As artificial intelligence (AI) takes on larger chunks of software work, more information technology (IT) services firms are working out new ways to charge for it. From AI tokens and usage-based rate cards to subscriptions and prepaid credits, large and mid-sized outsourcers are moving away from the traditional bill-by-the-hour model for work done by humans, as clients seek more predictable costs and outcome-based pricing for AI-led delivery.
Six months after two large IT services companies devised new billing structures for AI work, more large and mid-sized firms are adopting similar models, signalling that AI-linked pricing is becoming mainstream. EPAM Systems is factoring AI token costs into deal structures, while Tech Mahindra, India's fifth largest firm in the sector, is exploring pricing linked to token consumption.
Cognizant Technology Solutions Corp has now introduced a rate card for AI usage and Coforge Ltd launched a fixed monthly subscription for AI use. On the other hand, LTM Ltd (earlier LTIMindtree) allows clients to purchase a fixed number of units that are deducted each time an AI tool performs a task.
These new ways of charging clients for AI work come against the backdrop of advancements in technology. AI tools, once limited to automating tasks such as coding, can now handle entire processes such as creating marketing campaigns and predicting weather.
Nasdaq-listed Cognizant is the latest IT outsourcer to track AI usage through token consumption and base its rate cards on it. Tokens are small units of texts—including words, numbers and even punctuation marks as part of a prompt—that an AI tool can read. Companies pay for these tokens either directly to AI companies like Anthropic and OpenAI, or to their tech vendors who then pay these AI companies.
“We're beginning to see the emergence of AI-infused rate cards, where pricing reflects a blended model of human effort and digital effort. With several clients, we're exposing tokenized rate cards that prices work along a continuum from fully human-led discovery to hybrid to increasingly autonomous agentic delivery,” said Ravi Kumar, chief executive officer of Cognizant at the company’s post-earnings analyst call on 29 April 2026.
Cognizant is charging clients based on four metrics: work done by AI and verified by humans and vice-versa, work that is completely automated, and work done only by AI tools without any human intervention.
At least one of the analysts Mint spoke to said that AI tokens are a cost for companies and complex tasks require them to pay a higher amount for more tokens.
"Due to uncertain macro conditions, clients are moving away from paying AI companies and IT vendors separately. Instead, they want a single bundled deal where the IT vendor takes responsibility for the entire delivery, including managing AI model usage, within a predictable cost structure,” said Sushovon Nayak, lead IT analyst at Anand Rathi Institutional Equities.
“This means the IT vendor now has to decide which AI models to use for which tasks, how to keep token consumption efficient, and how to deliver the outcome within the agreed terms,” Nayak added.
And then there is Noida-based Coforge, which is selling monthly subscriptions of AI tools, offering clients a buffet of over 130 pre-built AI agents. Clients are given the option of paying a monthly fee for a package of agents and senior executives to oversee the work done by those agents. Coforge terms these packages as ‘AI Mod Squads.’
“With our AI Mod Squads, we are giving clients something the market has not offered before: the freedom to compose their own AI-powered delivery team, backed by domain intelligence built over three decades,” said Sudhir Singh, CEO of Coforge, in a company statement on 8 April.
The pricing of these AI teams is determined by the number, complexity and autonomy of the agents deployed, which range from code-writing bots to agents that handle multiple tasks.
“This model provides enterprises with the flexibility to pay for only what they consume, ensuring high transparency and cost predictability for annual engineering budgets,” Coforge said in its press release, adding that these AI packages are available for deployment.
While Cognizant and Coforge are selling packages for fixed AI uses, LTM plans to sell credits to clients that get deducted upon AI usage, similar to tokens given in arcade stores.
As part of this pricing model, clients can buy AI tools specializing in certain functions, including cloud migration and software modernization from a fixed set of credits. LTM is expected to roll out this feature to select clients in the current quarter ending June.
“BlueVerse credit is the new commercial model of LTM in the new agentic AI era. Clients will sign up for credit as subscription services, and it is used to consume value of work in agentic engineering or linked to outcome-based KPIs for managed services,” said Venu Lambu, CEO of LTM, in an emailed response to Mint's queries. "It will delink the effort-based pricing model to outcome or value of work delivered."
As of now, Cognizant, Coforge and LTM do not detail their AI revenues.
Cognizant, which follows a January-December financial calendar, ended last year with $21.1 billion in revenue, up 7% on a yearly basis. Homegrown LTM, the sixth largest IT services firm in India, and Coforge, the seventh largest, ended April-March 2026 with $4.76 billion and $1.87 billion in revenue, up 6% and 29%, respectively.
The rest of the country's largest IT services companies have not specified ways by which they charge clients for AI work.