Subscribe to Making Sense

No spam, no jargon. Unsubscribe anytime.
Makes Sense. You're subscribed to the AI signal for Canadian business leaders. Every two weeks. No hype, no jargon.
All issues Issue 09 · July 30, 2026

Making Sense.

AI, translated for Canadian business leaders.

Editor's Note

In this issue we have an exciting new case study to share from our work with MOSAIC, outlining how we equipped their settlement professionals to reclaim time from manual entry and burdensome administrative tasks. This is time that can now be used to better serve local immigrant workers through the important programs they offer here in BC. I continue to be impressed at how fast we can move with AI: mapping processes across multiple programs, quantifying task-based workloads, and delivering a prototype, all in ten weeks.

If you've been following the Canadian AI buildout, and the drama around new data centres in BC, we've also compiled the latest scoop to share with you. Read on for those updates, and a recap of the latest important AI releases. Enjoy!

In this issue
  1. 01
    Giving Frontline Staff Their Time Back.
    A national survey says admin work is the biggest barrier for people-serving teams. Learn how MOSAIC turned that problem into a quantified case and a validated prototype in ten weeks.
  2. 02
    Three stories from the past two weeks.
    Bell bets that Canada's AI future should run on homegrown infrastructure, Ottawa asks Canadians how transparent AI should be, and Vancouver's data-centre fight becomes a campaign issue.
  3. 03
    Release Radar.
    Claude Opus 5, the Chinese open-weight wave, Gemini's Flash lineup grows, and Clio's AI legal assistant lands in Canada.
  4. 04
    Four numbers for your next meeting.
    How Canadian leaders rate their own AI returns, how far AI use is spreading past job descriptions, and the compute deal underwriting it all.
01 — Feature 3 min read

Giving Frontline Staff Their Time Back

A national survey names admin burden the top barrier facing people-serving teams. Here's what one BC non-profit did about it.

A settlement worker reviewing paperwork with a client across an office desk
AI Prototyping Process Redesign Non-Profit

For any organization that serves people directly, administrative work is a tax on capacity. A survey of Canada's developmental and social services sector, published this year, found that 68% of organizations named limited staff time and capacity as the single biggest barrier to quality care, ahead of funding or staffing levels. Reclaiming time from administrative work topped the list of priorities for 70% of respondents for the year ahead. The goal is clear, but where to start is the harder question for many organizations.

MOSAIC, a Vancouver non-profit that helps newcomers to Canada navigate employment, housing, language, and immigration paperwork, was asking exactly that. Its settlement professionals work across nine disconnected systems, manually re-entering session details after every client conversation. Leadership knew the load was heavy, but lacked a clear picture of where the time was actually going.

“What stood out was how Sam worked with us. They involved our frontline staff at every step, and they behaved exactly the way they said they would. We came in with a documentation problem and came away with a clearer picture of our own work, a prototype to build on, and a partner we trust to keep going with.”

— Adrienne Bale, Senior Manager Settlement Programs, Family & Settlement · MOSAIC

Over ten weeks, we mapped the end-to-end settlement worker journey across two of MOSAIC's programs, quantified the workload task by task, and built a working prototype. Not every bottleneck we found called for AI, and our process reflected that. We found that scheduling across four offices was costing missed appointments, and all that required was a simpler booking process. MOSAIC walked away with a quantified case for where its time goes, a validated prototype to build on, and a booking process solution.

Read the full case study
02 — In The News

Three stories from the past two weeks.

Bell bets on an all-Canadian AI stack, Ottawa opens a second consultation on AI transparency, and Vancouver's data-centre fight becomes a campaign issue.

01

Bell's Bet on an All-Canadian AI Stack.

Downtown skyline at sunset reflected in the river, Saskatchewan

While Telus builds AI data centres in BC, construction is underway on a $1.7-billion, 300-megawatt AI data centre outside Regina, Bell's largest AI investment yet. Around the data centres, Bell has also assembled a coalition it calls the Canadian Sovereign AI Alliance. Among its members, Cohere supplies the foundation models, Coveo layers in search and generative AI for government services, and BUZZ HPC operates the GPU clusters, all running on Bell's own infrastructure. It's a different kind of coalition than the one in last issue's feature (where Scotiabank, Sun Life, and Telus pooled resources to build shared AI governance), but continues the pattern of Canadian companies choosing to build AI together.

Canadian corporate bond issuance is having a record year, with many tech firms borrowing to fund exactly this kind of AI infrastructure. Infrastructure providers are now explicitly selling “Canadian-owned” as a feature, and that pitch is only going to get louder as more of these deals close. That's good news for a market that's historically been short on tech exposure, but analysts are starting to flag it as a concentration risk if the AI capital cycle turns.

Read more →
02

Ottawa wants your opinion on how transparent AI should be.

Looking up at Parliament Hill's Peace Tower against a partly cloudy sky

Canada's AI Minister, Evan Solomon, opened a public consultation on July 23 asking Canadians, businesses, researchers, and Indigenous groups to weigh in on AI transparency rules. It's open until September 23, and it's the second time this year the federal government has gone to the public for input, following the consultation that shaped the national AI for All strategy released in June. The survey focuses on five areas: detecting AI-generated content, letting people know when they're talking to a system rather than a person, making information about how AI systems are developed and how they work easier to understand, tracking serious AI-related incidents, and monitoring what autonomous AI agents are actually doing.

Critics of the AI for All strategy released in June claim it was heavy on ambition and light on the actual privacy and safety detail needed to enforce it. This consultation will help shape the definition of “meaningful transparency” in Canada, which is worth watching for any organization already deploying AI-generated content or customer-facing agents.

Read more →
03

Vancouver's data-centre fight is now officially a campaign issue.

Vancouver's downtown skyline and BC Place viewed from a rooftop garden

We've covered Vancouver's AI data-centre pushback in previous issues, and now it has reached another turning point. On July 21, City Council unanimously voted to reverse its own decision to send the Telus-Westbank rezoning application (for the Mount Pleasant site at 111 East 5th Ave.) to a public hearing. The official reason was volume: nearly 600 written submissions and almost 200 registered speakers couldn't be heard in the time available. But the vote also spared councillors running for re-election on October 17 from a communications blackout that would have barred them from discussing the proposal during the campaign. The application now goes back to staff review, with no further public consultation until after the election and a possible return to Council as late as Q1 2027.

On July 26, the group No AI Vancouver held its third demonstration since May, this time a stationary rally at the Art Gallery rather than a march. Organizer Guerric Haché was explicit about the rally's strategy, saying “We hope to demonstrate a concrete electoral risk to politicians who refuse to commit to pausing hyperscale data centre development.” Speakers at the event also included BC Green Party leader Emily Lowan, alongside two city council candidates. On the other side of the story, Telus maintains the three-site BC project would generate roughly $9 billion in provincial economic activity, run on 98% renewable BC Hydro power, and recover enough waste heat to warm about 150,000 homes.

Read more →
03 — Release Radar
  • Anthropic released Claude Opus 5, pitched as coming close to flagship-level performance at half the price of a top-tier model. It matches Opus 4.8's pricing while leading several new benchmarks, including a 3x score over the next-best model on ARC-AGI 3.
  • Two notable open-weight releases: Moonshot AI's Kimi K3 (2.8 trillion parameters, the largest open model yet) and Thinking Machines Lab's Inkling (Apache 2.0-licensed, positioned as a strong base model for fine-tuning rather than a frontier contender).
  • Google expanded its Gemini Flash lineup with 3.5 Flash-Lite (built for high-throughput, low-latency agentic work) and 3.5 Flash Cyber (a specialized model for finding and patching security vulnerabilities, launching first through a limited pilot with governments and trusted partners).
  • Burnaby, BC-based Clio brought its AI work assistant, Clio Work, to Canadian lawyers, letting them summarize cases, pull relevant case law, and draft documents grounded in a database of more than 470,000 Canadian cases.
04 — Quick Facts
  • A Deloitte Canada survey found that 90% of senior Canadian leaders report positive AI productivity gains over the past year, but Deloitte's own read is that most are measuring value too narrowly, focusing on cost savings rather than revenue growth or risk reduction. Deloitte Canada
  • New OpenAI research on more than 800,000 messages found that 43.5% of non-generic ChatGPT work messages relate to tasks outside the user's own occupation, early evidence that AI is expanding what people do at work before job titles catch up. OpenAI
  • AMD is planning to invest up to $5 billion in Anthropic as part of a deal that will see Anthropic deploy 2 gigawatts of AMD's newest AI chips. CNBC
  • OpenAI's newly announced Project Camellia is a 3.2-gigawatt data-centre campus in Effingham County, Georgia. This project is more than 10x the entire capacity BC has capped for AI data centres provincewide over the next two years. OpenAI