Use cases
AI use cases, with the systems and the human checkpoint named.
Each use case is one task we automate end to end: what the system takes on, which of your tools it reads and writes, where a person stays in charge, and how it rolls out week by week. Not sure where to start? Score a task of your own with the four-question rubric.
Score a first use case
Pick one task and answer four questions.
Good candidate for a narrow pilot
Worth automating; cover the common cases first and hand the rest to a person.
- Oversight:
- People take the low-confidence cases, with full context.
- Engagement shape:
- A narrow pilot on one slice of the task, usually Starter, widened once the weekly review holds up. See pricing
Closest use cases
Sample data: guidance, not a promise of results. Your answers stay in this browser.
Answer
Questions with a written answer: customer support, internal knowledge, order status, multilingual support. Every answer cites its source; a person takes the rest.Qualify and route
Leads and calls sorted by rules you write and handed to the right person with a summary.Schedule and remind
Bookings, reminders and follow-ups that keep the calendar and CRM right, with opt-out built in.Extract and sync
Data pulled from documents and kept consistent between systems, checked by rules before it is written.Research and create
Research, outreach drafts and marketing video, with a person approving every send and every cut.
Tasks we automate
Browse by service, team and industry
- AI chatbots
Customer support automation
Deflect repetitive tickets with cited answers and hand the rest over with context.
- Systems
- Helpdesk, help centre, web chat and WhatsApp
- Human checkpoint
- Agents own refunds, disputes and low-confidence chats
- Workflow automation
Lead qualification and routing
Qualify in the first reply, score by your rules and route to the right owner.
- Systems
- Forms, chat, WhatsApp and your CRM
- Human checkpoint
- Sales owns the rules and reads every sales-ready lead
- AI chatbots
Appointment scheduling
Booking, reminders and one-tap rescheduling against your real calendar.
- Systems
- Your calendar, WhatsApp, SMS and the phone
- Human checkpoint
- The front desk handles exceptions and late cancellations
- Voice agents
Inbound call answering
An AI receptionist that answers, books, takes messages and transfers urgent calls.
- Systems
- Your phone line, calendar and job system
- Human checkpoint
- Urgent callers reach the on-call person with a summary
- Workflow automation
Document data extraction
Extract, validate by rules, hold doubtful fields and post the rest.
- Systems
- Mailbox, review queue and your accounting system
- Human checkpoint
- Finance clears held fields and every bank-detail change
- Custom AI
Internal knowledge assistant
Cited answers from SOPs and policies, limited to what each person may see.
- Systems
- Drives, wiki, single sign-on and team chat
- Human checkpoint
- Collection owners act on the weekly gap report
- Workflow automation
CRM data sync and enrichment
Sync between tools with conflict rules and a nightly reconciliation.
- Systems
- Your CRM, forms, billing and spreadsheets
- Human checkpoint
- Operations settles the conflicts the rules cannot
- AI chatbots
Order status and returns
Order lookups behind an identity check and policy-grounded returns.
- Systems
- Your store platform, carrier tracking and helpdesk
- Human checkpoint
- Staff approve every refund
- AI agents
Sales research assistant
Agents that research accounts and draft outreach for a rep.
- Systems
- Public company sources, your CRM and email drafts
- Human checkpoint
- Nothing is sent until a rep approves it
- Workflow automation
Follow-up and nurture automation
Consent-aware sequences that honour an opt-out at every step.
- Systems
- Your CRM, email and WhatsApp
- Human checkpoint
- Marketing approves every sequence and its wording
- AI video
Marketing video production
Brief to script to storyboard to render, with a brand review step.
- Systems
- Brand kit, asset library and the render pipeline
- Human checkpoint
- A brand reviewer signs off every cut
- AI chatbots
Multilingual customer support
Per-language answers, with a human speaker when needed.
- Systems
- Chat, WhatsApp and help content per language
- Human checkpoint
- Staff who speak the language take the rest
Method
How we pick a first use case
Before we propose a build, we score the task on the same four questions the scorer asks. The score is not about whether AI is clever enough; it is about whether automating this task first is worth the effort and safe to run.
| Question | What we look at | A strong signal | A warning sign |
|---|---|---|---|
| Volume | How often the task happens across your team | It happens many times a day, every day | A handful of cases a month |
| Repeatability | How alike the cases are | Same inputs, same steps, a few known exceptions | Every case needs fresh judgement |
| Data readiness | Whether the inputs are written down and reachable | Current documents, or a system with an API | It lives in one person’s head |
| Risk | What a single mistake would cost | Internal and easy to correct | Money moves, or health or legal exposure |
Why risk changes the shape, not the score
A risky task can still be a good first project: document extraction moves money, yet works well when a person approves what is held. Risk sets how much oversight the build carries, from spot checks to an approval on every step.
What usually makes a good first project
- One team owns the task and can say what a good result looks like.
- The task touches one or two systems, not your whole stack.
- You can measure today’s baseline, such as response time, keying time or missed calls, before anything changes.
- There is an obvious place for a person to step in when the system is unsure.
When we suggest not using AI
If the task is rare, purely rule-based or blocked by missing documentation, a checklist, a form or a plain automation often does more for less. We say so on the call, and so does the scorer.
The scorer is a planning aid: guidance for a first conversation, not a promise of results. Your answers never leave your browser.
FAQ
Frequently asked questions
Are the use cases on these pages client projects?
No. Each page explains how we build one task, with traces, chats and tables marked Sample data. Where a worked example exists it is labelled Concept build. None of them is presented as a client result.
Can one project cover more than one use case?
Yes, and some pairs belong together, such as appointment scheduling with inbound call answering. We still launch one task first and add the second once the first is measured and stable.
Our task is not listed here. Can you still help?
Probably. These pages cover the tasks we are asked about most, not everything we build. Score it with the rubric and bring it to a strategy call; if it is a poor fit for AI, we will say what would work better.
How long does a first use case take to go live?
Most rollouts on these pages run over four to six weeks, from discovery to a soft launch, with a shadow or parallel run before anything replaces how your team works. Each page shows its week-by-week plan.
Do we have to replace our existing software?
No. Every use case connects to the tools you already run through their APIs, and those tools stay the system of record. Your team keeps working in its helpdesk, CRM, calendar or accounting system.
What happens to the answers I give the scorer?
Nothing leaves your browser. The scorer runs entirely on the page and does not send, save or track your answers.
Next step
Have a workflow in mind? Let’s scope it together.
Book a free 30-minute strategy call. Bring one process you want to automate and we will talk through scope, integrations and a sensible first step.
Free 30-minute call · no commitment