KFC China’s post-order upselling feature, which prompts personalised add-on recommendations when customers check-out via the app or at in-store kiosks, has unlocked hundreds of millions of renminbi in revenue. It is underpinned by a loyalty ecosystem encompassing more than 1,200 behavioural tags linked to over 590 million members.

Earlier this year parent company Yum China introduced an artificial intelligence (AI)-powered ordering assistant called Smart K to the app, promising greater convenience and customisation. It has been adopted by more than 2 million members, especially breakfast and coffee regulars.

Primavera Capital Group, an investor in Yum China, divides the AI impact on the consumer sector into two distinct streams. Digital consumption-based businesses involving music, video, and software are expected to struggle as AI-generated content floods the market. Physical consumption, however, should be a net beneficiary because – like KFC China – it is rooted in tangible, human-centric experiences.

“Fundamental needs for tasty, safety, and healthy food have nothing to do with AI. No matter how powerful AI becomes, it cannot replace the satisfaction of a human enjoying something,” said William Wang, a founding partner at Primavera. “AI frees humans from execution, giving consumers the headspace to judge, appreciate, and enjoy, and allowing companies to focus on crafting the product.”

In this context, AI is an enabler rather than a long-term differentiator, especially as the technology becomes more commoditised over time. But it can still be enormously impactful sitting within applications, reshaping how consumer value is created, delivered, and captured.

Private equity investors in Asia should be highly attuned to these dynamics given the region is arguably at the bleeding edge of adoption. Market research firm Ipsos found that, while consumers in Europe and North America are often nervous about AI, their Asian and Latin American peers are more likely to embrace the benefits of the technology and be excited about products and services using it.

“The percentage of consumers in China, Japan, and India who believe that products and services using AI have changed their lives is higher compared to that in the US and the UK,” added Scott Chen, a managing partner at L Catterton. “More pertinently, consumers in Asia trust AI to a greater degree compared to those in the West. AI is therefore likely to be adopted more quickly here.”

Sensible screening

There are geographical nuances to all this that must be threaded into due diligence. China is a leader in e-commerce penetration and data accumulation, making AI tools more immediately effective. Southeast Asia is unburdened by legacy systems, so can leapfrog traditional digitalisation and adopt AI from the ground up. In Japan, a shrinking population and labour shortages have made AI a survival imperative.

Fundamentals remain the starting point in this process: category tailwinds, addressable market size, differentiated value propositions, sustainable competitive advantages, identifiable value creation opportunities, and favourable deal terms. But those geographical nuances are part of a wider AI module that influences the underwriting of any consumer sector investment.

Bain & Company describes an AI screening system that feeds into a three-tier classification: revolution for companies that may need to reinvent themselves to survive; transformation for situations in which AI poses significant challenges but also substantial upside, provided there is investment and rapid execution; and augmentation, where there can be measurable gains without major disruption.

The firm ran the rule over multiple sectors in its most recent Asia Pacific private equity report. Consumer products and retail clearly skewed towards augmentation – AI can help deliver cost reductions, efficiency improvements, and an enhanced customer experience.

Azusa Owa, a partner at Bain & Company, noted that investors are not demanding that companies have end-to-end digital infrastructure in place at time of acquisition to minimise their risk. Rather, they want to establish whether it is worthwhile building this infrastructure. The presence of proprietary data that can deliver additional upside is a key consideration.

“Some companies attracted premium valuations not because their data was fully leveraged, but because they possessed raw assets – proprietary transaction data linked to customer profiles – that PE firms could unlock post-acquisition. You know who bought what, at what frequency, and when people stopped using it,” Owa explained, while declining to comment on specific transactions.

Infocom Corporation, a Japanese digital content provider best known for manga platform Mecha Comic, fits this profile. As a direct-to-consumer business aimed at women aged 30 and above, the company has vast quantities of data intrinsically linked to an attractive consumer demographic. Blackstone took Infocom private in 2024 at an enterprise value of JPY 275bn (USD 1.7bn).

Reputation is equally critical, Owa added. Companies that have poured resources into traditional digital marketing, often through search engine optimisation (SEO), face a different outlook in the AI era.

“Instead of searching by words, people will search by the experience they’re looking for,” she explained. “If a brand is highly recognised on a trusted word-of-mouth platform, that’s a valuable tangible asset.”

Layers of value 

Global multi-strategy private equity firms have invested heavily in the resources required to answer these questions. Carlyle, for example, claims to have 80 data scientists and technologists embedded in its deal teams, helping assess AI opportunities and risks. Several firms have also forged partnerships with the likes of OpenAI and Anthropic to assist with implementation at the portfolio level.

Smaller managers lean more heavily on third-party expertise, which may come from specialist consultants or even from within the portfolio. Primavera paired Yum China with another of its investees, large language model (LLM) developer Mininglamp Technology in a joint venture intended to create enterprise-grade AI applications for the food and beverage industry.

AI-related value creation efforts are spread across three progressive layers. One comprises the non-negotiables – typically costs and efficiencies – that would be pursued regardless of sector.

McDonald’s China used AI to rewire its backend operations, notably a forecasting system that allows scheduling, ordering, and production management to be adjusted around real consumer demand. It can predict order volumes across approximately 8,000 restaurants “with an accuracy consistently above 90%,” according to Eric Xin, a senior partner at Trustar Capital, which owns McDonald’s China.

Another layer is for initiatives that are fundamentally transformative, driving entirely new capabilities or business models. There are few examples in the consumer sector, given the nascent status of AI.

Investors tend to find most traction in the layer that sits in between the other two: AI use cases that enhance existing operations. They encompass personalization of the consumer experience; marketing that evolves from SEO into answer engine optimization (AEO), AI optimization (AIO), and generative engine optimisation (GEO); and accelerated new product development through rapid, low-cost testing

Yum China’s AI-powered ordering assistant demonstrates personalisation of customer experience, while the company upgraded its marketing to GEO by feeding data from individual profiles into an engine that surfaces relevant ads as soon as the ap is opened or a customer walks past a store. Every subsequent interaction – click, buy, ignore – is logged and used to refine future targeting.

Primavera’s Wang described it as a transition from “people looking for products” to a new model where “products understand people.”

L Catterton-backed personal care brand Stenders reconfigured its marketing by swapping human livestreamers for AI-generated replacements during low-traffic periods on its e-commerce platforms. The company fine-tuned the look and voice of AI streamers, and modified their scripts, based on real-time sales data. This reduced costs while boosting sales, according to L Catterton’s Chen.

In the same vein, Traya, a direct-to-consumer hair loss treatment start-up in India, used human coaches for customer onboarding but handed the rest of the relationship management process to AI agents. “The benefit of the human doesn’t go away. The customer still wants to talk to a coach to understand,” said Vinay Singh, co-founder and a partner at Fireside Ventures, an investor in Traya.

Trustar’s work with TSK Group, a traditionally labour-intensive pest control business in China, showcases accelerated product development. When customers started asking for intelligent monitoring, TSK set about incorporating AI. Within nine months, it had a platform – incorporating hardware, service management and analytics – that was outperforming competitors in pest recognition scenarios.

“At the same time, by building a fully AI-enabled R&D model, TSK saves several millions in annual development cost compared with traditional approaches,” said Trustar’s Xin.

Measure it

Investors recognise the need for measurable outcomes from AI implementation, so they know which initiatives to prioritise. Bain & Company’s Owa observed that cost improvements are easiest to quantify – companies require fewer personnel, unit costs may fall. Impact on revenue is a more complicated proposition with a wider range of metrics and outcomes.

Private equity firms may emphasize specific themes or line items. Trustar, for example, looks at AI impact on business economics across four dimensions: revenue and client adoption, workflow automation and cost efficiency, product transformation, and team augmentation.

At Japan-based NSSK, customer conversion and customer lifetime value are the key revenue data points, while under costs it tracks spending on customer acquisition as well as selling, general and administrative (SG&A) outflows. Speed of decision making and accuracy of forecasting are also in scope, but these are of secondary importance.

“If AI doesn’t show up in EBITDA or cash flow, it is likely not creating real value,” said Jun Tsusaka, founder and CEO of NSSK. “AI is not a separate theme, it is embedded in every investment we make. The differentiator is not who uses AI, but who can systematically translate it into measurable operational improvement at scale.”

Fireside is still in the process of developing a measurement framework. It captures a range of data – from customer retention rates to fixed cost reduction to customer acquisition costs – but Singh is not where he would like to be in terms of specific benchmarks that demonstrate the value generated per LLM token consumed.

“Right now, I think we’re still in the early days,” he explained. “We’re at the point where encouragement is not so much about saying ‘Use AI only as much as required’ as ‘Use as much AI as required but impact these metrics.’”

Agent-to-agent

Fireside’s approach is to some extent shaped by its investment stage. As a venture capital firm, it is targeting less mature companies than Trustar and NSSK – on one hand, targets have fewer resources to throw at AI, on the other, they may not be burdened by legacy systems.

But the Fireside portfolio is still populated by traditional brands across beauty and personal care, food and beverage, and health and wellness. In this sense, Traya is a typical example of a company using AI to enable a consumer experience; AI is not the defining characteristic of the product.

Only a handful of VCs in Asia have exposure to AI-native consumer electronics. Most of this investment activity emanates from China where Even Realities and OdyssLife have gained traction with smart glasses and smart necklaces, respectively. Backers include Cyanhill Capital, HSG, and Monolith Management – but these investments sit in large, diversified portfolios.

Chen of L Catterton believes that the concept of AI-native consumer brands requires considerable scrutiny. What some may call as “AI-native brands” today were referred to as “consumer technology brands” not long ago, he observed.

VC investors are tracking an anticipated evolution from AI tools to agent-to-agent commerce, which could create more opportunities. A personal AI agent learns the consumer’s preferences, proactively negotiates with company agents, and presents a curated selection of branded products. At the same time, the brand’s AI agent proposes deals based on real-time inventory monitoring.

“Before all that, you need to have the basic agent-level infrastructure built out. Agent-focused search infrastructure, payments capability, inference costs coming down. When it gets built and is trusted, you have an inflexion point,” said Pinn Lawjindakul, a partner and advisor of Lightspeed Venture Partners.

Early signs are already emerging. McDonald’s China is working with electric car maker Nio on voice ordering in smart cockpits, whereby customers state what they want and the system places an order and navigates to the nearest restaurant.

The next step is moving from AI-ready to AI-native. McDonald’s China would turn core functions – ordering, inventory management, scheduling – into standardised building blocks, so that AI-enabled systems can execute complex tasks beyond chat-based support and do so at scale.

Fireside encourages portfolio companies to run small, low-risk experiments with emerging technologies to stay ahead of the curve. The firm learned this lesson when quick commerce platforms like Blinkit first emerged. It invited founder Albinder Dhindsa to workshop with 20 portfolio companies. Only four signed up initially; the rest dismissed it as irrelevant.

“You should invest 95% of your capital in the business of today and 5% you can invest in being fit for the future,” Singh said. “Assume your experiments will go to naught. But what if one of them works, you’re the first mover, and best placed to do a land grab?”