Six roles,
built end to end.
This is not a catalogue. These are simply six roles we have built end to end, shown so you can see what a finished AI employee looks like. Every one ships with a written job description, an authority matrix and the KPIs it is measured on. Every one we build is designed from scratch around your business, and most of what we build is a role that isn't on this page at all.
Maya AI Receptionist
Answers the phone at 3am, on a public holiday, and during your busiest hour.
Calls arrive after hours, during lunch, and while the team is already on another line. Voicemail hands the decision to the caller. This employee holds the switchboard.
- Answers inbound calls, identifies the caller, captures why they rang
- Qualifies against the criteria you set, in your tone of voice
- Books appointments directly into the right calendar
- Routes anything outside its remit to a named human, with context
- Writes every interaction back to your CRM before it hangs up
- Handles overflow when your team is already on the phone
Typically measured on
- Calls answered vs. abandoned
- Average time to answer
- Qualified appointments booked
- Routing accuracy
- Escalation rate
- Cost per handled call
Commonly connects to
RingCentral, Twilio or your phone system; Google or Outlook calendar; HubSpot or your CRM; SMS; and whatever booking tool you already run.
Alex AI Customer Service Employee
Reads the request, finds the answer, resolves it, and knows exactly when not to.
Support volume concentrates: the same questions arrive again and again, wearing different clothes. This employee clears those completely so your people only see the ones that genuinely need a person.
- Monitors the shared inbox, chat queue and ticket backlog
- Looks up the customer, the order, the history, the entitlement
- Resolves routine requests end to end and closes the ticket
- Drafts a reply for review when it isn't certain
- Escalates real exceptions with the research already done
- Refuses to invent a policy it was never given
Typically measured on
- First-contact resolution rate
- Autonomous resolution share
- Median response time
- Escalation correctness
- Customer satisfaction
- Cost per resolved ticket
Commonly connects to
Zendesk, Freshdesk or your helpdesk; Gmail or Outlook; live chat; Shopify or your order system; your knowledge base and returns portal.
Sloane AI Sales Development Employee
Works every inbound enquiry as it arrives, follows up on a schedule, and keeps the CRM record straight.
Reply speed and follow-up persistence slip when a rep's day fills up. This employee works every enquiry the moment it lands and keeps each follow-up sequence running until the prospect replies.
- Picks up inbound enquiries as they land, day or night
- Researches the company and contact before writing
- Qualifies against your criteria and disqualifies honestly
- Runs multi-step follow-up sequences and stops on a reply
- Books meetings onto the right rep's calendar
- Keeps CRM records accurate without being chased
Typically measured on
- Median lead response time
- Leads worked vs. leads received
- Meetings booked and held
- Qualification accuracy
- Pipeline created
- CRM data completeness
Commonly connects to
HubSpot, Salesforce or your CRM; Gmail or Outlook; calendars; Apollo or your enrichment tool; web forms; phone; LinkedIn.
Free, thirty minutes, no obligation.
Rachel AI Accounts Receivable Employee
Chases every overdue invoice, records every promise, and escalates disputes early.
Collections is easy to judge, because the output is cash rather than opinion. It is also work nobody enjoys, which is exactly why it gets skipped when everyone is busy.
- Monitors every overdue invoice, every day, without a reminder
- Prioritises accounts by value, age and likelihood of recovery
- Sends the appropriate reminder at the appropriate escalation level
- Answers routine payment questions and confirms remittance
- Records promises to pay and follows up on the exact date
- Escalates disputes to a human before they harden
Typically measured on
- Overdue balance recovered
- Days sales outstanding
- Collection rate
- Human interventions per 100 accounts
- Promise-to-pay conversion
- Customer complaint rate
Commonly connects to
QuickBooks, Xero or your accounting system; Stripe or your payment processor; your banking feed; email; CRM; Slack or Teams.
Sophia AI Recruiting Employee
Screens every application the day it lands, against a scorecard rather than a mood.
Hiring rarely stalls on the decision. It stalls on the applications nobody has read yet and the good candidate who accepted elsewhere while waiting.
- Sources candidates against a written scorecard
- Screens every inbound application within hours, not weeks
- Runs structured first-round interviews by chat or voice
- Scores consistently and shows its reasoning
- Schedules the shortlist without the email tennis
- Tells you honestly when the pipeline is too thin to shortlist
Typically measured on
- Time from application to first response
- Applications screened per week
- Shortlist quality (hiring manager accept rate)
- Interviews scheduled
- Time to shortlist
- Candidate experience rating
Commonly connects to
Greenhouse, Workable or your ATS; Indeed and the job boards; email; calendars; video interview tooling; your HRIS.
Marcus AI Operations Employee
The repetitive processing work that quietly consumes a department.
Order entry, document processing, data reconciliation, exception handling, the report that eats a morning every Monday. Individually trivial, collectively enormous.
- Ingests documents (PDFs, spreadsheets, emailed forms) and extracts structure
- Processes repetitive transactions across your systems
- Reconciles data between systems that don't talk to each other
- Detects exceptions and routes them to the right person
- Produces recurring reports on schedule, correctly
- Flags the pattern before it becomes a problem
Typically measured on
- Transactions processed
- Processing accuracy
- Exception detection rate
- Cycle time per item
- Human touches per 100 items
- Human hours avoided
Commonly connects to
NetSuite, SAP or the ERP you actually run; email; Excel and Google Sheets; document stores; databases; browser-based legacy portals.
Don't see the employee you need? That's the point.
Most of what we build has no name yet. Freight invoice reconciliation. Warranty claim processing. Permit tracking. Dealer onboarding. Fitment verification. Tender qualification. Compliance monitoring.
The named roles above are simply the ones we have built end to end. Yours may be a role we have never built before. Either way the method is identical: we map the job, design the authority, build it, test it, and put it on probation.
We'll tell you when it can't.
If the work is repetitive, digital, rules-and-judgement based and performed through software, there's a good chance an AI employee can hold meaningful parts of it. When it isn't, we'd rather lose the sale than take your money for something that will disappoint you in month three.
Strong candidates
- High frequency: happens hundreds or thousands of times a month
- Meaningful human time goes into it today
- Enough recurring structure to describe as a process
- The information needed is reachable in a system we can connect to
- Mistakes are recoverable and detectable
- Someone can tell us what "done correctly" looks like
Poor candidates
- Requires physical presence or handling
- Depends on a relationship that genuinely can't be delegated
- Rules change constantly and are written down nowhere
- The data lives somewhere we cannot reach, legally or technically
- A single error is catastrophic and undetectable
- It happens four times a year, so automating it never earns the cost back
A no from us costs you thirty minutes. A yes from someone who bills either way costs considerably more.
Tell us the job.
We'll tell you if we can fill it.
Free, thirty minutes, no obligation.
