Thursday, August 13, 2026 — Daily Business & AI Briefing
This morning's theme is AI moving into the commercial plumbing of business: discovery, payments, cloud infrastructure, power supply, and enterprise operations. For Canadian businesses, the immediate issue is still trade uncertainty ahead of the August 19 tariff deadline.
1. Canada rejects the latest U.S. tariff-relief offer as August 19 closes in
Canadian negotiators are reportedly dissatisfied with Washington's latest proposal. Negotiations are occurring daily, but the U.S. is still threatening 50% tariffs on a broad range of Canadian imports beginning August 19 unless its demands concerning autos, dairy, alcohol, and other trade issues are addressed.
Action before August 19: Identify your five most exposed inputs, suppliers, or customers and model three scenarios: Mild (+10%), Significant (+25%), Severe (+50%). A company can have relatively little direct U.S. revenue and still have substantial tariff exposure through equipment, materials, and suppliers.
2. AI shopping creates a new problem — who owns the customer?
Payments company Adyen warned that AI shopping assistants could increasingly handle the entire shopping journey — discovering products, comparing merchants, and potentially initiating payment — without a customer developing the same direct relationship with the retailer.
The emerging model: Customer asks AI → AI compares businesses → AI selects product/service → AI initiates transaction → Merchant fulfils order. The risk: becoming a fulfilment provider behind somebody else's interface.
Action: Make your business highly discoverable by AI — but simultaneously strengthen reasons customers come back directly. Prioritize customer accounts, email/SMS permission, service history, warranties, and loyalty programs.
New KPI: Direct Customer Rate = returning/direct customers ÷ total customers.
3. More than 80% of companies haven't deeply implemented AI
A Reuters survey of Japanese businesses found only 16% said AI had become an integral company-wide tool. About 60% use it only in limited parts of their businesses.
Better maturity scale:
| Level | Description |
|---|---|
| Level 1 | Ask AI questions |
| Level 2 | AI creates documents/content |
| Level 3 | AI accesses business data |
| Level 4 | AI participates in workflows |
| Level 5 | AI triggers controlled actions |
| Level 6 | Business measures ROI |
Action: Instead of deploying another general-purpose AI tool, move one existing workflow one level upward.
4. Cerebras shows AI growth alone isn't enough — margins matter
AI-chip company Cerebras fell more than 9% today after mixed quarterly results. Cloud business roughly quadrupled to US$126 million, but hardware revenue declined and adjusted gross margin fell to 40.6% from 46.5%.
Small-business lesson: Track Revenue + Gross Margin + Customer Retention + Cash Required + Delivery Capacity together. A C$500K business at 70% gross margin can be healthier than a C$1M business at 25% that constantly requires more employees and infrastructure.
5. Power companies turn existing industrial sites into AI infrastructure assets
Germany's RWE said it is close to agreements with data-centre operators at two of its existing sites. It owns about 30 locations with strong electrical infrastructure and sees roughly 10 sites with data-centre potential. UN-linked research projects global data-centre electricity consumption could reach around 945 TWh/year by 2030 — roughly double current levels.
For Canadian industrial operators: Underused infrastructure can itself be an asset. A site with spare power capacity, fibre access, or substantial electrical infrastructure may become more strategically valuable than its current operation suggests.
Key Numbers — August 13
| Signal | Metric |
|---|---|
| Threatened U.S. tariff on Canadian goods | 50% (Aug. 19) |
| AI integral company-wide (Reuters survey) | 16% |
| Limited AI adoption | ~60% |
| Cerebras cloud revenue | US$126M (~4×) |
| Cerebras gross margin | 40.6% |
| Big Tech expected 2026 capex | US$740B+ |
| RWE industrial sites | 30 |
| Projected 2030 data-centre electricity use | ~945 TWh/year |