Language Learning Software Procurement Guide (Schools)

Every AI language tutor pitch to a school procurement office promises the same thing: one flat price, unlimited practice. Look at the vendor's own infrastructure bill and that promise falls apart at scale, because the only AI language pricing a school can actually put in a budget is pricing built around a hard, published usage cap.

Why AI Language Software Breaks Normal Software Budgeting

Most edtech procurement instincts come from software where usage is free to the vendor. A learning management system, a Google Workspace seat, a PDF library — once it's built, an extra hour of student use costs the vendor close to nothing. So vendors happily sell those seats flat, unlimited, and hope you use them more. Usage is upside, not cost.

Real-time AI voice tutoring inverts that model. Every minute of live conversation runs speech recognition, a language model, and speech synthesis — three paid inference calls, back to back, for every minute a student talks. That is a genuine per-minute cost that scales directly with how much your students actually use the tool.

A vendor selling "unlimited" AI conversation practice at one flat per-seat price is making a bet: that most students won't use it enough to matter. That bet is fine for a consumer app monetizing casual, low-engagement users. It is a real problem for a school contract, where the entire point of the purchase is to drive daily usage across an entire cohort.

Caps Are a Feature, Not a Limitation

A contractual usage cap is what makes a per-student budget line real. Without a cap, your worst-case annual cost is unknowable — it depends entirely on how enthusiastically your students adopt the tool, which is exactly the outcome you're trying to encourage. With a cap, the worst case is arithmetic.

LinguaLive publishes its caps for this reason: the free tier is 10 live voice minutes a day, and Pro is $7.99/month or $79.99/year with a 20-minute daily cap. Run the worst-case math on the annual plan: $79.99 divided by a theoretical maximum of 7,300 minutes a year (20 minutes × 365 days) works out to roughly $0.011 per minute if a student maxed out the cap every single day, which almost none do. The number that matters isn't that per-minute figure — it's that the ceiling exists at all, so the total line item is bounded no matter how hard your rollout succeeds.

Question a budget holder asks	With a published cap	Without a cap ("unlimited")
Can I state a worst-case annual cost?	Yes — seats × price, fixed	No — depends on actual usage the vendor won't commit to
Does high adoption increase our bill?	No — price is per seat, not per minute used	Unclear — often resolved later via a "fair use" clause
Can I defend this line item to a finance committee?	Yes — the number doesn't move	Only if you assume usage stays low
Does the vendor's own margin depend on light usage?	No — the cap protects both sides	Yes — that's the entire pricing model

Red Flags in AI Pricing

Once you know caps are what make a budget defensible, "unlimited" reads differently. Here's what to watch for in any AI language vendor's pricing page or sales deck.

Red flag	What it actually signals
"Unlimited" plan, no published cap	The vendor is pricing on the bet that average usage stays low. If your rollout succeeds and daily use rises, their economics — and then yours — change, usually via an unannounced fair-use policy, a forced upsell, or a quiet quality downgrade.
No published per-seat price ("book a demo" only)	You can't build an approvable, comparable budget line without a fresh back-and-forth negotiation at every renewal.
Vague "fair use" language instead of a number	There is a cap — it's just undisclosed until you exceed it, at which point you have no negotiating position.
Overage terms left undefined	A successful rollout becomes a financial surprise instead of a planned expansion you priced in advance.
No data deletion clause	Post-contract liability for student data sits with your institution regardless of who is currently storing it.

The Real Comparison: Per-Seat AI vs Human Tutor Hours vs Doing Nothing

Procurement conversations often skip straight to "is this cheaper than a tutor?" without laying out what each option actually buys. For a cohort of 200 students, here's the honest shape of the comparison using canonical, published figures only.

Option	Illustrative annual cost (200 students)	What it buys
AI daily speaking practice (LinguaLive Pro, annual)	200 × $79.99 = $15,998/year	Daily conversation reps capped at 20 minutes/day/student, available any time, no scheduling, no instructor hours consumed
1:1 human tutoring (iTalki/Preply-style, typical listed rates as of mid-2026 run roughly $10-$30/hr for standard lessons, more for certified specialists)	200 × 30 hrs/year × ~$20/hr (midpoint) = $120,000/year	Personalized correction and curriculum-specific coaching, but scheduled, hour-limited, and expensive to scale to daily practice
Status quo — classroom-only speaking practice	$0 direct cost	No new line item, but speaking practice stays bottlenecked by class time; the opportunity cost is real, it's just not a budget number

These aren't substitutes for each other. AI voice practice covers the daily repetition volume no timetable can fit. Human tutors cover judgment calls AI tools genuinely can't make yet — nuanced accent coaching, high-stakes interview feedback, complex curriculum design. If your program includes interview or exam prep specifically, see how the economics compare directly at the LinguaLive vs Preply comparison.

And doing nothing is a real decision with a real cost, not a neutral default — we cover the pedagogical side of that gap, separate from the money question here, in how university language centres are closing the speaking-practice gap.

The RFP Question Bank: 15 Questions for Any AI Language Vendor

Use this as a literal question list in your RFP or vendor call. A vendor that answers all fifteen in writing has a procurement-ready product. One that dodges more than two or three is still figuring out its own unit economics — at your expense.

#	Category	Question
1	Pricing	What is the exact per-student (or per-seat) annual price at our anticipated enrollment?
2	Pricing	Is that price published, or does it require a custom quote we can't independently verify against a rate card?
3	Pricing	Does the contract price change mid-term if usage rises?
4	Usage caps	What is the exact usage cap per student per day or month — in minutes, sessions, or messages?
5	Usage caps	What happens when a student hits the cap: a hard stop, a throttled experience, or automatic paid overage?
6	Usage caps	What is the per-unit overage rate, and is it itself capped or open-ended?
7	Usage caps	Can we model our worst-case annual cost using only the published cap and overage rate, without your sales team?
8	Data & compliance	Where is student data — voice recordings, transcripts, personal identifiers — stored, and under which jurisdiction?
9	Data & compliance	Is a signed Data Processing Agreement available covering processor obligations under GDPR Article 28?
10	Data & compliance	If handling US student education records, how does your vendor agreement satisfy FERPA's "school official" exception (34 CFR § 99.31(a)(1)(i)(B)) — direct control, use limitation, and no redisclosure?
11	Data & compliance	What is the data deletion timeline after contract termination, and is it a contractual guarantee, not a policy statement?
12	Contract terms	What is the initial term length, and does it auto-renew without an opt-out notice window?
13	Contract terms	If we run a pilot, is the per-seat price locked for the first full-scale contract, or renegotiated once we depend on the tool?
14	Contract terms	What happens to student accounts and progress data if we don't renew — is export available?
15	Contract terms	Who is our named point of contact for a billing dispute, and what is the SLA for resolving one?

Contract Terms That Matter Beyond Price

Price and caps get the attention, but three other terms decide whether the contract is actually safe to sign.

Data deletion. Under GDPR Article 17, individuals have a right to erasure of personal data — your contract should mirror that as a hard deletion deadline after termination, not a vague "in accordance with our privacy policy" line. Pair it with a signed Data Processing Agreement under Article 28 if any student is an EU resident. For US institutions, according to the U.S. Department of Education's Protecting Student Privacy program (studentprivacy.ed.gov), FERPA's "school official" exception (34 CFR § 99.31(a)(1)(i)(B)) permits disclosure of education records to a vendor without consent only if the vendor performs a service the school would otherwise handle itself, stays under the school's direct control over how it uses the records, and is limited under § 99.33(a) to using that data solely for the disclosed purpose — confirm in writing, with counsel, that the vendor's contract actually meets those conditions.

Term length and auto-renewal. A one-year term that silently rolls into a second year without a real opt-out window locks you into next year's price before you've evaluated this year's results.

Pilot-to-licence conversion. If you're piloting before a full rollout, get the full-scale per-seat price locked in the pilot agreement itself. Vendors routinely quote favorable pilot pricing, then reprice upward once a school is dependent on the tool and switching costs are real.

Exit and portability. Confirm what happens to student progress data and accounts if you don't renew. A vendor unwilling to commit to an export path is asking you to treat switching costs as permanent.

Who Signs Off: Budget Holder, DPO, Academic Lead

Procurement for AI tools usually stalls because three different people are evaluating three different things without a shared document. Put them on one page before the vendor call, not after.

Role	Owns	RFP questions to lead on
Budget holder (bursar / finance director)	Per-seat price, usage cap, overage terms, worst-case annual cost	Questions 1-7
Data protection officer / compliance lead	Data location, DPA status, deletion guarantees, regulatory alignment	Questions 8-11
Academic / language-centre lead	Pedagogy fit, curriculum alignment, learner outcomes	Contract terms 12-15, plus the separate pedagogy criteria in how institutions should evaluate AI language tools

This guide covers the money and contract layer deliberately, on its own — pedagogy fit is a real, separate evaluation with its own criteria, and conflating the two is how a good pedagogical fit gets waved through on bad contract terms, or a great contract gets rejected because nobody checked whether it actually teaches speaking well.

For reference, LinguaLive's own institutional pricing and cohort licensing terms are published — not gated behind a demo call — at LinguaLive pricing, as an example of the transparency this guide is asking every vendor for.

FAQs

How much does language learning software cost per student?

There's no single reliable market-wide figure, because most AI language vendors don't publish a per-student annual rate at all — that opacity is exactly what this guide is pushing back on. What you can compare are published examples: LinguaLive Pro, for instance, is $79.99/year with a 20-minute daily cap. Always convert any vendor quote to an annual per-student figure before comparing, since the cap changes what that price actually buys.

What should schools ask in an RFP for AI tools?

At minimum, demand four numbers in writing: the exact per-seat price at your enrollment size, the contractual usage cap per student, the overage rate once a student exceeds it, and the data deletion timeline after the contract ends. A vendor that can answer all four in writing has priced the product; one that can't is still pricing you.

Why do AI tools have unpredictable costs?

Real-time AI voice tutoring runs speech recognition, a language model, and speech synthesis for every minute of conversation — a genuine per-minute infrastructure cost. Traditional software has near-zero marginal cost per additional user-hour, so vendors can safely sell flat, unlimited access. AI vendors selling "unlimited" at a flat price are betting your students won't actually use it much.

What is a usage cap in an AI software contract?

A usage cap is a contractual ceiling — usually daily or monthly minutes, sessions, or messages — beyond which access stops, slows, or triggers a defined overage charge. It's what turns an AI vendor's variable, usage-based cost into a fixed number a school can put in a budget line and defend to a finance committee.

Should schools buy unlimited AI plans?

Generally no, unless the vendor can show exactly what "unlimited" means technically — a specific fair-use ceiling, defined overage handling, and a published price that survives heavy classroom-wide adoption. Without those specifics, "unlimited" is marketing language for an uncapped liability, not a procurement-ready contract term.

If you're evaluating AI language software for a school, language centre, or district, start with the numbers in this guide before you take a demo. See LinguaLive for Education for published per-seat pricing, cohort licensing, and a pilot structure we'll walk your budget holder and DPO through directly — no unlimited plan, no undisclosed cap, no surprise invoice at renewal.