Information technology (IT) services companies are beginning to change their billing strategies to account for artificial intelligence (AI) work as the new technology increasingly takes over routine tasks.
For instance, two large information technology (IT) services companies – US-based EPAM Systems Inc, and homegrown Tech Mahindra Ltd—are looking to include artificial intelligence costs in billing rates and charge clients based on AI usage, instead of the traditional model of billing for engineers’ time, as automation tools disrupt the industry.
At its investor day on 12 March, EPAM Systems said the company plans to factor the prices of artificial intelligence (AI) tokens, which clients currently pay for separately, into its billing model.
“Going forward, as we are migrating away or transitioning away from time and materials to more advanced capabilities or more advanced contracting models, we will be seeing that it’s ( tokens) going to be part of our commercial model. We’re going to factor in the price of the tokens into our model itself, on top of it,” said Balazs Fejes, chief executive officer (CEO) of EPAM, during the company’s analyst day.
AI tokens are small units of texts that an AI tool can read. These include words, numbers, and even punctuation marks that are part of a prompt.
Fejes said that clients traditionally pay for these tokens and while EPAM works as a service integrator, this could change going forward.
“It’s a work in progress, how we’re going to charge our clients because all the tokens’ price is very volatile, so it’s very difficult to figure out how to price it in at this point of time but we expect that once we are more in the fixed price or more advanced models, the compute will be included in our price,” added Fejes.
Fejes’ comments come as the technology’s pace of evolution raises questions about the relevance of IT outsourcers. Shares of providers have fallen by more than 15% since the start of the year as AI tools increasingly eat into their core work of software development and maintenance, coding and customer support.
EPAM was one of the fastest-growing companies last calendar year. It ended 2025 with $5.46 billion in revenue, up 15.46% on a yearly basis. This makes the company almost the size of Tech Mahindra Ltd, which is India’s fifth-largest IT services company.
A fifth of EPAM’s workforce, or 12,200 employees, is based out of India, which also became the company’s largest talent hub in 2024.
EPAM reported $105 million in AI-native revenue during the October-December 2025 period and expects $600 million in AI revenue by year-end.
EPAM is not the first to shed light on billing AI work. Just two months ago, Tech Mahindra CEO Mohit Joshi spoke about the company’s methods to bill AI-led work.
“Essentially you have a pricing model where you are partly charging for human labour, and that is very clearly called out, and partly charging for digital labour. The problem is that how do we specify digital labour, and what we have said is that the lowest common denominator for digital labour is token consumption. So, we will try and define how many tokens are consumed and have pricing linked to that,” Joshi said during the company’s post-earnings press conference on 16 January.
“So we’re giving a lot of thought to how we can change our pricing model for clients to clearly charge for digital labour distinct from human , but I feel that adoption will obviously take time,” said Joshi.
The Pune-based company ended the last fiscal year with $6.26 billion in revenue, down 0.21% on a yearly basis. Unlike EPAM, the company has yet to share revenue and orders from AI.
The rise of automation tools has raised questions about existing billing models in the IT industry. Traditionally, IT services companies are billed on the hours spent by humans on completing certain IT tasks. Automation tools are forcing a shift to outcome-based models where IT services companies charge their clients on the outcome achieved.
EPAM follows the January-December financial calendar, while Indian IT services companies follow April-March accounting year.
Experts said that the billing models of Indian IT had to include platform subscription, and fees tied to business results.
“The current time-and-materials pricing model is poorly suited to AI because productivity gains can be exponential. If a small team supported by AI can deliver the work of fifty engineers, by the hour quickly becomes misaligned with value. Over time, the market will shift toward hybrid models that combine platform subscription, consumption pricing for AI workloads, and outcome-based fees tied to business results,” said Phil Fersht, chief executive of HFS Research.