Black Friday Prep: What Burnaby Shops Should Automate

October is the last month in which you can still change how Black Friday goes. By mid November your staffing is locked, your inventory is committed, and every spare hour goes to shipping. So the useful question for a shop in Burnaby or along Columbia Street in New Westminster isn't whether AI is worth trying. It's which specific jobs you can hand over in the next four weeks and still have time to unwind if they go badly.

Here is what the evidence actually supports, including the parts that argue against automating.

Check your assumption about online sales first

Statistics Canada reported that in November 2025, on a seasonally adjusted basis, retail e-commerce sales fell 2.8% to $4.0 billion, "accounting for 5.7% of total retail trade, compared with 6.0% in October" (Statistics Canada, Retail trade, November 2025). Total retail sales rose 1.3% that month, to $70.4 billion.

Seasonal adjustment already removes the expected holiday bump, so this doesn't mean nobody shopped online in November. But it's a useful reality check. Across Canadian retail, e-commerce is somewhere around one dollar in seventeen, and in the Black Friday month it didn't outrun the storefront. If you run a physical shop, the automation that pays back is usually the kind that removes friction around the store, not a rebuild of your online funnel three weeks before the rush.

Do the bulk text work now, while it's cheap

The most underrated tool for October is asynchronous batch processing. Anthropic's documentation describes the Message Batches API as processing large request volumes "with most batches finishing in less than 1 hour while reducing costs by 50% and increasing throughput," and states that "All usage is charged at 50% of the standard API prices" (Anthropic, Batch processing).

The catch is in the same document: "You can access batch results when all messages have completed or after 24 hours, whichever comes first. Batches expire if processing does not complete within 24 hours." That makes batching an October tool, not a Black Friday morning tool. Queue it when nothing depends on the answer arriving in ten seconds.

On price, check the published page rather than anyone's memory. As of September 2026, Anthropic's pricing page lists Claude Haiku 4.5 at $1 per million input tokens and $5 per million output tokens, with batch pricing at $0.50 and $2.50 respectively (Anthropic, Pricing). The same page carries a vendor worked example, which is illustrative rather than a promise about your shop: roughly 10,000 support ticket conversations averaging about 3,700 tokens each, run on Haiku 4.5, comes to about $37.00. Model names and prices move, so re-check before you budget.

Jobs that batch well in October:

• Product descriptions and image alt text for the catalogue you never finished
• Returns, exchange, holiday hours and shipping cutoff pages, rewritten so a customer can answer their own question at 11pm
• Draft replies to the twenty questions you get every single year
• Tagging last year's support inbox by theme, so you find out what actually broke instead of guessing
• Translating your top five pages for the languages your Metro Vancouver customers actually speak

None of that touches money, inventory or a live customer. That's the point.

The constraint on email and SMS is consent, not copy

This is where good AI copy gets businesses in trouble. Canada's Anti-Spam Legislation governs commercial electronic messages, and the government's guidance is specific: express consent has no time limit unless the recipient withdraws it, while implied consent from an existing business relationship is time limited, running up to two years from a purchase and only six months from an inquiry or application (ISED, Getting consent to send email).

Every message needs your business name, a current mailing address, and a phone number, email or web address, plus an unsubscribe mechanism. Unsubscribe requests must be "actioned within 10 business days or less and at no cost to the recipient."

So the automation worth building isn't a subject line generator. It's the plumbing: date stamp every consent record, expire implied consent automatically at 6 and 24 months, and route unsubscribes to your send list the same day rather than the tenth. A rented list plus a great AI writer is just a faster way to break the law.

If you put a bot on the site, say it's a bot

Canada's federal, provincial and territorial privacy authorities issued joint principles for generative AI on December 7, 2023. Two lines matter for a retail chatbot. System outputs that could significantly affect someone should be "meaningfully identified as being created by a generative AI tool." And "accountability for decisions rests with the organization, and not with any kind of automated system used to support the decision-making process," with an effective challenge mechanism for significant decisions (OPC, Principles for responsible, trustworthy and privacy-protective generative AI technologies).

In practice: label the widget, keep a visible path to a human, and never let it be the last word on a refund.

Where this doesn't apply

Anything on the critical path during peak load. A 2026 benchmark study of agent reliability found that even semantically equivalent rephrasings of the same task dropped success from 96.9% to 88.1%, and that rate limiting was the most damaging fault type tested (Gupta, ReliabilityBench, 2026). Black Friday is the day your rate limits get hit. Treat that finding cautiously, since it used simulated domains and a small model set, but the direction is consistent enough: keep AI out of checkout, payment and inventory writes.

Shops where AI genuinely isn't the bottleneck. Statistics Canada found 19.2% of businesses used AI to produce goods or deliver services in the twelve months to the second quarter of 2026, up from 6.1% two years earlier. Among those using it, 28.2% used virtual agents or chatbots. But 40.0% of all businesses said AI simply isn't relevant to what they make or do, and 13.4% cited cybersecurity or privacy concerns (Statistics Canada, Analysis on artificial intelligence use by businesses in Canada, second quarter of 2026). If your constraint is foot traffic, shelf space or one supplier who ships late, sitting in that 40% is a defensible call.

Anything you can't test twice. If it isn't running and observed by mid November, it ships in January instead. That rule has saved more Black Fridays than any tool.

Sources

• Statistics Canada. *The Daily, Retail trade, November 2025*. Released January 23, 2026. https://www150.statcan.gc.ca/n1/daily-quotidien/260123/dq260123a-eng.htm
• Statistics Canada. *Analysis on artificial intelligence use by businesses in Canada, second quarter of 2026*. Released June 11, 2026. https://www150.statcan.gc.ca/n1/pub/11-621-m/11-621-m2026010-eng.htm
• Anthropic. *Batch processing, Claude Platform documentation*. Accessed September 2026. https://platform.claude.com/docs/en/build-with-claude/batch-processing
• Anthropic. *Pricing, Claude Platform documentation*. Accessed September 2026. https://platform.claude.com/docs/en/about-claude/pricing
• Innovation, Science and Economic Development Canada. *Getting consent to send email, Canada's Anti-Spam Legislation*. https://ised-isde.canada.ca/site/canada-anti-spam-legislation/en/getting-consent-send-email
• Office of the Privacy Commissioner of Canada and federal, provincial and territorial privacy authorities. *Principles for responsible, trustworthy and privacy-protective generative AI technologies*. December 7, 2023. https://www.priv.gc.ca/en/privacy-topics/technology/artificial-intelligence/gd_principles_ai/
• Gupta, A. *ReliabilityBench: Evaluating LLM Agent Reliability Under Production-Like Stress Conditions*. arXiv, 2026. https://arxiv.org/abs/2601.06112

If you want a second opinion on which of these is worth doing before November, we're happy to give one. Autana Solutions builds AI employees for small and mid sized businesses around Burnaby and New Westminster, and we'll tell you when the honest answer is to skip it this year. Book a free call and bring your actual bottleneck.