Bilingual WhatsApp Support Agent for an E-Commerce Retailer
A self-hosted Arabic-English WhatsApp support agent with per-customer memory, structured handoffs, human takeover, and catalog grounding.
Read the projectCustomer operations
Give every customer request a clear next step.
Connect onboarding, support handoffs and customer updates so work reaches the right owner. Keep people in control of exceptions and sensitive decisions.
- 20–30 minute review
- No preparation needed
- One workflow mapped
Start with a defined scope. Explore the relevant package.Automation Sprint
Organizations we’ve worked with
- Airbnb
- TheAMLab
- Hood
- NAMA
- Klaro
- GoBoon
- Studio Arlington
Where customer response slips
The customer work your team is doing by hand
Every one of these is a handoff a system should own. Today, people are the glue between your customer tools.
The diagnosis
All six are one problem: your team acting as the queue between customer tools that were never connected. Remove that one job, and response times start fixing themselves.
- 01Manual customer onboardingEvery new customer walked through the same steps by hand, from welcome to first value.
- 02Ticket triage and handoffsSupport requests read, categorized, and reassigned manually before anyone can act.
- 03Updates across CRM and support toolsThe same customer status re-keyed into the CRM, the helpdesk, and internal trackers.
- 04Missed follow-upsCheck-ins and next steps that depend on someone remembering to send them.
- 05Status chasingTeam members pinging each other to find out where a customer request actually stands.
- 06Slow response timesCustomers waiting because work sits in a queue no system is moving forward.
Live customer runtime
One customer, zero manual handoffs
Follow one request through the system. It is classified on arrival, routed to the right owner, and every tool it touches is updated in sequence and confirmed — no one re-keys status between systems, and nothing waits in a queue.
The runtime sequence
- 1Request classified by intent and priority on arrival
- 2Routed to the right owner — human or system
- 3CRM and helpdesk updated together, then verified
- 4Follow-ups scheduled and confirmed, nothing forgotten
- 5Exceptions paused into a review queue with full context
What stays human
Not everything should be automatic. An angry customer, a judgment call, a case the rules cannot answer — the run pauses into a review queue with full context, a person decides, and the workflow resumes. Empathy stays human; the routing does not.
- Frustrated customers and judgment calls → routed to a person
- The decision is recorded for audit and agreed rule reviews
- The approved workflow resumes, with unresolved steps tracked
ILLUSTRATIVE WALKTHROUGH
A customer needs a person, not another automated reply
- 01
Recognise the request
A refund request arrives with the customer and order context.
- 02
Assign an owner
The policy exception goes to an authorised teammate, with a visible status and next action.
- 03
Close the loop
After review, the customer receives the approved response and the outcome is recorded.
Check that escalation, ownership and customer updates still work when a connected service is unavailable.
PUBLISHED PROJECTSee the work
See the work
behind the approach.
A published example of related work. Your systems, scope and acceptance checks are agreed separately.
Pick your workflow
See the handoffs disappear
Pick the one that hurts most — today on the left, as a production system on the right. A representative pattern; your map will be exact.
Today — by hand
- 1Request lands in a shared inboxsits in queue
- 2Someone reads it and guesses the category10 min each
- 3Ownership decided in a Slack threadping-pong
- 4Status re-keyed into CRM and helpdeskkeyed twice
- 5Customer emails again asking for an updateinbox limbo
Hours to first touch · 4 manual steps
With the system
- Classified by intent and priority on arrivalAI + rules
- Routed to the right owner instantlyauto
- CRM and helpdesk updated togetherauto
- Customer gets a status without askingauto
- Escalations flagged to a person with contextverified
Minutes to first touch · 0 manual handoffs
Works with the tools you already run
No rip-and-replace — the system wires into your stack.+ any tool with an API
- Zendesk
- Intercom
- HubSpot
- Salesforce
- Slack
- Microsoft Teams
- Gmail
- Stripe
- Calendly
- Notion
- Zapier
- Make
How it works
The capabilities that deliver it
Customer Operations Automation is built from three engineering capabilities working together.
Primary
Automation
Trigger, routing, and execution logic that moves onboarding, tickets, and follow-ups end-to-end without chasing.
TriageRoutingFollow-upsEscalation
Explore capabilitySupportingAI Systems
Triage, classification, and next-action selection for the customer decisions that stay manual today.
IntentPriorityNext action
Explore capabilitySupportingIntegration
The connective layer that keeps your CRM and support tools in sync instead of your team.
CRM ↔ helpdesk syncReliable APIsShared state
Explore capabilityDo the math
How much of support is manual routing?
Slide to your reality. This is only arithmetic — the review tells you which of those minutes a system can actually own.
10500
130
Handling overhead
16hrs/week
≈ 736 hours a year spent triaging, re-keying status, and chasing updates instead of the conversations that need judgment.
Planning estimate based on the inputs above, not a forecast of savings.
Assumes ~46 working weeks a year. We will tell you which of those hours a system can own — and which it should not.
How we deliver
From manual handoffs to running system
Four steps, and each one hands you a concrete artifact — not a status update.
- 01 · Discover
Map the customer journey
We trace how onboarding, tickets, and updates move today — who touches them, where they stall, and which handoffs are manual.
You get
A map of how customer work really moves
Done when
The worst customer workflow is chosen, its constraint named
- 02 · Design
Design the system
We define triage rules, routing, follow-up logic, and the integrations that keep every customer tool in sync.
You get
The system design — triage, routes, follow-ups
Done when
Every handoff has an owner — human or system
- 03 · Build
Build and connect
We build the automation and wire it into your CRM and support stack with production safeguards.
You get
A running automation in your support stack
Done when
The first real ticket flows end-to-end
- 04 · Operate
Operate and verify
We ship with monitoring and verification so every handoff and follow-up is confirmed and nothing slips through.
You get
Monitoring with every follow-up confirmed
Done when
Agreed routing, follow-up and escalation checks pass
Is this the right starting point?
Where Customer Operations Automation fits
We would rather map the wrong workflow out than sell you a system you do not need.
Good fit when
Start here- The same customer work repeats on a predictable trigger — onboarding, triage, follow-ups
- People manually route, categorize, or reassign inbound requests
- Response speed matters and work stalls in a queue no system is moving
- You can point to one customer workflow that breaks most often
Not a fit when
Not yet- Every case needs human judgment with no repeatable rules to encode
- No system of record exists for customers to build against yet
- You need a one-off script, not a monitored production system
- The real constraint is headcount or strategy, not coordination
Not sure which side you are on? Most teams aren’t — until the workflow is mapped.
YOUR FIRST ENGAGEMENT
Start where the customer journey stalls.
Choose one journey: a new customer, an incoming request or a case waiting on another team. Make the next action and owner visible.
Start with a conversation
Bring an example of a delayed request and the tools your team uses to resolve it.
An example helps; a finished specification is not required.
BEFORE IMPLEMENTATION
Know what you are agreeing to.
- A request-to-resolution map with ownership rules
- An agreed escalation and customer-update flow
- A scoped first release with completion checks
We agree the scope, delivery approach and price before implementation. Where discovery is needed, its scope and fee are agreed first.
Automation routes and updates the work. Your team handles cases that need judgement, approval or a personal response.
BEFORE YOU GET STARTED
Customer Operations Automation questions
Scope, delivery and what happens next. The details that help you decide.
A RELEVANT WAY TO WORK WITH US
Automation Sprint
Turn a defined workflow into an agreed automation scope. Review the package, deliverables and eligibility before choosing your next step.
Explore Automation SprintAsk about your projectWhich customer journey should we start with?
Choose a repeatable journey such as onboarding, ticket routing or a customer status update. We map ownership and the desired outcome before deciding which actions can run automatically.
When does a person take over?
You agree the escalation rules, authorised owners and approval boundaries. Sensitive decisions and policy exceptions can remain with your team while the system provides context and tracks the next step.
Can you connect our helpdesk and CRM?
We assess the available APIs, identifiers and permissions in your existing tools. The scope identifies which records move, how conflicts are handled and what happens when a connection is unavailable.
What happens before implementation starts?
We review the intended outcome, existing systems and dependencies, then agree deliverables, acceptance checks, responsibilities and price. If discovery is needed first, its scope and fee are agreed separately before that work begins.
How long will the work take?
The delivery plan depends on scope, access, integrations and review dependencies. We agree milestones after that assessment rather than promise the same timeline for every system.
What access or preparation do you need?
Bring an example of the workflow or user journey and an overview of the systems involved. We agree the access needed for the work. Do not send passwords or private customer data through the enquiry form.
What happens after handover?
Handover, documentation, defect coverage and any ongoing support follow the chosen offer and written scope. New features, additional integrations and continuing maintenance are not assumed to be included.
Start here
Start with the customer workflow that improve
We will map one manual customer workflow, find the real constraint, and show you the first system to build.
20–30 minutes · No preparation needed · Not ready for a call? Send a note instead