/ ai booking engine · 2026
Bookings that read between the lines.
Enquiries arrive over email, WhatsApp, web forms and voice. Our model parses intent, picks the best slot under your real-world constraints, and confirms — with a human in the loop only when the model isn't sure.
auto-handled
73%
avg first-reply
14s
languages
4
/inbox · enquiry → confirmed
handled in 14sFrom: ops@coldchain-ltd.co.ukEN
“Reefer at the Manchester DC has been flagging temps. Can someone come out before Thursday? Vaccine consignment due Friday morning.”
ai analysis
intent: service-call
priority: high
site: Manchester DC
deadline: Thu 17:00
asset: REEFER-12
confidence: 94%
→ Auto-replyCONFIRMED
“Booked Patel J. — Wed 14 May, 10:00–12:00. Replacement compressor pre-loaded on van. Reply STOP to escalate.”
/ playbook
From inbox to confirmed in five steps.
step 01
Detect
New message arrives. Channel-specific webhook fires.
step 02
Extract
LLM extracts intent, entities, deadlines, location, asset references.
step 03
Schedule
Constraint solver picks slot — engineer skills, parts stock, travel time, SLAs.
step 04
Confirm
Auto-reply in the customer's language. Calendar invites sent. Stock reserved.
step 05
Escalate
Below-threshold confidence → human takes over. Every action explainable & reversible.
What the scheduler considers
engineer skill
match job type → certified engineer
travel
live traffic, fuel, hours-of-service
parts stock
van, regional depot, supplier lead time
sla / deadline
customer contract, severity
utilisation
target 80–90%, no churn-burn
customer preference
site access windows, contacts
weather
outdoor work risk-adjusted
opt. score
human-readable per slot
/ next step
Send us a week of your inbox.
We'll show you exactly which enquiries the Booking Engine would have auto-handled, which it would have escalated, and the slots it would have picked.