled pass re-reads the day's notes, promotes what’s durable, retires what’s stale and files its own follow ups for tomorrow.
Morning · sharper
Picks up where it left off
A new session starts with refreshed knowledge. The same questions get better answers than they did the day before.
Over time
Measurably more useful
An agent six months into your business isn’t the agent you deployed, and none of that depends on the model underneath changing.
Get started
Which role would your business hire first?
Tell us the workflow that’s eating your best people's week and we’ll show you what an agent for it looks like.