What we automate

Start with the work your team repeats.

These are examples, not fixed products. We begin with the process you already have, then decide what should be automated, what should stay manual and whether AI is useful at all.

Four common patterns

Where repetitive work usually hides

Each example can connect to the tools your team already uses, with a person kept in control where judgment matters.

01

Enquiries and inboxes

Every request understood, organised and sent to the right place.

Turn forms, shared inboxes and incoming requests into structured work without asking someone to read, copy and route every item by hand.

  1. Form or email
  2. Extract and classify
  3. Route
  4. Update system
  5. Prepare next step

Examples

  • Lead intake and qualification
  • Shared inbox and request triage
  • Priority and ownership assignment
  • Acknowledgement and response drafts

Unclear requests and consequential replies stay with a person.

02

Documents and data

Useful information extracted once and ready for review.

Read invoices, forms, CVs and contracts, validate the fields you need and move the result into the right business system.

  1. Document arrives
  2. Classify and extract
  3. Validate
  4. Update system
  5. Store and notify

Examples

  • Invoice and form processing
  • Selected contract fields and summaries
  • Document classification and filing
  • Structured data validation

Exceptions remain visible beside the source document; automation does not approve payments or legal decisions.

03

Sales and follow-up

Cleaner records and a consistent next step after every interaction.

Keep CRM records, meeting outcomes, tasks and follow-up drafts moving without another round of administration.

  1. Activity or transcript
  2. Summarise
  3. Create tasks
  4. Update CRM
  5. Draft follow-up

Examples

  • Meeting summaries and action items
  • CRM record creation and updates
  • Opportunity briefs and reminders
  • Customer follow-up drafts

An owner confirms decisions and customer commitments before anything is sent.

04

Operational workflows

Scattered recurring work connected into one reliable process.

Bring together onboarding, reporting, synchronisation and other repeated processes using rules, APIs and AI only where each one helps.

  1. Your trigger
  2. Rules, APIs or AI
  3. Human checkpoint
  4. Business system
  5. Clear outcome

Examples

  • Customer and supplier onboarding
  • Recurring reports and notifications
  • Data synchronisation between tools
  • Custom internal workflow automation

Scope, exception handling and ownership are agreed before the build starts.

Illustrative example

Enquiry intake to CRM — a worked example

This is an illustrative example, not a client result. It shows how a typical enquiry-to-CRM workflow might be automated, what stays manual and where the measurable improvements appear.

The problem

A shared inbox receives enquiries from several channels. Someone reads each one, decides what it is, copies information into the CRM, assigns an owner and drafts a response. Duplicate records are common, routing is inconsistent and responses take hours.

What changes

An automated workflow handles intake, validation, classification, routing, CRM update and response drafting. A person still approves outgoing replies and handles exceptions that the rules cannot resolve.

  1. 01

    Intake

    Enquiries arrive from forms, email or integrations and are normalised into a consistent structure. Deterministic rules.

  2. 02

    Validation

    Required fields, formats, duplicates and known spam patterns are checked automatically. Deterministic rules.

  3. 03

    Classification

    The enquiry type, urgency and topic are identified from the message content. AI classification where free-text makes rules impractical.

  4. 04

    Routing

    The enquiry is assigned to the right owner or team based on type, region and workload. Deterministic rules.

  5. 05

    CRM update

    Contact, company and opportunity records are created or updated without manual data entry. Direct API integration.

  6. 06

    Response drafting

    A reply is drafted using the enquiry context and standard templates. AI drafting with template guardrails.

  7. 07

    Human approval

    The assigned owner reviews the draft, edits if needed and sends. Person decides.

  8. 08

    Exception handling

    Unclassifiable, ambiguous or high-value enquiries are flagged for manual review rather than processed automatically. Person decides.

Measurement

Measurable outcomes to track

These are the metrics you would measure after deployment — not invented results.

  1. 01

    Handling time — time from enquiry arrival to owner assignment.

  2. 02

    Routing accuracy — percentage of enquiries reaching the correct owner on first assignment.

  3. 03

    Response time — time from arrival to first reply sent.

  4. 04

    Duplicate records — CRM duplicates created per period.

  5. 05

    Exception rate — percentage of enquiries requiring manual intervention.

See how projects like this are delivered, or describe your own workflow to start.

A useful constraint

Not everything needs AI.

Sometimes the reliable answer is a rule, a direct integration, a better form or one less step. The goal is not to add a model to every process; it is to remove repetitive work without creating a harder system to operate.

Recognise one of these patterns?

Describe the workflow in your own words. We will start with the process, not a predetermined tool.