SPEC 03 · AI AUTOMATIONExample · customer message
AI in your day-to-day business.
We automate customer experience and build AI into the way your team already works: answering customers, moving work between tools and reading the paperwork.
- TriggerA customer writes inEmail, web chat or WhatsAppStatus: Done
- AIReads it and finds the orderLooks up the customer in your CRMStatus: Done
- RuleRefund under your limit?The limit is yours to set
- If yes: Refund sent, customer toldStatus: Done
- If no: Passed to a person with a drafted replyStatus: Waiting
By team
Pick a team. See what we'd automate.
Customer service
What we automate
- Answers to common questions, from your own help pages
- Order, booking and delivery status on request
- Sorting and routing every incoming message
- Drafted replies a person checks and sends
What changes
- Customers get answers at any hour
- Your team handles the cases that need a person
- Every conversation logged in one place
Works withEmailWeb chatWhatsAppYour CRMHelp desk
Sales
What we automate
- Lead details checked and filled in
- Follow-up emails drafted after each call
- Quotes prepared from your price list
- Weekly pipeline summary
What changes
- No lead waits for a reply
- Sellers spend time selling, not typing
- Managers see the week at a glance
Works withYour CRMEmailCalendarSpreadsheets
Operations
What we automate
- Hand-offs between your tools without copy-paste
- Approvals and routing that run themselves
- Reports that build and send on schedule
- Alerts when something needs a person
What changes
- Fewer steps lost between teams
- Clear fallbacks when a step fails
- You can see and change every rule
Works withAPIsEmail & chatSpreadsheetsYour systems
Finance
What we automate
- Invoices and receipts read into structured data
- Matching payments to invoices
- Expense checks against your policy
- Month-end summaries for review
What changes
- Less manual entry
- People review low-confidence results
- Exports to the systems you already use
Works withAccounting softwareBank exportsScans & photos
People & HR
What we automate
- An assistant that answers policy questions
- Onboarding checklists that run themselves
- CVs read and sorted against your criteria
- Leave and request routing
What changes
- Staff find answers without waiting
- New hires get the same start every time
- A person makes every hiring decision
Works withHR systemEmailYour documents
What we build
Pick a starting point. See what we'd build.
AI agents & assistants
What we build
- Assistants that answer from your own documents and policies
- Agents that take actions in your systems, with approvals
- Chat for customers on web and in apps
- Helpers for support, sales and HR teams
What matters here
- Answers grounded in your sources, with citations
- A person approves anything that matters
- Every action logged
- Clear rules on where your data goes
LLM APIsRetrievalTool useSSOAudit logs
Workflow automation
What we build
- Hand-offs between your tools without copy-paste
- Approvals and routing that run themselves
- Reports that build and send on schedule
- Alerts when something needs a person
What matters here
- Start with one workflow, measured before and after
- Clear fallbacks when a step fails
- You can see and change every rule
- Runs in your cloud or ours
APIsWebhooksQueuesEmail & chatMonitoring
Document AI
What we build
- Read invoices, forms and contracts into structured data
- Classify and route incoming documents
- Search years of files in plain language
- Summaries and drafts for review
What matters here
- People review low-confidence results
- Sensitive data handled by your rules
- Works with scans and photos, not just PDFs
- Exports to the systems you already use
OCRExtractionSearchRedactionIntegrations
AI inside apps
What we build
- Smart search and recommendations
- Chat and voice inside your web or mobile app
- Writing and summarising features for your users
- Image and document understanding
What matters here
- Fast responses on real devices
- Predictable cost per user
- Safe defaults and content filters
- Tested against real user questions
WebiOS / AndroidLLM APIsOn-device optionsAnalytics
Process
Four steps, no surprises.
- 01
01 Find
We pick one workflow worth automating and see how it runs today.
You get: a written proposal with scope, timeline and price.
- 02
02 Pilot
A working version on real data with a small group of users.
You get: a pilot you can test.
- 03
03 Build
We harden it, connect it to your systems and add controls.
You get: a production system with logs and approvals.
- 04
04 Run
Monitoring, tuning and support as your data changes.
You get: reports, fixes and improvements.
Control
Secure by default, not as an extra.
- Human approvalPeople sign off on actions that matter.
- Audit logsEvery action and answer is recorded.
- Data boundariesYour data stays where you decide.
- MonitoringQuality and cost tracked after launch.