APAC clients more open to outcome-based AI deals than the West, says NTT DATA
NTT DATA’s global AI strategy head says Asia Pacific buyers want proof of value before they spend, as the company tries. to price its “services as software” offerings by outcome rather than headcount.
Businesses in Asia Pacific are more willing than those in North America or Europe to pay for AI services only when the work delivers results, according to Nitin Bajaj, global head of AI strategy and product at NTT DATA.
Speaking to CRN Asia at the NTT R&D Forum 2026 in Tokyo, Bajaj said these arrangements are central to what NTT DATA calls “services as software”. Under this model, AI agents carry out work that would once have been billed by the number of people or hours assigned to it.
“In the services as software model, what we’re trying to do is get away from a headcount model and into one that is very much focused on outcomes,” he said
Three ways to charge for AI work
Bajaj said NTT DATA uses different pricing models depending on the kind of work involved. The first is consumption-based pricing, which suits high-volume tasks where each transaction can be counted. In an insurance company, for example, claims could be grouped as simple, medium or complex, with the client charged per claim.
The second ties fees to a measured result. If a client wants fewer system outages, NTT DATA could charge according to whether uptime rises from 90% to 98.5%, he said.
The third is a revenue or profit share, used when the work is meant to bring in new business. Bajaj gave the example of an outbound sales call centre for a telecommunications company. If the service signs up more customers than an agreed baseline, NTT DATA would take a percentage of the resulting profit.
He said outcome-based contracts are the hardest to run, because the provider can control its own process but not everything in the client’s business that affects the final result.
“You might have done everything, or the services software might have done everything, to improve process efficiency, and you’re seeing all the metrics that you can control, but you might not be seeing the end-to-end business impact because there might be confounders in the client’s world,” he said.
Why APAC clients are more open to outcome-based AI deals
Bajaj said projects in Asia Pacific were “very, very interesting” because of the region’s appetite for these arrangements, which he said had grown in the past year or two.
“Especially in the AI space, we’re seeing a lot more appetite for outcome-based models, or even revenue-sharing models, in APAC clients as compared to North American and European clients,” he said.
He gave two reasons. The first is the mix of AI models available. In North America, he said, frontier AI labs, the developers of the largest and most advanced models, heavily dominate the market and skew its dynamics. In Asia Pacific, clients choose from frontier models, privately developed open-weight models and Chinese open-weight models. Open-weight models are those whose trained parameters are published, so companies can run and adapt them on their own systems.
“That mix is much more rich and fluid,” he said.
The second reason is how clients in the region fund technology. Bajaj said capital structures and access to capital are shaping how Asia Pacific clients engage, with more of them asking providers to prove value first and tying spending to that value.
From pilots to paid projects
Bajaj told the media roundtable earlier the same day that NTT DATA ran more than 600 AI pilot projects in 2024. He said pilots on their own were almost always “anti-climactic” because many did not produce measurable business results. He cited market research estimates that 80% of AI projects fail to deliver, adding that NTT DATA does “a little bit better than that”. He also cited estimates that 30% to 50% of AI projects run into cost overruns.
He told the roundtable that these failures usually come from the data underneath the AI rather than the AI model itself. The data is often fragmented across older systems that are hard to connect, and it is often unclear who owns it.
In the interview, Bajaj said the 2024 pilots led to 1,350 paid projects last year, even though only a minority of the individual pilots went live.
“I’d say probably like 20% may have gone into production for those specific use cases, but it opens the field to a lot more use cases,” he said. He added that the pilots helped clients get to know and trust NTT DATA, which led to further work.
He said Asia Pacific made up “probably 15 to 20%” of the pilot volume in 2024 and was “ramping up very fast.”
Picking models on cost and performance
Asked which AI models NTT DATA offers when an Asian bank or government asks for sovereign AI, meaning AI systems that keep data and control within a country or organisation, Bajaj said “the answer is it depends.”
When NTT DATA builds systems in a client’s own environment, the choice combines NTT DATA’s technical advice with the client’s own preferences. A bank that is heavily invested in Google Cloud, for example, would likely focus on Google’s frontier models, he said.
For its own services-as-software products, NTT DATA mixes frontier, open-weight and in-house models, choosing for each task based on cost and performance.
Bajaj gave the example of intelligent document processing, in which AI reads and extracts information from documents such as insurance claim forms that were previously scanned and coded by hand. He said the industry benchmark for a generative AI document-processing agent was about 43 cents. NTT DATA AIVista, an AI firm the company founded in Silicon Valley, has achieved the same outcome for about 17 cents by routing each step to the most suitable model.
NTT DATA does not sell AIVista as a platform to clients. Bajaj said it is used internally to build NTT DATA’s services-as-software products, because a platform model, where different parties build on it, makes it harder to guarantee results.
“If you bake it end to end into a full software stack, you’re able to guarantee every single variable that might impact the outcome,” he said.
Built on the hyperscalers
Bajaj said NTT DATA’s services-as-software offerings run end to end on hyperscalers, the large cloud providers such as Amazon Web Services (AWS). The company works with each one in industries where that provider has the most influence.
“In banking we have AWS as an outsized influence, especially in the APAC market,” he said. NTT DATA goes to market with AWS on banking services, which in turn drives consumption of AWS cloud services, he added.
At the roundtable, Bajaj said NTT DATA was also expanding its partnership with Google Cloud and expected to have more than 5,000 engineers certified on Gemini Enterprise, Google’s AI platform for businesses, within the next few months. He also cited a partnership with data platform company Databricks.