Context

A large, privately owned construction and civil infrastructure group operating nationally, with distinct business units spanning estimating, project delivery, payroll and finance, and a property development arm. Leadership could see generative AI moving their industry, but usage across the group was shallow, inconsistent and individual — pockets of curiosity, no shared capability.

The problem

The group's most valuable workflows were heavily manual. Estimators were assembling subcontractor requests for quote by hand — extracting scope from project documents, finding the right subcontractors, chasing responses by email, and comparing quotes line by line. Meanwhile, teams had access to AI tools but no structured way to build skill, and no view of where AI could genuinely move the needle versus where it was hype.

What we did

We started in the boardroom: Leaders Exploration Sessions with the executive and general management teams to build fluency, set direction and agree where to focus. From there we ran hands-on Accelerator Labs across six working groups, then role-based Claude training for the estimating team (around 30 people), the payroll and finance team, and the property development business — each session built around that team's real work, not generic demos.

Alongside the training, we worked with the group's AI lead to design a Claude-powered automation for the subcontractor RFQ process end to end: extracting scope and location from the estimators' RFQ list, retrieving the right subcontractors from their database, generating and distributing RFQ packages, drafting contextual follow-ups, filing incoming quotes, and producing a structured comparison with a risk-assessed recommendation for each package.

The outcome

• AI capability established across five parts of the business, from executives to estimators

• A phased, buildable design for automating the RFQ process — combining rules-based automation with Claude for drafting, extraction and reasoned recommendations

• Completion certificates issued to trained team members as capability milestones

• The engagement was extended twice, and renewed for a second year