AI consulting | Based in Perth, working across Australia
AI implementation that takes repetitive work off your team.
We connect AI to how you handle enquiries, find internal information and report on the business, inside the CRM and systems you already run. Perth is our base, and we work with businesses across Australia.
Free 30-minute conversation. Send a short form and we'll reply to arrange a time.
Who it's for
For teams ready to put AI into real processes
Most teams already use AI chat tools. The value comes from connecting AI to the processes, data and systems the business runs on.
Sound familiar?
- Enquiries wait in the inbox until someone has time to read, sort and reply to them.
- Details from emails, forms and paperwork are typed into your systems by hand.
- Staff keep asking the same few people how things are done.
- Reports are pulled together by hand, so questions about the numbers wait.
- You've experimented with AI tools, but nothing has stuck in day-to-day operations.
- Staff use AI informally and you're not sure what business data is going where.
Where AI helps
What AI can take on, and what changes for your team
Each use is built into the systems you already run. For each use case we agree what AI handles automatically, when it hands over to your team, and which actions need approval first.
- 01
Enquiry handling
AI reads each new enquiry, works out what it's about and how urgent it is, routes it to the right person and drafts a first reply.
What changesEnquiries are sorted as they arrive, and your team starts from a draft instead of a blank reply.
- 02
Automated customer conversations
Answering routine questions, qualifying enquiries, booking appointments and following up past customers by chat, SMS or email.
What changesRoutine questions and bookings don't wait for someone to be free, and anyone who asks for a person is handed over.
- 03
Internal information
Staff ask questions in plain English and get answers from your own procedures, with the source shown and access limited by role.
What changesFewer interruptions for the people who know how things work, and answers staff can check against the source.
- 04
Documents and email
AI reads invoices, job sheets and emails with varied layouts and enters the details into your systems.
What changesLess retyping. Unclear fields, and anything that changes an amount owed, wait for a person.
- 05
Reporting
Plain-English questions and summaries over agreed metrics from your CRM, pipeline or job data.
What changesAnswers about the numbers without waiting for someone to build a report.
Plenty of automation needs no AI. If rules alone do the job, we'll recommend that.
Not sure which of these would help most?
That's what the free growth review is for. We'll look at one process, the systems it touches and a sensible first step.
Request a growth reviewExample workflow
From incoming request to a checked CRM record
AI reads and sorts each request, independent checks test the result, and a person reviews what fails. Only checked or approved items update the CRM or reach a customer.
Synthetic example, not a client workflow
A service business receives job requests by email and web form, many with a PDF work order attached.
- 01Handled by: Automation
Receive
Each request is logged with an ID, and the original is kept unchanged.
- 02Handled by: AI model
Extract and triage
AI pulls out the customer, site, service, date and reference, and sorts the request by type and urgency.
- 03Handled by: Rule checks
Validate
Checks that don't rely on the AI test required fields, service area, the CRM match and duplicates.
- 04Handled by: Your team
Review where needed
A coordinator checks anything that failed or was flagged, and approves the next step.
- 05Handled by: CRM and automation
Update the CRM and act
Only checked or approved items update the CRM and send an acknowledgement.
- 06Handled by: Owner and reports
Monitor
A weekly sample is checked against the originals, and any change is retested first.
Each stage in full: input, output, exceptions and approval rules
- 01
Receive
Each request is logged with an ID, source and time, and the original is kept unchanged.
- Input
- Email, web form or PDF
- Output
- A logged item with a unique ID
- Exception
- Unreadable file: sent straight to review.
- 02
Extract and triage
AI reads the message and attachment, extracts the customer, site, service, date and reference, and classifies the request type and urgency.
- Input
- The logged item
- Output
- Structured fields and a request type
- Exception
- Missing or ambiguous fields are flagged for review. If the AI service is down, items queue for manual handling.
- 03
Validate
Field checks independent of the AI test the result: required fields, service area, a CRM match by email or ABN, and no duplicate job. Review thresholds come from testing against labelled examples, not the model's own confidence.
- Input
- Structured fields
- Output
- A pass, or a list of named failures
- Exception
- Any failed check: sent to review with the reason attached.
- 04
Review where needed
A coordinator compares the original with the extracted fields, corrects or confirms them, and approves the next step.
- Input
- Failed or flagged items
- Output
- Confirmed fields and an approved action
- Exception
- A complaint, or not a job request: passed to the right person.
Approval boundaryItems cross this line only after passing every check or being approved in review.
05Update the CRM and act
The contact and job are created or updated, an owner is assigned, and an approved acknowledgement is sent.
- Input
- A validated or approved item
- Output
- CRM record, task and logged acknowledgement
- Exception
- CRM unavailable: retried without duplicates, then an alert to a named person.
- 06
Monitor
A weekly sample of automatic items is checked against the originals, and any prompt, model or rule change is retested on the labelled examples first.
- Input
- Processed items, corrections and failures
- Output
- A quality report and change log
- Exception
- Results outside the agreed range: the automatic path pauses until it's fixed.
Approval rules in this example
Runs automatically
- Logging, extraction and checks
- Routing and CRM updates for validated requests
- Acknowledgements from approved templates
Needs approval first
- Failed or flagged items
- New customer records
- Non-standard replies
- Changes to dates, scope or amounts
Always handled by a person
- Quotes and pricing
- Complaints and disputes
- Urgent or safety-related requests
Stages, checks and approval rules are agreed for each workflow, and boundaries widen only as checked results support it.
How we work
How we deliver AI implementation
We start with one process, prove it on real work with your team checking the output, then widen it. Once it's live, we keep monitoring and improving it as models and your systems change.
- 01
Discovery
Walk through the process with the people who do it, check data and system access, and agree who owns the result.
- 02
Shortlist and pilot plan
Rank uses by value, effort, running cost and risk, then agree the first pilot's measures and approval rules.
- 03
Pilot
Build one workflow on real data, run it with full review or alongside the current process, and measure it.
- 04
Roll out and run
Widen what passed, train the team, and keep monitoring as models and connected systems change.
You're left with a working AI workflow connected to your systems, a runbook, and a team that knows how to review and override it.
Why Evolve
- Built into your systemsAI connected to the CRM, inbox, documents and reporting you already run.
- AI only where it earns its placeRules handle routing, timing and checks. AI handles the reading and writing.
- You set the boundariesYou agree what runs automatically, what needs approval and what always stays with a person.
- Part of how you growAs a growth and technology partner, we treat AI as part of how you win, deliver and report on work.
Free 30-minute growth review
Talk through where AI could help
Tell us which process you'd like AI to help with and the systems involved. We'll come back to arrange a growth review, a practical conversation about priorities, constraints and whether we're the right fit.
What the conversation covers
- 01The processWhich process you'd like AI to help with, how often it runs and where it gets stuck.
- 02Your systems and dataThe CRM, inbox, documents and other tools involved, and what AI could safely reach.
- 03The next practical stepWhere it makes sense to start, or whether simpler automation would do the job.
Send the form and we'll reply by email or phone to arrange a time.
Every engagement is bespoke: we agree the scope and terms once we understand your business and priorities.
Implementation detail
The detail behind an AI implementation
For the people who will check, run or approve the work. Each item gives the short answer, with the full detail underneath.
Where AI fits, and when simpler automation will do
AI earns its place where inputs are messy or a reply needs judgement. Often the answer is a mix: rules handle routing, timing and checks, and AI handles the reading or writing.
Compare the five uses: where AI helps, when rules are enough and what stays with your team
| Use case | Where AI helps | When rules are enough | Stays with your team |
|---|---|---|---|
| Automated customer conversations | Where AI helpsAnswering routine questions, qualifying enquiries, booking appointments and following up past customers by chat, SMS or email. | When rules are enoughFixed messages on a schedule, such as a quote reminder. | Stays with your teamComplaints, unusual requests and anyone who asks for a person. |
| Enquiry triage and replies | Where AI helpsClassifying free-text enquiries, routing them and drafting a first response. | When rules are enoughThe form already captures the service and location. | Stays with your teamNon-standard replies, and high-value or urgent enquiries. |
| Document and email processing | Where AI helpsReading invoices, job sheets and emails with varied layouts, and entering the details into your systems. | When rules are enoughFixed forms or structured exports that map field by field. | Stays with your teamUnclear fields, values outside agreed limits and anything that changes an amount owed. |
| Internal knowledge assistant | Where AI helpsAnswering staff questions from your procedures, with the source shown and access limited by role. | When rules are enoughA few questions with fixed answers on a help page. | Stays with your teamQuestions the documents don't answer, and HR, legal or safety judgements. |
| AI reporting | Where AI helpsPlain-English questions and summaries over agreed metrics from CRM, pipeline or job data. | When rules are enoughA scheduled dashboard already answers the weekly questions. | Stays with your teamJudgements about individual performance, and decisions with real consequences. |
Data access and integration constraints
What AI can do depends on what it can reach. Before agreeing scope, we check each system's API, exports, permissions and usage limits. Where a system has no usable API or export, we scope around it.
How access and data are limited
- Least access: read-only where possible, with separate credentials per integration, limited to the fields the workflow uses.
- Provider terms: data retention and training terms checked before anything is connected.
- Sensitive details: personal, health or financial information stays out of prompts unless the workflow needs it and the handling is agreed.
Choosing a first AI pilot that's worth running
Start with one process and settle the business case first. The pilot then has one commercial question to answer: is this worth running and looking after once it ends?
Rare or high-stakes processes are better left until there's a checked result to build on.
What to settle before approving a pilot
- The process: it happens often enough to give a useful sample, and a person can check the output before it reaches a customer or changes an amount owed.
- Today's cost: the staff time, rework and missed work involved now, timed or sampled so there's a real baseline.
- Your team's time: a process owner, plus the people who check labelled examples and review output during the pilot.
- Cost after the pilot: estimated running costs, review time and upkeep at expected volumes, set against today's cost.
- The decision: the result that would justify continuing, agreed up front, and the one person who makes the call.
A pilot passes on measures agreed before it starts
Measures are compared with the current process where it can be measured. If the pilot misses one, we fix the cause, narrow the scope or recommend stopping.
Typical acceptance measures
- Accuracy: results match labelled examples checked by your team, to a threshold agreed for each field that matters.
- Errors that get through: mistakes reaching the CRM or a customer, counted separately with a lower tolerance.
- Review rate: the share of items that need a person, and why.
- Time and effort: from receipt to a confirmed record or reply, including review time.
- Running cost: model, platform and integration costs per item at expected volumes.
- Team confidence: the people using it understand the output and know when to override it.
What you receive
A working AI workflow, plus what your team needs to run it and change it safely.
Everything that's handed over
- 01
AI opportunity review
A prioritised shortlist of uses, with effort, running costs, risks and measures for each.
- 02
Workflow specification
Inputs, outputs, checks, review queues and approval boundaries, with the labelled examples used for testing.
- 03
The working implementation
Connected to your CRM, inbox, documents or other systems, and tested against agreed scenarios.
- 04
Pilot report
Results against each acceptance measure, the failures found and what was changed.
- 05
Runbook and change log
Ownership, the review queue, what to do when something fails, and every change made.
- 06
Staff guidelines and training
Which AI tools to use, what not to paste into them, and how to review and override output.
Client work involving AI-assisted follow-up
Scope of related past work. Each write-up sets out what was built and how its results were measured.
Questions we're often asked
Do we need our data cleaned up before starting?
Not completely. Part of discovery is checking whether the data a use case depends on is good enough. If it isn't, we'll say what needs fixing and whether it's worth doing first.
Which AI tools or models do you use?
We choose per use case, typically business offerings from established AI providers, connected to the CRM, automation and document tools you already use. We'll explain the trade-offs in cost, accuracy and data handling before anything is built.
Will our data be used to train AI models?
That depends on the provider and the plan. We prefer business and API offerings whose terms say customer data isn't used for model training by default, and we'll show you the relevant terms for anything we recommend.
Can AI reply to our customers on its own?
Yes, for suitable tasks such as answering routine questions, qualifying enquiries or booking appointments, within the approval rules agreed for that use case. Many businesses start with a narrow scope and widen it once results have been checked.
How long does an AI project take?
It depends on the scope and your systems. We prefer a small pilot on one workflow before wider rollout, and the proposal sets out the timeline for each stage.
Do you work with businesses outside Perth?
Yes. We're based in Perth and work with businesses across Australia.
Free 30-minute growth review
Find out where AI would take the most work off your team.
A free 30-minute conversation about one process, the systems involved and the next practical step.
Request a growth review