By: Audrey Denise B. Cachuela
A contract that looks safe at signing rarely stays that way. The subscription fee fits the budget, the rollout timeline looks doable, and the vendor promises a fast return. What rarely gets priced is what happens the day the company wants to leave. This is vendor lock-in, and it seldom shows up in the business case. Rajeev Jaswal, founder and CEO of Argo Intelligence, has spent his career watching companies pay for it for years after signing.
Few buyers raise the exit question during the pitch. Sales reps have no reason to, and on day one there is nothing to leave yet. Three years later, the company’s data sits in the platform, hundreds of people are trained on its screens, and the renewal deadline is six weeks out. Lock-in builds one workflow at a time, until leaving costs more than staying.
Jaswal learned this from the buying side, in CIO roles at Rapid7 and Red Hat across 25 years in enterprise IT. “The invoice is the smallest number in the conversation,” he says. “The number that matters is the exit cost.”
In the FinOps Foundation’s 2025 survey of organizations responsible for more than $69 billion in cloud spending, 63% of respondents said they manage AI spending, more than double the 31% that did so a year earlier (Source: FinOps Foundation, 2025). As AI moves from pilots into production, that is where lock-in grows, because most procurement teams still price AI one contract at a time.
The Cost of Vendor Lock-In Sits Outside the Contract
Jaswal once ran the numbers on retiring a platform most employees wanted to replace. The license fee was the smallest line on the page. Leaving would have meant retraining several thousand employees, rebuilding integrations that only one engineer understood, and absorbing weeks of lower output while everyone adjusted. The company renewed at a higher price instead. “That is not a negotiation,” Jaswal says. When leaving costs more than staying, a vendor does not have to work to keep a customer, and that math holds companies in place long after a platform stops serving the business.
Enterprises already feel the pressure. Nearly half of tech professionals, 47%, worry about depending too heavily on the three largest cloud providers, and 59% say their cloud costs went up over the past year (Source: Civo, 2024).
Part of the reason this stays hidden is how the cost gets tracked. Annual contract value sits on one line of the finance system, easy to see and compare. Exit cost is spread across the organization. IT absorbs the cost of rebuilding integrations, HR carries the training load, finance absorbs the productivity dip, and security repeats vendor reviews and rebuilds access controls. Split across five budgets, the full number never appears in one place, and a renewal looks cheaper than a migration even when the platform stopped serving the business years ago.
Regulators have noticed the same pattern. The UK’s Competition and Markets Authority spent close to two years examining cloud infrastructure, including technical switching barriers, licensing practices, committed spend agreements, and egress fees. Its final report recommended prioritizing strategic market status investigations into Amazon Web Services and Microsoft (Source: UK Competition and Markets Authority, 2025). The European Union’s Data Act, in effect since September 12, 2025, targets the same problem by making provider switching easier (Source: European Commission, 2026).
Regulation can remove contractual and technical barriers, but the rest stays with the buyer. Documenting integrations and reducing workflow dependence has to happen before the contract is signed, because after signing, the vendor holds the leverage.
How to Calculate Software Switching Costs and Protect the Exit in Your Contract
Build the exit-cost model around one question: if this vendor shut down tomorrow or doubled its price in three years, what would it take to keep the business running?
Procurement prices the data extraction, including cleaning, validation, storage, and the egress charges buyers often overlook. Technical teams map every connection the system has to internal databases, identity providers, and reporting tools. The human cost is training hours times the fully loaded cost of every employee who uses the system, plus lost productivity while people relearn their jobs. Add parallel operations, since most enterprises run two platforms for months while teams reconcile records by hand. Then add a contingency for the failed transfers and compliance delays that show up in most migrations.
Consider a mid-sized system. A $600,000 annual subscription might be undercut by a $450,000 competing platform, which looks like $150,000 in yearly savings on paper. Technical migration at 4,000 hours and $120 an hour costs $480,000. Training 2,000 employees for four hours each at $60 an hour adds another $480,000. Three months of parallel operation costs $150,000. Outside support, compliance validation, and contingency add $240,000. Total switching cost comes to $1.35 million, which, at $150,000 a year in savings, takes nine years to break even.
These figures are illustrative. The value is in running the exercise, because it forces a buyer to weigh the contract against everything underneath it and shows which dependencies need to shrink before signing. The point is not to avoid switching. It is to know the number before you sign, so the next contract costs less to leave than the last one.
Once that number exists, put it in the contract. Treat data portability as a financial term: specify the export format, the timeline, and what the vendor can charge, and state whether custom workflows, prompts, configurations, and audit records transfer with the data or must be rebuilt. Every integration needs a named owner and documentation, because an undocumented integration is migration debt that comes due at renewal.
Review volume discounts tied to growing usage targets. They look generous at signing and become restrictive when a company wants to scale down or split workloads. That risk is higher with AI vendor pricing, where workloads run across public cloud, SaaS tools, and private infrastructure. Among companies already investing in AI, 97% say that investment spans multiple infrastructure environments (Source: FinOps Foundation, 2025).
Run an exit test alongside the usual demos and security reviews. Ask the vendor to walk through a complete data export, have engineers confirm another platform can use the output, and get a written estimate of termination and migration fees. Then divide projected switching cost by annual contract value. A $500,000 platform that costs $2 million to leave has a four-to-one ratio. A high ratio argues for a shorter initial term and a phased rollout.
AI Vendor Pricing Adds a New Kind of Lock-In
That calculation works for conventional software, where the costs sit in contracts, data, and integrations. AI adds model behavior, prompt libraries, retrieval pipelines, and automated actions that reach into a company’s own systems. Two platforms can both accept natural-language instructions and behave very differently once real work runs through them. The underlying model, the data it retrieves, and how it interprets an instruction are all vendor choices behind a similar chat interface. A company that built its approval workflows around one model’s risk assessments will find that a replacement flags a different set of warnings, and every control built on top has to be retested and recalibrated.
The risk rises when a system takes action. A platform that drafts summaries is easier to replace than one that approves workflows, updates records, or moves money. Replacing the second kind means rebuilding the model interaction, the audit trail, the approval chain, and every permission that lets the system act on its own.
Argo is one example of building around this. Its architecture lets customers use the models that ship with the platform, bring their own, or run the system on infrastructure they already control. The approach is designed to address the model-dependency problem described above, so the underlying model can be changed without a full rebuild of the workflow. Buyers can hold any vendor to that standard by requiring a clean separation between business processes and vendor components, keeping their own copies of prompts and audit records, and testing a backup provider before the dependency becomes too expensive to unwind.
Price Vendor Lock-In Before It Becomes One
The best time to calculate switching costs is before employees build habits around a system and before critical data accumulates in it. Exit-cost analysis belongs on the same calendar as budget reviews and security audits, with a fresh review at every renewal. A platform that was cheap to leave in year one can be expensive to leave by year three as more of the business depends on it.
Before signing the next enterprise software or AI contract, bring procurement, finance, and technology leaders together to price vendor lock-in under at least two realistic exit scenarios. Make model choice, data portability, and exit costs explicit evaluation criteria. The companies best positioned in a renewal negotiation are the ones that priced the exit before they signed.
