Article archive — page 2 of 22
Your sales history already says which customers are missing something. Three patterns AI can find in it, and the line between a useful suggestion and a nuisance.
A no-show costs the whole slot, not the deposit. What to send, when to send it, and the one thing that matters more than the reminder itself.
Anthropic and OpenAI both announced cheaper models on the same day. Three things in the pricing structure decide whether any of it reaches your invoice.
AI rollouts fail differently from software rollouts, because nobody is forced to use them. Four things that predict whether adoption holds past month two.
Most late invoices are late because nobody followed up, not because the customer refused. Which parts of collections to automate, and which to keep human.
ISED is asking whether Canada needs AI disclosure rules, incident reporting and agent tracking. Five areas, what each could mean for you, and how to respond.
The best purpose-built legal research AI scores 54% on its own benchmark. A tiering rule for which documents you can draft with AI and which need a lawyer.
If a task took four hours and now takes one, billing by time turns efficiency into a pay cut. Four pricing responses, and how to pick before a client asks.
OpenAI published its own benchmark for Astra for Law: 54% correctness, up from 38.7%. Publishing that number is to its credit, and it should recalibrate buying.
Who can log into what, and who can approve what. Four controls a small business can put in place this month, now that agents are holding credentials too.
Assistants now summarise your reviews before a buyer sees one. What that changes about which reviews matter, and how to reply so the summary improves.
Password managers have started handing credentials to AI agents. The design is better than the headlines suggest, and the real risk sits somewhere else.
A survey found 27% of Canadian owners are unsure how AI applies to them. Usually nobody has written down how the work happens. The one-hour fix.
SOPs written for people lean on judgment nobody wrote down. What changes when software follows the same document, and how to write one in an hour.
A survey of 1,000 Canadian owners found most feel no impact from AI at all. The numbers, the caveats, and what the small group seeing returns did differently.
Insurers now ask about patching cadence, MFA and AI tools before they quote. What a policy actually pays for, and the controls that decide your premium.
Awareness training stops some attacks and is useless against others. Which category your real risks fall into, and what to spend on instead of more modules.
One attacker used AI agents to compromise PaperCut servers at 395 organisations. The fix had shipped weeks earlier. What the AI did, and what it did not do.
BDC’s LIFT program ties a preferential rate to Canadian content, which turns vendor nationality into a financial question. What counts, and how to verify it.
A PIA is seven questions, not a legal document. What to write for an AI use case, the five parts generic templates miss, and when you actually need one.
Edge0 ran a 35B model on an iPhone using 2.9 GiB of memory. It also needs 23 GB of storage and runs at 15 tokens a second. What that means for field work.
Predicting failures needs failure history most businesses never recorded. What the data requirement actually is, and the cheaper thing to do first.
Most warehouse time goes on walking, looking and paperwork, not lifting. Four software-first changes that cut hours before anyone quotes you for machinery.
Guidance for GPT-6 Astra says to strip out instructions written for older models. What transfers to ordinary business use, and what is developer-only advice.
Weekly case studies, new tools, and ROI playbooks for Canadian SMEs. One email, zero spam.