How an architecture firm saved 12 hours a week with AI

When a 120-person architecture practice approached us, they were not looking for a technology overhaul. They had a simpler problem: their senior architects and project leads were spending too much time on repetitive documentation and communication tasks, and not enough time on the design and client work that actually drives revenue.

They had tried AI informally. A few people were using ChatGPT for the occasional email draft. But there was no structured approach, no consistency, and no measurable impact.

Here is what happened when we worked with them to embed AI into their actual workflows.

The challenge

Like most professional services firms, the practice faced a familiar set of pressures:

• Senior staff buried in admin. Project leads were spending significant portions of their week writing meeting minutes, status reports, project briefs, and client correspondence. This was skilled work requiring judgement, but the actual writing was slow and repetitive.
• Inconsistent quality across teams. Different teams produced documentation in different formats and to different standards. There was no easy way to maintain consistency across a growing practice.
• Fee pressure and utilisation targets. Every hour a senior architect spent on documentation was an hour not spent on billable design work. The business needed to improve utilisation without hiring more support staff.

What we did

We ran a focused engagement over four weeks, working primarily with two project teams.

Week 1: Discovery

We sat with project leads and team members to understand their actual workflows. Not what they thought they should be doing, but what they actually spent their time on day to day.

The biggest insight was that the most time-consuming tasks were not the complex ones. They were the routine documentation tasks that happened every week on every project: meeting minutes, progress reports, consultant coordination emails, and briefing documents.

Week 2: Training

We designed and delivered targeted training sessions for the two pilot teams. Each session was built entirely around their real project documentation. We used their actual meeting notes, their actual report templates, and their actual client communication styles.

The training covered:

• Drafting meeting minutes from rough notes using AI
• Generating first drafts of project status reports
• Writing consultant coordination emails
• Creating briefing documents from project information

Every exercise used real examples from their current projects. Nobody was learning in the abstract.

Weeks 3 to 4: Embedding and support

This is where the real value was created. We worked alongside the teams as they started using AI in their daily work. We helped them refine prompts, build reusable templates, and troubleshoot the inevitable "this output is not quite right" moments that happen in the first few weeks.

We also helped the practice put a simple AI usage policy in place, covering what tools were approved, how to handle client-sensitive information, and quality assurance expectations for AI-assisted work.

Specific workflows that changed

Meeting minutes: Previously took 30 to 45 minutes to write up after each meeting. With AI, team members now dictate or paste rough notes and generate a well-structured set of minutes in under five minutes. A project lead running four meetings a week saves roughly two hours.

Weekly status reports: These were a consistent pain point. Each project lead spent 20 to 30 minutes per project writing weekly updates. With AI-assisted drafting from project notes and timesheets, this dropped to five minutes per report.

Consultant coordination emails: Writing clear, detailed emails to subconsultants (engineers, surveyors, planners) used to take 15 to 20 minutes each. With AI handling the first draft from bullet-point instructions, this dropped to under five minutes, with the architect focusing on reviewing and refining rather than writing from scratch.

Briefing documents: New project briefs that previously took an hour or more to draft from scratch are now generated in 15 minutes using AI with a structured template and project-specific inputs.

The results

Across the two pilot teams (approximately 20 people), the measurable outcomes after four weeks were:

• 12 hours per week of time saved on routine documentation across the two teams
• Faster turnaround on project documentation, with reports and minutes going out same-day instead of the following day
• More consistent quality across teams, with AI-assisted drafts following standardised formats
• Higher senior utilisation as project leads redirected documentation time to design and client work

The less measurable but equally important outcome was confidence. Teams went from "I have heard about AI but do not really use it" to "this is just part of how I work now." That shift is what makes adoption stick.

Lessons learned

Start with the boring tasks. The highest-impact workflows were not the most intellectually interesting ones. They were the repetitive documentation tasks that everyone dislikes but nobody had thought to automate. This is where AI delivers the most immediate, tangible value.

Real work beats demos every time. The training sessions that used actual project documents were dramatically more effective than any generic demo could have been. People could see the value immediately because the output was directly relevant to their next task.

Follow-up is not optional. The teams that had regular check-ins during weeks three and four adopted AI far more thoroughly than they would have from training alone. The support phase is where habits are built.

Keep governance simple. A one-page policy was enough to give people confidence to use AI without worrying about doing something wrong. Overcomplicating governance at this stage would have slowed adoption significantly.

Measure from the start. Having baseline time estimates for each workflow made it easy to demonstrate value to the broader practice. "We save 12 hours a week" is a far more compelling story than "people seem to like it."

What happened next

Based on the results from the two pilot teams, the practice expanded AI training across all project teams over the following two months. The approach was the same: targeted, workflow-specific training followed by structured support.

They are now exploring more advanced applications, including AI-assisted design brief analysis and automated project setup documentation.

If your practice or business is facing similar challenges, we would be happy to discuss what a practical approach looks like for your specific situation.