What this looks like in practice

The problem

Clients send documents in whatever format they have. Before any work can start, someone has to sort through it, rename it, and get it into the right place.

Potential approach

• •When a file arrives, an agent identifies the document type, extracts vendor/date/amount, renames it to a consistent format, and moves it to the right subfolder.
• •Unreadable files get flagged for review.

The problem

Someone has to manually key invoice / bill data into Xero. For clients with high invoice volumes, that’s hours of low-value data entry.

Potential approach

• •Agent reads the designated folder, extracts vendor, date, amount, and GST from each file, and generates a table formatted for direct import into Xero as bills.
• •Accountant reviews and imports.

The problem

Once year-end is done in Xero, the numbers exist but the narrative doesn’t. Writing a plain-English summary for each client takes time most practices don’t have.

Potential approach

• •Agent pulls P&L and balance sheet from Xero, runs year-on-year comparisons, and drafts a plain-English management summary.
• •Output is a doc the accountant edits and sends. Can include commentary on GST position, provisional tax, and KiwiSaver.
• •Audit layer to review accuracy.

The problem

Setting up client folders and generating emails takes time, and it’s easy for mistakes to slip through.

Potential approach

• •When a new client is identified, the agent confirms client details then provisions appropriate folder templates.
• •Agent checks what’s been provided against an onboarding checklist and identifies gaps.

The problem

Writing up qualitative research is one of the heaviest back-end tasks in a consulting engagement. With six to eight interviews and a workshop, synthesis easily takes a day and a half before a draft exists. That time comes out of either quality or margin.

Potential approach

• •Agent reads transcripts and notes, extracts themes and supporting evidence, and produces a structured synthesis document with key findings and suggested implications.
• •You review, adjust, and curate. The agent doesn't produce a final deliverable — it produces a strong first draft.

The problem

Pricing bespoke work by gut feel worked when the baseline was stable. AI has changed what's deliverable in a given timeframe, and the person with enough context to scope accurately is usually too senior to be doing it manually every time.

Potential approach

• •You describe the client outcome and the activities you'd typically recommend. The agent maps these against your rate card and past engagement data to produce a structured estimate.
• •Output is a scoping document ready for review. The consultant checks, adjusts, and uses it as the basis for the proposal.

The problem

On a multi-workstream engagement, compiling updates, normalising the format, and producing a steerco-grade summary is easily a half-day task per reporting cycle. It lands on whoever is holding delivery, and it doesn't bill.

Potential approach

• •Workstream leads submit a short structured update. The agent pulls these together and produces a full-pack draft and a compressed exec summary.
• •Leads review their own section. The programme lead reviews the full pack. Neither starts from scratch.

The problem

The notes exist in Zoom or Teams, but turning them into a follow-up email, updating the pipeline, and logging next steps takes another 20 to 30 minutes per meeting. Across a week of client calls, that adds up.

Potential approach

• •Agent reads the transcript, extracts actions by owner, drafts a brief follow-up email, and produces a plain-language CRM activity note.
• •You review the draft email and the CRM update before anything sends or gets logged.

The problem

Rewriting supplier descriptions for every new product is one of the biggest time costs in running an online store. At five or six products a day, it adds up fast and it’s hard to delegate without losing your store’s voice.

Potential approach

• •Agent is built with a system prompt based on your existing copy — your product language, tone, and what you typically highlight.
• •You read, approve or edit, and the listing moves to the next stage. Writing becomes checking.

The problem

Once a listing is approved, it still needs to get onto the site. Logging into the admin portal, formatting, uploading — it’s repetitive work that sits between the content being ready and it being live.

Potential approach

• •Best case: supplier data in, approved listing live — no admin portal required.
• •If a full integration isn’t feasible, a formatted template makes it a simple copy-paste into the portal. That alone frees up meaningful time.

The problem

Posting content consistently can be tough when you’re starting from scratch across multiple channels every week. The result is irregular output and a pipeline that goes quiet.

Potential approach

• •Agent triggers weekly or fortnightly — not constantly. Surfaces ideas based on what’s changed: new stock, seasonal timing, recent engagement.
• •Delivers a draft post, image direction, and clear instructions for each platform — not just a prompt to work from.
• •You stay in control of what goes out. The agent does the preparation, not the publishing.

The problem

Chasing ETAs, confirming stock availability, and following up on orders generates a steady stream of low-value email. It has to happen, but it doesn’t need to come from you every time.

Potential approach

• •Agent is built with context about your suppliers and how you typically communicate with them. Point it at your product folders and it can answer questions about stock before drafting.
• •You describe what you need in plain language — “chase [supplier] on the order that’s overdue” — and it drafts the email ready for your review.
• •Once you’re happy, you send. The agent handles the drafting; you stay in control of what goes out.

The problem

Clients describe work in rough terms. An email, a few lines in a message, a forwarded brief from an architect or project manager. Someone has to decode that and translate it into line items. It's how most jobs start.

Potential approach

• •Agent reads the brief, matches the described scope against your rates, and drafts a formatted quote with line items, GST, and your standard terms.
• •You review the numbers, adjust if needed, and send. The job of quoting shifts from writing to checking.

The problem

Most scope changes happen informally. A message, an email, a conversation relayed from site. By the time it gets logged as a variation, the work is often already done and the invoice looks like a surprise.

Potential approach

• •Agent monitors incoming messages and emails for language that looks like a scope change.
• •When it finds one, it drafts a variation notice with the additional scope, estimated cost, and an approval request. The client confirms before work starts.

The problem

Jobs finish on site but invoices get raised in the office, days later. Materials need confirming, timesheets need checking, and it often lands on the one person managing everything else.

Potential approach

• •When job status changes to complete, the agent checks that open variations are accounted for and pulls materials and labour logged against the job.
• •No per-job configuration required. The trigger is the status change.

The problem

Most jobs require a Site-Specific Safety Plan and Safe Work Method Statements before work starts. Most platforms offer a blank template. Filling them in for each job is manual, repetitive work that gets rushed under time pressure.

Potential approach

• •When a job is created, the agent reads the scope and generates a pre-filled SSSP and relevant SWMS documents with hazards, controls, and contractor requirements based on the work type.
• •Under the Health and Safety at Work Act 2015, these documents are a compliance requirement. An agent that generates a consistent first draft reduces the chance something gets missed.