AI software and services are entering a phase of “pandemonium” in licensing and pricing, according to Gartner VP analyst Jo Liversidge. Speaking at Gartner’s Symposium in Australia, Liversidge highlighted the lack of consistency and transparency among vendors, warning enterprises to closely examine contracts to avoid unexpected risks and escalating costs.
Contracts often lack clear AI-specific clauses. Liversidge noted that AWS only mentions AI in clause 50 of its supplementary terms, while Adobe initially required customers to accept liability for copyright infringements caused by its AI tools—a stance it later reversed following backlash. She argued buyers will need to pressure vendors to adopt responsible AI principles and comply with standards such as ISO 42001, pointing out that just 1% of vendors currently embed such provisions.
Pricing models are equally inconsistent:
- Credit- and token-based schemes: Some vendors charge credits up front, applying shifting “multipliers” that increase usage costs. Others sell tokens but fail to clarify that output tokens cost more than input ones.
- Inconsistent pricing within platforms: Salesforce’s Agentforce platform, for example, launched with one pricing model, later added another, and now allows customers to choose between them.
- Hidden costs and forced upgrades: Slack cited for retiring a product in favor of an AI-infused version that increased seat prices by 40%.
Buyers also face challenges monitoring consumption. Traditional software monitoring tools cannot track generative AI’s complex pricing, and vendors often delay releasing calculators until after contracts are signed.
Looking ahead, Gartner expects the next major licensing hurdle will be the cost structures for AI agents, with at least four competing models already emerging. Enterprises are urged to demand transparency and safeguards to avoid costly surprises.
Source:
https://www.theregister.com/2025/09/10/ai_software_licensing_immature/
