Migration runs in stages.
Your current platform keeps operating until the replacement is proven in production, rolled out gradually with data kept in sync the whole way, so the cutover happens as a series of small, reversible steps.
The per-seat model was built for humans in chairs. AI agents don't sit in chairs.
Every seat an agent replaces is pure margin, unless your vendor keeps charging you for it.
The Double Ransom
They charge for the seats. Then they charge for the AI that replaces the seats.
When AI handles 80% of your tier-1 support, the vendor still charges for all 200 seats. Then one major SaaS vendor layers a $550/user/month AI seat upgrade, mandatory data-layer infrastructure at $65K–$175K/year, and per-conversation or per-action usage fees on top of the enterprise license you already pay. Another restructured its AI pricing into three paid tiers with metered token pools. You're paying for humans and the machines that replace them, simultaneously. That's the double ransom.
| What Your Vendor Charges | What We Build | |
|---|---|---|
| User access | $175–$500/user/mo | Access for your whole team |
| AI seat upgrade | $550/user/mo (on top of enterprise tier) | Agents native to the system |
| Required agent infrastructure | $65K–$175K/yr data-layer tier | Included, not a separate tier |
| Conversation / action fees | $2/conv or $0.10/action | Agents run as much as the work needs |
| Agent architecture ceiling | 15 topics × 15 actions per topic | Topics and actions scale with the spec |
| Data access | Other systems only via their paid, metered data tier | Reads across the systems you connect |
| Platform limits | Governor limits, API caps | Sized to your workload |
A $550/user/month AI seat upgrade times 500 users equals $3.3M per year, before data-layer infrastructure, before conversation fees, before credit consumption. The same agent capability on a purpose-built system from us carries no per-seat ransom and no per-conversation metering. And the agent reads across the systems your business runs on, your CRM, ERP, support records, documents, and data warehouse, without a separate metered data tier to reach across them.
This is not one vendor's quirk. A second major enterprise SaaS vendor's April 2026 API policy formalized the agent-toll model at the contract layer: third-party AI agents acting on its data must now route through its own AI runtime, where access is metered. Whatever your incumbent has said about openness, the binding text is the policy.
After Migration
The subscription is just the entry fee.
A SaaS license buys access. Making it do your job is a separate budget: an implementation partner to stand it up, administrators to keep it configured, integration work to connect it, and a new services bill every time the business moves. That layer routinely costs more than the license itself. Our MSA folds it into one relationship: we build the system, host and operate it, and keep developing it on your direction, under one contract with defined SLAs.
InfrastructureSecurityComplianceIncident responseImplementationAdministrationContinuous development
Set this against your true SaaS cost, the license plus the implementation and admin budget around it. Our MSA replaces that whole stack with one relationship that includes active development at a 5-day change cadence, not just access. The fee is a function of the system's scope and service level, confirmed in the intake.
The Per-Seat Math
What 500 seats actually cost over five years.
Enterprise SaaS pricing is designed to grow faster than your headcount. Escalators, add-on SKUs, AI surcharges, and platform fees compound every renewal cycle. Here's the math your vendor hopes you won't do.
| Scenario | Year 1 Cost | 5-Year Total |
|---|---|---|
| Major SaaS platform (moderate add-ons) | $1.2M – $1.5M | $6M – $7.5M |
| Major SaaS platform (extensive add-ons) | $1.5M – $2M | $7.5M – $10M |
| Custom-built replacement (conservative) | $675K – $1.1M | $1.3M – $2.1M |
| Custom-built replacement (comprehensive) | $875K – $1.6M | $1.9M – $3.6M |
These ranges are illustrative for a deployment this size; your exact numbers come out of the Assessment. The vendor figures assume a mid-market deployment of one of the largest SaaS platforms, with other major-vendor deployments typically higher. The custom build covers AI-assisted development, infrastructure, and ongoing operation under the MSA.
They were quoted $54M. They built it for $2.64M.
Alberta's Ministry of Infrastructure killed a $54M four-year vendor procurement and built both replacement systems with a small internal team and modern AI tooling. $858K spent, 643 active users today, $2.64M projected to fully complete both. Roughly 95% less than the vendor proposal. The math isn't unique to Alberta. It's what AI-augmented engineering does to enterprise software economics.
Published April 2026 by Nate Glubish, Alberta Minister of Technology and Innovation.
The Return
Stop renting access. Start funding an asset.
Both models carry an ongoing cost; the difference is what the money buys. Vendor spend is rent that climbs at every renewal and still leaves you unable to change the platform on your own terms. Our MSA keeps the system being built for you, customized and extended on your direction at a 5-day cadence, for a fraction of what the vendor escalation costs. On the illustrative figures above, the gap opens in Year 1 and widens every year after. That difference is capability you direct, where the renewal was capital you simply lost.
The Risk Question
Lower cost, and lower risk.
The system stays current.
We operate and improve it under contract at a 5-day change cadence, so it evolves with your business instead of aging toward the next replatform.
Your position stays portable.
Your data is contractually yours and exportable in standard formats on demand; on exit, the specification and a managed transition are deliverables. The relationship holds on results.
Build the case to leave.
70–120 hours. 2–3 weeks. You get a complete migration business case: what it costs to leave, what it costs to stay, and whether it's viable. If it's not, we tell you.
A demo-able artifact every day, working software in your hands every week.