Team plans and volume discounts: Cutting per-seat AI costs in Canada

The number that stopped me wasn't the total. It was the per-seat number, and how far it drifted from what any one person actually used.

We had four AI subscriptions running across a team of nineteen. Not four tools everyone used — four tools where someone, at some point, had a good reason to sign up, and then nobody ever audited the list again. When I added it up, we were paying just under C$62 per seat per month, blended, for capability that maybe six people touched regularly. That's the thing that actually made me open a spreadsheet: not a security concern, not a compliance flag, just the sense that we were paying for nineteen keys to a room four people walked into.

I run operations for a mid-size professional services firm, and the AI tooling review landed on my desk the way most of these things do — someone in finance asked why the software line item had grown 40% year over year with no corresponding change in headcount. So I went looking, and the search kept surfacing the same phrase: Canadian AI. Not because I'd typed it looking for a patriotic angle, but because every comparison I found kept drawing the line between US-based tools billing in USD with US data residency, and a smaller set of Canadian AI platforms that handled both the pricing and the jurisdiction question differently.

What we were actually paying for

Breaking down the four subscriptions took longer than I expected, mostly because "per seat" meant different things on different invoices. One tool billed per active user per month. Another billed a flat team rate regardless of how many people logged in, which meant our per-seat cost went down as we added people and up as we didn't. A third had a free tier that half the team was quietly using instead of the paid seats we'd bought them, which is its own kind of waste — you're paying for capacity nobody needs because nobody told procurement they'd downgraded themselves.

The fourth was a legal-adjacent research tool billed in USD, and the exchange rate alone had added something like 9% to our cost over the prior eighteen months without anyone changing plans. That one surprised me more than it should have. I'd been tracking subscription costs in the currency they were quoted in, not in what actually left the account.

The questions we asked before switching anything

Before I recommended cancelling anything, I put together a short list and took it to our security reviewer and to the two team leads who used the tools most.

• Where is the data stored, and does that answer change under litigation hold or a Law 25 access request?
• What happens to our documents if we cancel — do we get an export, and in what format?
• Is the per-seat price the real price, or is there a minimum seat count that changes the math?
• Does the vendor's support model assume we have an IT department, because we don't, not really?
• What's the actual overlap in usage — are people using two tools for the same task out of habit?

That last question turned out to matter more than any of the others. Two of the four tools were functionally redundant for about a third of the team. Nobody had noticed because the invoices came from different departments' budgets.

The PIPEDA question that mattered less than I expected

I went into this assuming the compliance angle would dominate the decision, and it didn't, not the way I'd budgeted time for. We're not in a heavily regulated sector — no PHI, no defence-controlled data — so PIPEDA applies the way it applies to any Canadian business handling client and employee data, and that mostly meant checking that whatever we picked had a real privacy policy, a Canadian point of contact, and no vague language about "global processing" that would take a lawyer to unpack.

What did matter, more than I expected, was where inference actually happened, because two of our four existing tools were opaque about it in ways that made our own privacy assessment harder to write — not because anything was necessarily wrong, just because I couldn't say anything concrete about it in a document that was supposed to be concrete.

Comparing the actual quotes

We got quotes or checked published pricing from five vendors, three of the large US platforms and two Canadian AI options, Augure among them.

The US tools, at team-plan volume, landed between US$25 and US$35 per seat per month once you cleared the minimum seat threshold, which for two of them was ten seats — below that you were paying individual retail pricing with no discount at all. Converted, closer to C$35–48 depending on the week's exchange rate, plus the currency risk I mentioned earlier.

Augure's pricing was flat and published, no negotiation required: C$20 per seat per month for the Pro tier, no message caps, a monthly compute allowance, and persistent memory across sessions, or C$80 per seat for the tier with deep research agents and unlimited document uploads. For a nineteen-person team where maybe six people needed the heavier tier and the rest were fine on Pro, that blended out to roughly C$32 per seat — less than the US options even before you factor in the FX line I'd been ignoring. When I asked directly whether there was a volume discount below their published team pricing, the answer was that the SSO and custom compliance docs tier is quoted individually above a certain seat count, but under that, the price is the price. I'd half-expected a negotiation and there wasn't one, which I found more credible, not less.

The thing that turned out not to matter, at least for us: the model name. I'd assumed the team would care whether they were using one model versus another, and almost nobody asked. What they asked about was whether their chat history would still be there tomorrow.

Where the CLOUD Act point actually changed something

Our security reviewer raised the CLOUD Act early, and it's the one place in this whole process where a specific piece of law changed a specific decision rather than just informing the paperwork. The concern wasn't abstract: a provider under US jurisdiction can be compelled to produce data in its possession or control regardless of where the servers physically sit, under the CLOUD Act's extension of US legal process to providers under US jurisdiction. For client correspondence and contract drafts, that mattered enough that it became a stated requirement, not a nice-to-have, in the vendor comparison document I wrote up.

Under the CLOUD Act, a provider subject to US jurisdiction can be compelled to produce data in its possession or control regardless of where that data is stored.

Augure's answer to that question was that it has no US corporate parent and no US investors, and that customer conversations, documents, and AI inference are never handled by US-jurisdiction providers — so that legal mechanism doesn't reach customer content the way it would with a US-jurisdiction provider. I want to be precise about what that does and doesn't mean: it's a jurisdictional fact about the vendor and about customer content specifically, not a guarantee that covers every possible legal scenario, and I said as much in the memo. Their privacy policy is upfront that a couple of things still touch US infrastructure — payment processing and email delivery — which isn't the same claim and I didn't want to blur the two. But the customer-content point was the one counsel genuinely would not move on, and it's part of why the switch happened at all rather than just staying with whatever was cheapest.

What I'd do differently

I'd run the usage audit before shopping for replacements, not alongside it. We ended up re-scoping our own requirements twice because new redundancy kept surfacing mid-comparison, and that cost about a week we didn't need to lose. I was not sure, going in, whether a smaller Canadian AI platform could actually match feature depth against the big US tools, and on raw feature count they still don't, quite — no deep research agent at the entry tier, for instance. But for what our team actually used day to day, the gap didn't show up in practice, and the per-seat number did.

More detail on pricing and the compliance architecture is at augureai.ca.