Product
Meet Axia — your company's force multiplier.
An operating model for AI employees — not a chat window.
Not a prompt and a cron job. A persistent record for every AI employee — live, auditable, and yours to tune.
- One record: role, mission, authority, systems, KPIs, budget, reports
- Every task is a work object — owner, state, policy, evidence
- The Fabric shows it live. Widen or tighten authority anytime.
The AI employee record
Everything you expect from an employee, on one record.
Not a prompt and a cron job. One persistent record per AI employee.
- Job description, reporting line, authority per action
- Systems it may touch, schedule, triggers that wake it
- KPIs, budget, the reports it owes
- Same shape for every role — GTM, recruiting, marketing, IT, concierge
- Role
- GTM — pipeline creation
- Mission
- Create qualified pipeline for the mid-market healthcare segment
- Department · manager
- Revenue · reports to VP Sales (human)
- Responsibilities
- Research · outreach · follow-up · reply handling · booking · CRM hygiene
- Authority
- first touch: approval · follow-up: auto · pricing: escalate · schedule: auto
- Systems
- Salesforce · Outlook · M365 Calendar · ZoomInfo
- Schedule · triggers
- 07:00 daily list build · on new lead · on inbound reply · on missed follow-up
- KPIs
- pipeline created · meetings booked · cost per meeting · autonomous share
- Budget
- $1,200 / month · alerts at 80%
- Reports
- daily at 17:30 · weekly Monday 08:00 · to VP Sales
Work objects
Every action is a thing you can point at.
A work object is the unit of everything.
- One research task. One drafted message. One booking.
- Carries its inputs, its reasoning, the policy that governs it.
- Reasoned. Governed. Auditable — not merely observable.
- proposed
- awaiting approval
- approved
- executing
- done
- verified
- 09:41:09outreachdrafted from brief wo-88213 · tone: warm-direct
- 09:41:09policyfirst-touch → requires approval (role: rep)
- 09:43:52human:dmapproved · edited line 2 · edit saved as pattern
- 09:43:53outreachsent via gmail:rep-01 · msg-id 7f3a…
- 09:43:54adapter:crmactivity logged · contact 00Q…
- Owner
- Outreach worker → rep-01 (approver)
- Inputs
- Account brief wo-88213 · CRM contact · sequence template v3
- Policy
- first-touch (approval) · send window 08:00–18:00 local · max 2 follow-ups
- Outputs
- Message · CRM activity · follow-up wo-88220 (scheduled)
- Cost
- $0.07 · 4.1s compute
AI workers
Four workers cover the outreach loop today — and run it again.
Each worker is narrow on purpose.
- One job, known inputs, known outputs — a clear place for policy to gate
- Research hands to Outreach, Outreach to Reply, Reply to Booking, then the loop starts over
- More workers ship as more of the loop earns automation
Research
Builds and enriches target lists, shortlists candidates against role criteria, gathers the context the next step needs.
- inputs
- CRM/ATS records · public signals · approved data sources
- outputs
- Enriched records · shortlists · account briefs
- gate
- List changes (configurable)
Outreach
Drafts personalised first touches and sequences in your voice, sends through your reps' or recruiters' own inboxes.
- inputs
- Briefs · tone & criteria config · templates you approve
- outputs
- Sequences · sent messages · CRM/ATS activity
- gate
- First touch · re-engagement
Booking
Proposes slots, confirms, reschedules, and updates the calendar and the CRM/ATS stage — then the loop starts again.
- inputs
- Calendars · reply intent · scheduling rules
- outputs
- Confirmed meetings · invites · stage updates
- gate
- Usually auto-approved
Reply
Classifies inbound, answers what policy allows, escalates the rest to a person with a suggested response.
- inputs
- Inbound email/SMS · thread history · policy
- outputs
- Replies · escalations · sentiment & intent labels
- gate
- Pricing · compensation · objections you name
Autonomy policy
Policy decides. You set the dial. Workers execute.
You write policy in plain language.
- Which actions run alone, which wait, for whom, under what conditions
- Gated actions land in one queue with reasoning and evidence attached
- Autonomous actions are logged with the same evidence
- Widen the policy as trust builds — you decide when
require approval when action is first_touch or pricing or compensation
auto-approve when action is schedule and both calendars free
never send outside 08:00–18:00 recipient local
- GTMOutreachFirst-touch policySend first-touch sequence → Meridian Logistics · VP Operations
Matches ICP (3PL, 200–800 staff). Two ops roles posted this month; CRM shows no prior contact.
CRM: no activity 180dCareers page: 2 ops openingsEnrichment: HQ Denver, 410 staff - RecruitingOutreachRe-engagement policyMessage shortlisted candidate → Senior Controller · candidate #4471
92% criteria match. Open to relocation per ATS notes. Last contact 14 months ago — re-engage as warm.
ATS: prior stage = final interviewSkills: 6/6 must-havesAvailability: 30-day notice - GTMReplyPricing policyOffer pricing range in reply → Harbor Health · Director of Talent
Prospect asked for a ballpark. Policy requires approval before any pricing leaves the building.
Inbound: 'what does this cost?'Deal stage: discoveryProposed: standard tier range
Mission Control
Two views of the same ledger.
Computed from the same work objects — so the numbers never disagree.
- Executives get outcomes, cost per outcome, and what needs them
- Operators get the queue, the AI employees and the full trail
- Both roll up across companies for multi-entity operators
- 09:41:07researchenriched account acme-logistics · 3 sources · 1.2swo-88213
- 09:41:09outreachdrafted seq/first-touch v3 → queue (policy: first-touch)wo-88214
- 09:43:52humanapproved wo-88214 · edited line 2 · pattern savedusr-dm
- 09:43:53outreachsent via gmail:rep-01 · crm activity loggedwo-88214
Capabilities
Three things a platform has to do before it can be trusted with a workflow.
Three things, before a workflow moves.
- Execute the work
- Give you control over how much runs alone
- Show you exactly what happened
Execute
The work itself, done by AI workers as discrete, traceable work objects.
Accounts, contacts, candidates — from your systems and approved sources.
Personalised sequences in your voice, against your criteria.
Classify, respond, escalate — the thread never goes cold.
Propose, confirm, reschedule. Calendar and CRM/ATS updated.
Control
How much runs on its own, where, and when — configured per role, and changed when you’re ready.
Work routed across workers, systems and people by policy.
Gate the actions you name; let the rest run. Reasoning and evidence attached either way.
Escalate to a person, hand off between workers, retry safely.
Adapters with least-privilege scopes. Two-way sync, no migration.
See
If you can't see what it did, you can't trust what it does next.
What each worker is doing right now, and what it did last week.
Meetings, response time, cost per outcome — per worker, per workflow.
Executive view and operator view on the same numbers.
Every action, input and decision — timestamped, exportable.
Walk through it on your own workflow.
Bring a real role or a real segment. We'll run the loop end to end and show you the ledger.