Case study - construction

NZ builder: 29 AI use cases, mapped and ranked

A New Zealand builder doing residential and commercial work. Before building any AI tools, we ran two workshops with their AI cohort to find where AI fits, then ranked every use case.

Talk to us about your projectSee what we did
The top eight rows of the use case register: quote review scores highest at 5.33, then sub-contractor invoice reading, progress claims, site diaries, email drafting, voice dictation and Claude in Word, Outlook and Excel at 4.00, each with hours saved, confidence and weeks of effort
Client
A New Zealand construction firm
Industry
Residential and commercial construction
Our role
AI audit: two workshops, a use case register, an AI policy and strategy
Delivered
April 2026

The audit in numbers

  • 29AI use cases found, scored and ranked
  • 17people in the AI cohort, from leadership to site
  • 12items in the 90-day plan: Claude setup plus eleven builds

The brief

The firm set up a 17-person AI cohort, from leadership and quantity surveying to project and site managers. Some were already using AI day to day.

The brief had two parts: train the team up, then find where AI would actually save time and rank it.

A workshop whiteboard titled Completion doc and O&M manual: steps like reviewing consent and client requirements, emailing subcontractors for documents and sending the pack to the client, most tagged AI
Completion documents, mapped on Day 1. Seven steps tagged AI, two tagged O.

How we found where AI fits

  1. Map the real workEach role group maps its own processes, step by step.
  2. Tag every stepHuman only, AI with a person checking, or fully automated.
  3. Sit with the teamsFour small-group sessions walking through their actual work.
  4. Score every use caseHours saved a week, how sure we are, weeks to build.
  5. Directors set the orderThe scores are one input. The firm's own priorities are the other.

Start where the work is slow, not where AI fits.

Every step tagged by the people who do it

On Day 1, each role group mapped its own processes on the whiteboard, one step at a time. Then they tagged every step: H for human only, O for AI with a person checking, AI for fully automated. Very few steps stayed human only.

A workshop whiteboard titled Quote Review and Procurement: steps like confirming the scope of works, checking subcontractor quotes cover it and drafting clarification emails, each tagged O, beside a key reading H = Human, O = AI/Human oversight, AI = Full AI automation
Quote review, tagged step by step, with the key at the top.
A workshop whiteboard titled Site Management: QA reminders and a to-do list for site teams tagged AI, and steps like daily site diaries and toolbox reports tagged O
Site management: two steps tagged AI, six tagged O.

29 use cases, scored and ranked

Both days fed one register: 29 use cases across QS, project, site, admin and leadership. Most are admin and document handling: invoices, progress claims, reports and email.

Each use case is scored on hours saved per week, weighted by how sure we are the saving is real, against weeks to build. Quote review came out on top.

A bar chart of all 29 use cases ranked by priority score, from quote review at 5.33 down to a reception voice agent at 0.35, with the 11 builds in the 90-day plan highlighted in blue
Blue builds made the 90-day plan. The scores guide the order, they do not set it.

A 90-day plan, with the directors deciding

The top of the register became a 90-day plan: Claude setup and team training first, then quick wins like email drafting and voice dictation, then the QS builds.

The directors set the final order. Where staff had already started with Claude, the plan builds on their work instead of replacing it.

Alongside it, the firm got an AI use policy: which tools are approved, what data can go where, and when a person must check the output.

A 12-week roadmap chart: Claude setup and training in weeks 1-2, then staggered builds for voice dictation, email drafting, invoice extraction, quote review, site diaries, progress claims, toolbox talks and regional reporting, with meeting notes and tender submissions starting in week 10
The 90-day plan, from the same register.

Not sure where AI fits in your business?

We start with the work, not the tools. Your team maps how the week actually runs, tags every step, and you get a ranked list of where AI saves time.

Talk to us about your projectAI for construction and field services

Last updated 20 September 2026