Making Sense | Issue 04

01 — Feature Story 5 min read

Jobs have always changed. What's different now?

A diverse team collaborating around a table in a modern office

Long-term AI returns are best generated by working closely with your people.

Workforce AI ROI Change Management

Every wave of new technology has reshaped work. The printing press, electricity, the personal computer: each one made a category of work obsolete and created a category that hadn't existed before. The people in the middle of each transition rarely got to see how it ended.

The difference now is speed. Stanford's 2026 AI Index reports that AI passed 50% global user adoption in just three years, faster than either the personal computer or the internet reached the same threshold. 88% of organizations are now using it.

"Looking only at layoffs is shortsighted in terms of getting value from AI."

That's Helen Poitevin, a Gartner VP analyst, summarizing a survey of 350 executives at companies above $1B in revenue. Gartner found 80% of AI-piloting firms had reduced staff, with no correlation to ROI. KPMG's Canadian survey landed in the same window: only 3% of Canadian organizations report measurable AI ROI versus 8% globally, with the workforce skills gap as the top blocker. Anthropic's Dario Amodei has walked back his earlier prediction that AI would eliminate half of entry-level white-collar roles.

The productivity gains are real, but they aren't appearing where the firms are cutting. The Index cites a 14% productivity boost in customer service and 26% in software development for organizations using AI well. The firms getting durable returns are amplifying what their people do and redesigning roles around the new capability, while the ones cutting first are leaving that productivity on the table.

The practical move for any AI program right now is to design with your people, not around them.

Read the full piece →

Three Canadian AI stories from May.

BC sovereign data centres, OpenAI's PIPEDA finding, and Cohere's sovereign-AI positioning.

Telus and Ottawa announce three AI data centres for BC.

On May 11, AI Minister Evan Solomon and Telus CEO Darren Entwistle announced three AI factories for BC. The cluster will host 60,000+ GPUs across Kamloops and two Vancouver sites by 2032, run on 98% BC Hydro renewable power, and is projected to generate $9 billion in economic activity. Vass Bednar of the Canadian Shield Institute flagged an important caveat: residency without governance isn't sovereignty, and US CLOUD Act exposure doesn't disappear just because your provider has a Vancouver address.

Canadian privacy commissioners rule that OpenAI violated PIPEDA.

Filing records representing federal privacy regulation

On May 6, federal Privacy Commissioner Philippe Dufresne, with provincial counterparts from BC, Alberta, and Québec, announced that OpenAI violated PIPEDA and provincial privacy laws while collecting training data for GPT-3.5 and GPT-4. The investigation concluded that OpenAI's collection from publicly accessible websites was "overbroad and therefore inappropriate." The federal case is "conditionally resolved"; the BC and Alberta findings remain unresolved.

The signal underneath this is broader: Canadian privacy regulators aren't waiting for new legislation, and they're treating existing laws as already binding on foreign AI providers. That changes the regulatory backdrop for any AI procurement going forward.

Cohere positions as the practical Canadian sovereign-AI option.

Toronto skyline representing Canadian tech

Toronto-headquartered Cohere, now valued at $20 billion after acquiring Germany's Aleph Alpha, spent May positioning itself as the practical sovereign-AI alternative to its US counterparts. Chief AI Officer Joelle Pineau described the company as "a very low drama company" with an internal motto of "ROI over AGI".

If you're a BC regulated entity with a hard Canadian-residency requirement, and especially if you're already on SAP, Cohere via SAP's Sovereign Cloud is now the cleanest fully-Canadian procurement path. The trade-off to know going in is capability. Cohere isn't competing at the frontier of Claude or GPT-5.5, so the call comes down to whether sovereign residency matters more than capability for your specific workload.

Anthropic's distribution week, plus OpenAI's counter-push.

What shipped in the first two weeks of May, and what it means for procurement decisions.

Featured · Anthropic

Anthropic leans into distribution.

Anthropic ran its biggest distribution push of the year between May 4 and May 15. The list includes Claude Platform on AWS (full native API on AWS auth and billing, with same-day feature parity), a $1.5B enterprise services joint venture with Blackstone, Hellman & Friedman, and Goldman, Claude for Small Business, a $200M Gates Foundation partnership, and a new suite of financial-services agents. One of the biggest pieces is the expanded PwC alliance, which will train 30,000 PwC professionals on Claude Code and roll out Claude Cowork across hundreds of thousands of staff.

Ramp's May 13 AI Index also puts Anthropic at 34.4% business adoption on their 50,000-company panel, ahead of OpenAI's 32.3%. It's the first time that Anthropic has led, and the pattern is distribution into work rather than new models.

OpenAI is responding with DeployCo, its own new consulting arm, plus the acquisition of applied-AI firm Tomoro.

From our team

We're a member of Anthropic's Claude Partner Network and have been deploying Claude Code and Claude Cowork across BC client environments. If your team is moving from "we have Claude licenses" to "Claude is part of how the work gets done," our AI Literacy Program is built around that transition.

Other Releases

• OpenAI's Codex push and new voice models. Across May, OpenAI shipped Codex on Windows, mobile and IDE deployment, and around 15 role-specific case studies (NVIDIA, Databricks, AutoScout24, Sea). New voice models in the API include GPT-Realtime-Whisper for low-latency streaming speech-to-text, plus updated transcription and translation.

• Google's Gemini 3.5 Flash at I/O. Google used I/O 2026 to launch Gemini 3.5 Flash, its first model designed to combine frontier-level capability with low-latency execution. It outperforms the older Gemini 3.1 Pro across virtually all internal benchmarks, with a significant jump in complex coding and agentic tasks, and claims 4x the output token speed of competing frontier models.

• ByteDance Seedance 2.0 in CapCut. ByteDance integrated its state-of-the-art video generation model into the CapCut editor, while OpenAI's Sora consumer product is retreating. The fastest-moving creative AI category is now video.

Four numbers for your next meeting.

• 3% of Canadian organizations report measurable AI ROI versus 8% globally, per KPMG Canada's May 2026 survey of 306 C-suite leaders. KPMG Canada

• 95% of AI transformation failures trace to organizational factors rather than technology, across 51 successful enterprise deployments. Stanford Digital Economy Lab

• $9B in projected economic activity from Telus's BC AI factory cluster, with 60,000+ GPUs by 2032. BetaKit

• 73% vs 23% of AI experts versus US public who expect AI to positively impact jobs. Pew Research